identifier
stringlengths 7
18
| space
stringclasses 4
values | uid
stringlengths 1
6
| arch_str
stringlengths 1
32
| input
stringlengths 8.51k
461k
| target_metric
stringclasses 1
value | val_accuracy
float64 0
95.1
| flops
float64 31.1M
14.7B
| params
float64 227k
50M
| metadata
stringlengths 0
1.46k
| metainformation
stringclasses 1
value |
|---|---|---|---|---|---|---|---|---|---|---|
NASBench101_285736
|
NASBench101
|
285736
|
acf5f6b39edaf899624e9109a697f1e7
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_779[FLOAT, 128x3x3x3]
%onnx::Conv_780[FLOAT, 128]
%onnx::Conv_782[FLOAT, 128x128x1x1]
%onnx::Conv_785[FLOAT, 128x128x3x3]
%onnx::Conv_788[FLOAT, 128x128x3x3]
%onnx::Conv_791[FLOAT, 128x128x1x1]
%onnx::Conv_794[FLOAT, 128x128x1x1]
%onnx::Conv_797[FLOAT, 128x128x1x1]
%onnx::Conv_800[FLOAT, 128x128x3x3]
%onnx::Conv_803[FLOAT, 128x128x3x3]
%onnx::Conv_806[FLOAT, 128x128x1x1]
%onnx::Conv_809[FLOAT, 128x128x1x1]
%onnx::Conv_812[FLOAT, 128x128x1x1]
%onnx::Conv_815[FLOAT, 128x128x3x3]
%onnx::Conv_818[FLOAT, 128x128x3x3]
%onnx::Conv_821[FLOAT, 128x128x1x1]
%onnx::Conv_824[FLOAT, 128x128x1x1]
%onnx::Conv_827[FLOAT, 256x128x1x1]
%onnx::Conv_828[FLOAT, 256]
%onnx::Conv_830[FLOAT, 256x256x3x3]
%onnx::Conv_833[FLOAT, 256x256x3x3]
%onnx::Conv_836[FLOAT, 256x256x1x1]
%onnx::Conv_839[FLOAT, 256x128x1x1]
%onnx::Conv_842[FLOAT, 256x256x1x1]
%onnx::Conv_845[FLOAT, 256x256x3x3]
%onnx::Conv_848[FLOAT, 256x256x3x3]
%onnx::Conv_851[FLOAT, 256x256x1x1]
%onnx::Conv_854[FLOAT, 256x256x1x1]
%onnx::Conv_857[FLOAT, 256x256x1x1]
%onnx::Conv_860[FLOAT, 256x256x3x3]
%onnx::Conv_863[FLOAT, 256x256x3x3]
%onnx::Conv_866[FLOAT, 256x256x1x1]
%onnx::Conv_869[FLOAT, 256x256x1x1]
%onnx::Conv_872[FLOAT, 512x256x1x1]
%onnx::Conv_873[FLOAT, 512]
%onnx::Conv_875[FLOAT, 512x512x3x3]
%onnx::Conv_878[FLOAT, 512x512x3x3]
%onnx::Conv_881[FLOAT, 512x512x1x1]
%onnx::Conv_884[FLOAT, 512x256x1x1]
%onnx::Conv_887[FLOAT, 512x512x1x1]
%onnx::Conv_890[FLOAT, 512x512x3x3]
%onnx::Conv_893[FLOAT, 512x512x3x3]
%onnx::Conv_896[FLOAT, 512x512x1x1]
%onnx::Conv_899[FLOAT, 512x512x1x1]
%onnx::Conv_902[FLOAT, 512x512x1x1]
%onnx::Conv_905[FLOAT, 512x512x3x3]
%onnx::Conv_908[FLOAT, 512x512x3x3]
%onnx::Conv_911[FLOAT, 512x512x1x1]
%onnx::Conv_914[FLOAT, 512x512x1x1]
) {
%onnx::Conv_915 = Identity(%onnx::Conv_873)
%onnx::Conv_912 = Identity(%onnx::Conv_873)
%onnx::Conv_909 = Identity(%onnx::Conv_873)
%onnx::Conv_906 = Identity(%onnx::Conv_873)
%onnx::Conv_903 = Identity(%onnx::Conv_873)
%onnx::Conv_900 = Identity(%onnx::Conv_873)
%onnx::Conv_897 = Identity(%onnx::Conv_873)
%onnx::Conv_894 = Identity(%onnx::Conv_873)
%onnx::Conv_891 = Identity(%onnx::Conv_873)
%onnx::Conv_888 = Identity(%onnx::Conv_873)
%onnx::Conv_885 = Identity(%onnx::Conv_873)
%onnx::Conv_882 = Identity(%onnx::Conv_873)
%onnx::Conv_879 = Identity(%onnx::Conv_873)
%onnx::Conv_876 = Identity(%onnx::Conv_873)
%onnx::Conv_870 = Identity(%onnx::Conv_828)
%onnx::Conv_867 = Identity(%onnx::Conv_828)
%onnx::Conv_864 = Identity(%onnx::Conv_828)
%onnx::Conv_861 = Identity(%onnx::Conv_828)
%onnx::Conv_858 = Identity(%onnx::Conv_828)
%onnx::Conv_855 = Identity(%onnx::Conv_828)
%onnx::Conv_852 = Identity(%onnx::Conv_828)
%onnx::Conv_849 = Identity(%onnx::Conv_828)
%onnx::Conv_846 = Identity(%onnx::Conv_828)
%onnx::Conv_843 = Identity(%onnx::Conv_828)
%onnx::Conv_840 = Identity(%onnx::Conv_828)
%onnx::Conv_837 = Identity(%onnx::Conv_828)
%onnx::Conv_834 = Identity(%onnx::Conv_828)
%onnx::Conv_831 = Identity(%onnx::Conv_828)
%onnx::Conv_825 = Identity(%onnx::Conv_780)
%onnx::Conv_822 = Identity(%onnx::Conv_780)
%onnx::Conv_819 = Identity(%onnx::Conv_780)
%onnx::Conv_816 = Identity(%onnx::Conv_780)
%onnx::Conv_813 = Identity(%onnx::Conv_780)
%onnx::Conv_810 = Identity(%onnx::Conv_780)
%onnx::Conv_807 = Identity(%onnx::Conv_780)
%onnx::Conv_804 = Identity(%onnx::Conv_780)
%onnx::Conv_801 = Identity(%onnx::Conv_780)
%onnx::Conv_798 = Identity(%onnx::Conv_780)
%onnx::Conv_795 = Identity(%onnx::Conv_780)
%onnx::Conv_792 = Identity(%onnx::Conv_780)
%onnx::Conv_789 = Identity(%onnx::Conv_780)
%onnx::Conv_786 = Identity(%onnx::Conv_780)
%onnx::Conv_783 = Identity(%onnx::Conv_780)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_779, %onnx::Conv_780)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_782, %onnx::Conv_783)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_785, %onnx::Conv_786)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_788, %onnx::Conv_789)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/Add_3_output_0, %/layers.1/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.1/Add_5_output_0 = Add(%/layers.1/Add_4_output_0, %/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_5_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_6_output_0 = Add(%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_6_output_0, %onnx::Conv_797, %onnx::Conv_798)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_800, %onnx::Conv_801)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_803, %onnx::Conv_804)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/Add_3_output_0, %/layers.2/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.2/Add_5_output_0 = Add(%/layers.2/Add_4_output_0, %/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_5_output_0, %onnx::Conv_806, %onnx::Conv_807)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_6_output_0, %onnx::Conv_809, %onnx::Conv_810)
%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_6_output_0 = Add(%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_6_output_0, %onnx::Conv_812, %onnx::Conv_813)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_815, %onnx::Conv_816)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_818, %onnx::Conv_819)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/Add_3_output_0, %/layers.3/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.3/Add_5_output_0 = Add(%/layers.3/Add_4_output_0, %/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_5_output_0, %onnx::Conv_821, %onnx::Conv_822)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_6_output_0, %onnx::Conv_824, %onnx::Conv_825)
%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_6_output_0 = Add(%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Add_6_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_827, %onnx::Conv_828)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_830, %onnx::Conv_831)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_833, %onnx::Conv_834)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/Add_3_output_0, %/layers.5/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.5/Add_5_output_0 = Add(%/layers.5/Add_4_output_0, %/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_5_output_0, %onnx::Conv_836, %onnx::Conv_837)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_839, %onnx::Conv_840)
%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_6_output_0 = Add(%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_6_output_0, %onnx::Conv_842, %onnx::Conv_843)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_845, %onnx::Conv_846)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_848, %onnx::Conv_849)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/Add_3_output_0, %/layers.6/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.6/Add_5_output_0 = Add(%/layers.6/Add_4_output_0, %/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_5_output_0, %onnx::Conv_851, %onnx::Conv_852)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_6_output_0, %onnx::Conv_854, %onnx::Conv_855)
%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_6_output_0 = Add(%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_6_output_0, %onnx::Conv_857, %onnx::Conv_858)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/Add_3_output_0, %/layers.7/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.7/Add_5_output_0 = Add(%/layers.7/Add_4_output_0, %/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_5_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_6_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_6_output_0 = Add(%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Add_6_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/Add_3_output_0, %/layers.9/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.9/Add_5_output_0 = Add(%/layers.9/Add_4_output_0, %/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_5_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_6_output_0 = Add(%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_6_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/Add_3_output_0, %/layers.10/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.10/Add_5_output_0 = Add(%/layers.10/Add_4_output_0, %/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_5_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_6_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_6_output_0 = Add(%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_6_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_908, %onnx::Conv_909)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/Add_3_output_0, %/layers.11/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.11/Add_5_output_0 = Add(%/layers.11/Add_4_output_0, %/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_5_output_0, %onnx::Conv_911, %onnx::Conv_912)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_6_output_0, %onnx::Conv_914, %onnx::Conv_915)
%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_6_output_0 = Add(%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Add_6_output_0)
%777 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %777
}
|
val_accuracy
| 93.810093
| 6,310,340,608
| 21,384,074
|
{'zcp_epe_nas': 146.4725014790919, 'zcp_fisher': 10.792214393615723, 'zcp_flops': 100965449728.0, 'zcp_grad_norm': 60.5468635559082, 'zcp_grasp': -4.58758544921875, 'zcp_jacov': -16.048882502204414, 'zcp_l2_norm': 1030.8115234375, 'zcp_nwot': 232.24084908137468, 'zcp_params': 21384074.0, 'zcp_plain': 0.135533943772315, 'zcp_snip': 534.9972534179688, 'zcp_synflow': 129.91178911071094, 'zcp_zen': 103.5826416015625, 'zcp_val_accuracy': 0.915665090084075}
| |
NASBench101_120876
|
NASBench101
|
120876
|
4906a1ead66e4bbc3ef43fb019e02c97
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_917[FLOAT, 128x3x3x3]
%onnx::Conv_918[FLOAT, 128]
%onnx::Conv_920[FLOAT, 43x128x1x1]
%onnx::Conv_921[FLOAT, 43]
%onnx::Conv_923[FLOAT, 43x43x3x3]
%onnx::Conv_926[FLOAT, 43x128x1x1]
%onnx::Conv_929[FLOAT, 43x43x1x1]
%onnx::Conv_932[FLOAT, 43x43x3x3]
%onnx::Conv_935[FLOAT, 43x128x1x1]
%onnx::Conv_938[FLOAT, 43x128x1x1]
%onnx::Conv_941[FLOAT, 43x43x3x3]
%onnx::Conv_944[FLOAT, 43x128x1x1]
%onnx::Conv_947[FLOAT, 43x43x1x1]
%onnx::Conv_950[FLOAT, 43x43x3x3]
%onnx::Conv_953[FLOAT, 43x128x1x1]
%onnx::Conv_956[FLOAT, 43x128x1x1]
%onnx::Conv_959[FLOAT, 43x43x3x3]
%onnx::Conv_962[FLOAT, 43x128x1x1]
%onnx::Conv_965[FLOAT, 43x43x1x1]
%onnx::Conv_968[FLOAT, 43x43x3x3]
%onnx::Conv_971[FLOAT, 43x128x1x1]
%onnx::Conv_974[FLOAT, 86x128x1x1]
%onnx::Conv_975[FLOAT, 86]
%onnx::Conv_977[FLOAT, 86x86x3x3]
%onnx::Conv_980[FLOAT, 85x128x1x1]
%onnx::Conv_981[FLOAT, 85]
%onnx::Conv_983[FLOAT, 85x85x1x1]
%onnx::Conv_986[FLOAT, 85x85x3x3]
%onnx::Conv_989[FLOAT, 85x128x1x1]
%onnx::Conv_992[FLOAT, 86x256x1x1]
%onnx::Conv_995[FLOAT, 86x86x3x3]
%onnx::Conv_998[FLOAT, 85x256x1x1]
%onnx::Conv_1001[FLOAT, 85x85x1x1]
%onnx::Conv_1004[FLOAT, 85x85x3x3]
%onnx::Conv_1007[FLOAT, 85x256x1x1]
%onnx::Conv_1010[FLOAT, 86x256x1x1]
%onnx::Conv_1013[FLOAT, 86x86x3x3]
%onnx::Conv_1016[FLOAT, 85x256x1x1]
%onnx::Conv_1019[FLOAT, 85x85x1x1]
%onnx::Conv_1022[FLOAT, 85x85x3x3]
%onnx::Conv_1025[FLOAT, 85x256x1x1]
%onnx::Conv_1028[FLOAT, 171x256x1x1]
%onnx::Conv_1029[FLOAT, 171]
%onnx::Conv_1031[FLOAT, 171x171x3x3]
%onnx::Conv_1034[FLOAT, 171x256x1x1]
%onnx::Conv_1037[FLOAT, 171x171x1x1]
%onnx::Conv_1040[FLOAT, 171x171x3x3]
%onnx::Conv_1043[FLOAT, 171x256x1x1]
%onnx::Conv_1046[FLOAT, 171x512x1x1]
%onnx::Conv_1049[FLOAT, 171x171x3x3]
%onnx::Conv_1052[FLOAT, 171x512x1x1]
%onnx::Conv_1055[FLOAT, 171x171x1x1]
%onnx::Conv_1058[FLOAT, 171x171x3x3]
%onnx::Conv_1061[FLOAT, 171x512x1x1]
%onnx::Conv_1064[FLOAT, 171x512x1x1]
%onnx::Conv_1067[FLOAT, 171x171x3x3]
%onnx::Conv_1070[FLOAT, 171x512x1x1]
%onnx::Conv_1073[FLOAT, 171x171x1x1]
%onnx::Conv_1076[FLOAT, 171x171x3x3]
%onnx::Conv_1079[FLOAT, 171x512x1x1]
) {
%onnx::Conv_1080 = Identity(%onnx::Conv_1029)
%onnx::Conv_1077 = Identity(%onnx::Conv_1029)
%onnx::Conv_1074 = Identity(%onnx::Conv_1029)
%onnx::Conv_1071 = Identity(%onnx::Conv_1029)
%onnx::Conv_1068 = Identity(%onnx::Conv_1029)
%onnx::Conv_1065 = Identity(%onnx::Conv_1029)
%onnx::Conv_1062 = Identity(%onnx::Conv_1029)
%onnx::Conv_1059 = Identity(%onnx::Conv_1029)
%onnx::Conv_1056 = Identity(%onnx::Conv_1029)
%onnx::Conv_1053 = Identity(%onnx::Conv_1029)
%onnx::Conv_1050 = Identity(%onnx::Conv_1029)
%onnx::Conv_1047 = Identity(%onnx::Conv_1029)
%onnx::Conv_1044 = Identity(%onnx::Conv_1029)
%onnx::Conv_1041 = Identity(%onnx::Conv_1029)
%onnx::Conv_1038 = Identity(%onnx::Conv_1029)
%onnx::Conv_1035 = Identity(%onnx::Conv_1029)
%onnx::Conv_1032 = Identity(%onnx::Conv_1029)
%onnx::Conv_1026 = Identity(%onnx::Conv_981)
%onnx::Conv_1023 = Identity(%onnx::Conv_981)
%onnx::Conv_1020 = Identity(%onnx::Conv_981)
%onnx::Conv_1017 = Identity(%onnx::Conv_981)
%onnx::Conv_1014 = Identity(%onnx::Conv_975)
%onnx::Conv_1011 = Identity(%onnx::Conv_975)
%onnx::Conv_1008 = Identity(%onnx::Conv_981)
%onnx::Conv_1005 = Identity(%onnx::Conv_981)
%onnx::Conv_1002 = Identity(%onnx::Conv_981)
%onnx::Conv_999 = Identity(%onnx::Conv_981)
%onnx::Conv_996 = Identity(%onnx::Conv_975)
%onnx::Conv_993 = Identity(%onnx::Conv_975)
%onnx::Conv_990 = Identity(%onnx::Conv_981)
%onnx::Conv_987 = Identity(%onnx::Conv_981)
%onnx::Conv_984 = Identity(%onnx::Conv_981)
%onnx::Conv_978 = Identity(%onnx::Conv_975)
%onnx::Conv_972 = Identity(%onnx::Conv_921)
%onnx::Conv_969 = Identity(%onnx::Conv_921)
%onnx::Conv_966 = Identity(%onnx::Conv_921)
%onnx::Conv_963 = Identity(%onnx::Conv_921)
%onnx::Conv_960 = Identity(%onnx::Conv_921)
%onnx::Conv_957 = Identity(%onnx::Conv_921)
%onnx::Conv_954 = Identity(%onnx::Conv_921)
%onnx::Conv_951 = Identity(%onnx::Conv_921)
%onnx::Conv_948 = Identity(%onnx::Conv_921)
%onnx::Conv_945 = Identity(%onnx::Conv_921)
%onnx::Conv_942 = Identity(%onnx::Conv_921)
%onnx::Conv_939 = Identity(%onnx::Conv_921)
%onnx::Conv_936 = Identity(%onnx::Conv_921)
%onnx::Conv_933 = Identity(%onnx::Conv_921)
%onnx::Conv_930 = Identity(%onnx::Conv_921)
%onnx::Conv_927 = Identity(%onnx::Conv_921)
%onnx::Conv_924 = Identity(%onnx::Conv_921)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_917, %onnx::Conv_918)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_920, %onnx::Conv_921)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_923, %onnx::Conv_924)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_926, %onnx::Conv_927)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_929, %onnx::Conv_930)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_932, %onnx::Conv_933)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_935, %onnx::Conv_936)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/Add_3_output_0, %/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_4_output_0)
%/layers.1/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.1/Slice_output_0 = Slice(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_5_output_0, %/layers.1/Constant_6_output_0, %/layers.1/Constant_4_output_0, %/layers.1/Constant_7_output_0)
%/layers.1/Constant_8_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_5_output_0 = Add(%/layers.1/Slice_output_0, %/layers.1/Constant_8_output_0)
%/layers.1/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_5_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.4/maxpool/MaxPool_output_0, %/layers.1/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_938, %onnx::Conv_939)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_941, %onnx::Conv_942)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_944, %onnx::Conv_945)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_947, %onnx::Conv_948)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_950, %onnx::Conv_951)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_953, %onnx::Conv_954)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/Add_3_output_0, %/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_4_output_0)
%/layers.2/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.2/Slice_output_0 = Slice(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_5_output_0, %/layers.2/Constant_6_output_0, %/layers.2/Constant_4_output_0, %/layers.2/Constant_7_output_0)
%/layers.2/Constant_8_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_5_output_0 = Add(%/layers.2/Slice_output_0, %/layers.2/Constant_8_output_0)
%/layers.2/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_5_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.4/maxpool/MaxPool_output_0, %/layers.2/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_956, %onnx::Conv_957)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_959, %onnx::Conv_960)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_962, %onnx::Conv_963)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_965, %onnx::Conv_966)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_968, %onnx::Conv_969)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/Add_3_output_0, %/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_4_output_0)
%/layers.3/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.3/Slice_output_0 = Slice(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_5_output_0, %/layers.3/Constant_6_output_0, %/layers.3/Constant_4_output_0, %/layers.3/Constant_7_output_0)
%/layers.3/Constant_8_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_5_output_0 = Add(%/layers.3/Slice_output_0, %/layers.3/Constant_8_output_0)
%/layers.3/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_5_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.4/maxpool/MaxPool_output_0, %/layers.3/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_974, %onnx::Conv_975)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_977, %onnx::Conv_978)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/Add_3_output_0, %/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_4_output_0)
%/layers.5/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_5_output_0 = Add(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_4_output_0)
%/layers.5/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_5_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.4/maxpool/MaxPool_output_0, %/layers.5/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/Add_3_output_0, %/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_4_output_0)
%/layers.6/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_5_output_0 = Add(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_4_output_0)
%/layers.6/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_5_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.4/maxpool/MaxPool_output_0, %/layers.6/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_1025, %onnx::Conv_1026)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/Add_3_output_0, %/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_4_output_0)
%/layers.7/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_5_output_0 = Add(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_4_output_0)
%/layers.7/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_5_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.4/maxpool/MaxPool_output_0, %/layers.7/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1028, %onnx::Conv_1029)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_1031, %onnx::Conv_1032)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1034, %onnx::Conv_1035)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_1037, %onnx::Conv_1038)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_1040, %onnx::Conv_1041)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1043, %onnx::Conv_1044)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/Add_3_output_0, %/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_4_output_0)
%/layers.9/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.9/Slice_output_0 = Slice(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_5_output_0, %/layers.9/Constant_6_output_0, %/layers.9/Constant_4_output_0, %/layers.9/Constant_7_output_0)
%/layers.9/Constant_8_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_5_output_0 = Add(%/layers.9/Slice_output_0, %/layers.9/Constant_8_output_0)
%/layers.9/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_5_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.4/maxpool/MaxPool_output_0, %/layers.9/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_1046, %onnx::Conv_1047)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_1049, %onnx::Conv_1050)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_1052, %onnx::Conv_1053)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_1055, %onnx::Conv_1056)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_1058, %onnx::Conv_1059)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_1061, %onnx::Conv_1062)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/Add_3_output_0, %/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_4_output_0)
%/layers.10/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.10/Slice_output_0 = Slice(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_5_output_0, %/layers.10/Constant_6_output_0, %/layers.10/Constant_4_output_0, %/layers.10/Constant_7_output_0)
%/layers.10/Constant_8_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_5_output_0 = Add(%/layers.10/Slice_output_0, %/layers.10/Constant_8_output_0)
%/layers.10/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_5_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.4/maxpool/MaxPool_output_0, %/layers.10/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1064, %onnx::Conv_1065)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_1067, %onnx::Conv_1068)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1070, %onnx::Conv_1071)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_1073, %onnx::Conv_1074)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_1076, %onnx::Conv_1077)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1079, %onnx::Conv_1080)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/Add_3_output_0, %/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_4_output_0)
%/layers.11/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.11/Slice_output_0 = Slice(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_5_output_0, %/layers.11/Constant_6_output_0, %/layers.11/Constant_4_output_0, %/layers.11/Constant_7_output_0)
%/layers.11/Constant_8_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_5_output_0 = Add(%/layers.11/Slice_output_0, %/layers.11/Constant_8_output_0)
%/layers.11/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_5_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.4/maxpool/MaxPool_output_0, %/layers.11/vertex_op.5/maxpool/MaxPool_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%915 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %915
}
|
val_accuracy
| 92.147434
| 930,990,208
| 3,078,206
|
{'zcp_epe_nas': 101.59745068328914, 'zcp_fisher': 44.61912536621094, 'zcp_flops': 14895843328.0, 'zcp_grad_norm': 137.30580139160156, 'zcp_grasp': -41.287109375, 'zcp_jacov': -16.0628344813698, 'zcp_l2_norm': 958.3563232421875, 'zcp_nwot': 218.6060468550378, 'zcp_params': 3078206.0, 'zcp_plain': 0.11170092225074701, 'zcp_snip': 713.6679077148438, 'zcp_synflow': 85.39374243110042, 'zcp_zen': 93.83417510986328, 'zcp_val_accuracy': 0.926382184028625}
| |
NASBench101_132487
|
NASBench101
|
132487
|
501bbf33bd18b2de7f85e3d4d07c2114
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_653[FLOAT, 128x3x3x3]
%onnx::Conv_654[FLOAT, 128]
%onnx::Conv_656[FLOAT, 64x128x1x1]
%onnx::Conv_657[FLOAT, 64]
%onnx::Conv_659[FLOAT, 64x64x3x3]
%onnx::Conv_662[FLOAT, 64x64x3x3]
%onnx::Conv_665[FLOAT, 64x64x3x3]
%onnx::Conv_668[FLOAT, 64x128x1x1]
%onnx::Conv_671[FLOAT, 64x64x3x3]
%onnx::Conv_674[FLOAT, 64x64x3x3]
%onnx::Conv_677[FLOAT, 64x64x3x3]
%onnx::Conv_680[FLOAT, 64x128x1x1]
%onnx::Conv_683[FLOAT, 64x64x3x3]
%onnx::Conv_686[FLOAT, 64x64x3x3]
%onnx::Conv_689[FLOAT, 64x64x3x3]
%onnx::Conv_692[FLOAT, 128x128x1x1]
%onnx::Conv_695[FLOAT, 128x128x3x3]
%onnx::Conv_698[FLOAT, 128x128x3x3]
%onnx::Conv_701[FLOAT, 128x128x3x3]
%onnx::Conv_704[FLOAT, 128x256x1x1]
%onnx::Conv_707[FLOAT, 128x128x3x3]
%onnx::Conv_710[FLOAT, 128x128x3x3]
%onnx::Conv_713[FLOAT, 128x128x3x3]
%onnx::Conv_716[FLOAT, 128x256x1x1]
%onnx::Conv_719[FLOAT, 128x128x3x3]
%onnx::Conv_722[FLOAT, 128x128x3x3]
%onnx::Conv_725[FLOAT, 128x128x3x3]
%onnx::Conv_728[FLOAT, 256x256x1x1]
%onnx::Conv_729[FLOAT, 256]
%onnx::Conv_731[FLOAT, 256x256x3x3]
%onnx::Conv_734[FLOAT, 256x256x3x3]
%onnx::Conv_737[FLOAT, 256x256x3x3]
%onnx::Conv_740[FLOAT, 256x512x1x1]
%onnx::Conv_743[FLOAT, 256x256x3x3]
%onnx::Conv_746[FLOAT, 256x256x3x3]
%onnx::Conv_749[FLOAT, 256x256x3x3]
%onnx::Conv_752[FLOAT, 256x512x1x1]
%onnx::Conv_755[FLOAT, 256x256x3x3]
%onnx::Conv_758[FLOAT, 256x256x3x3]
%onnx::Conv_761[FLOAT, 256x256x3x3]
) {
%onnx::Conv_762 = Identity(%onnx::Conv_729)
%onnx::Conv_759 = Identity(%onnx::Conv_729)
%onnx::Conv_756 = Identity(%onnx::Conv_729)
%onnx::Conv_753 = Identity(%onnx::Conv_729)
%onnx::Conv_750 = Identity(%onnx::Conv_729)
%onnx::Conv_747 = Identity(%onnx::Conv_729)
%onnx::Conv_744 = Identity(%onnx::Conv_729)
%onnx::Conv_741 = Identity(%onnx::Conv_729)
%onnx::Conv_738 = Identity(%onnx::Conv_729)
%onnx::Conv_735 = Identity(%onnx::Conv_729)
%onnx::Conv_732 = Identity(%onnx::Conv_729)
%onnx::Conv_726 = Identity(%onnx::Conv_654)
%onnx::Conv_723 = Identity(%onnx::Conv_654)
%onnx::Conv_720 = Identity(%onnx::Conv_654)
%onnx::Conv_717 = Identity(%onnx::Conv_654)
%onnx::Conv_714 = Identity(%onnx::Conv_654)
%onnx::Conv_711 = Identity(%onnx::Conv_654)
%onnx::Conv_708 = Identity(%onnx::Conv_654)
%onnx::Conv_705 = Identity(%onnx::Conv_654)
%onnx::Conv_702 = Identity(%onnx::Conv_654)
%onnx::Conv_699 = Identity(%onnx::Conv_654)
%onnx::Conv_696 = Identity(%onnx::Conv_654)
%onnx::Conv_693 = Identity(%onnx::Conv_654)
%onnx::Conv_690 = Identity(%onnx::Conv_657)
%onnx::Conv_687 = Identity(%onnx::Conv_657)
%onnx::Conv_684 = Identity(%onnx::Conv_657)
%onnx::Conv_681 = Identity(%onnx::Conv_657)
%onnx::Conv_678 = Identity(%onnx::Conv_657)
%onnx::Conv_675 = Identity(%onnx::Conv_657)
%onnx::Conv_672 = Identity(%onnx::Conv_657)
%onnx::Conv_669 = Identity(%onnx::Conv_657)
%onnx::Conv_666 = Identity(%onnx::Conv_657)
%onnx::Conv_663 = Identity(%onnx::Conv_657)
%onnx::Conv_660 = Identity(%onnx::Conv_657)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_653, %onnx::Conv_654)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_656, %onnx::Conv_657)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_659, %onnx::Conv_660)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/Add_2_output_0, %/layers.1/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_662, %onnx::Conv_663)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/Add_5_output_0 = Add(%/layers.1/Add_4_output_0, %/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_5_output_0, %onnx::Conv_665, %onnx::Conv_666)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_668, %onnx::Conv_669)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_671, %onnx::Conv_672)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/Add_2_output_0, %/layers.2/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_674, %onnx::Conv_675)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/Add_5_output_0 = Add(%/layers.2/Add_4_output_0, %/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_5_output_0, %onnx::Conv_677, %onnx::Conv_678)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_680, %onnx::Conv_681)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_683, %onnx::Conv_684)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/Add_2_output_0, %/layers.3/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_686, %onnx::Conv_687)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/Add_5_output_0 = Add(%/layers.3/Add_4_output_0, %/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_5_output_0, %onnx::Conv_689, %onnx::Conv_690)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_692, %onnx::Conv_693)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_695, %onnx::Conv_696)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/Add_2_output_0, %/layers.5/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_698, %onnx::Conv_699)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/Add_5_output_0 = Add(%/layers.5/Add_4_output_0, %/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_5_output_0, %onnx::Conv_701, %onnx::Conv_702)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_704, %onnx::Conv_705)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_707, %onnx::Conv_708)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/Add_2_output_0, %/layers.6/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_710, %onnx::Conv_711)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/Add_5_output_0 = Add(%/layers.6/Add_4_output_0, %/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_5_output_0, %onnx::Conv_713, %onnx::Conv_714)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_716, %onnx::Conv_717)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_719, %onnx::Conv_720)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/Add_2_output_0, %/layers.7/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_722, %onnx::Conv_723)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/Add_5_output_0 = Add(%/layers.7/Add_4_output_0, %/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_5_output_0, %onnx::Conv_725, %onnx::Conv_726)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_728, %onnx::Conv_729)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_731, %onnx::Conv_732)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/Add_2_output_0, %/layers.9/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_734, %onnx::Conv_735)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/Add_5_output_0 = Add(%/layers.9/Add_4_output_0, %/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_5_output_0, %onnx::Conv_737, %onnx::Conv_738)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_740, %onnx::Conv_741)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_743, %onnx::Conv_744)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/Add_2_output_0, %/layers.10/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_746, %onnx::Conv_747)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/Add_5_output_0 = Add(%/layers.10/Add_4_output_0, %/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_5_output_0, %onnx::Conv_749, %onnx::Conv_750)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_752, %onnx::Conv_753)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_755, %onnx::Conv_756)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/Add_2_output_0, %/layers.11/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_758, %onnx::Conv_759)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/Add_5_output_0 = Add(%/layers.11/Add_4_output_0, %/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_5_output_0, %onnx::Conv_761, %onnx::Conv_762)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%651 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %651
}
|
val_accuracy
| 90.885419
| 2,191,796,224
| 7,421,066
|
{'zcp_epe_nas': 146.9794391889311, 'zcp_fisher': 36.863739013671875, 'zcp_flops': 35068739584.0, 'zcp_grad_norm': 109.95054626464844, 'zcp_grasp': 9.24462890625, 'zcp_jacov': -16.05442477319688, 'zcp_l2_norm': 649.1217651367188, 'zcp_nwot': 218.15846358613103, 'zcp_params': 7421066.0, 'zcp_plain': 0.08718677610158901, 'zcp_snip': 667.1046752929688, 'zcp_synflow': 132.8684704727392, 'zcp_zen': 88.02941131591797, 'zcp_val_accuracy': 0.9034455418586731}
| |
NASBench101_257834
|
NASBench101
|
257834
|
9c1d41d7ecac665b414487dc804e4b5f
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_788[FLOAT, 128x3x3x3]
%onnx::Conv_789[FLOAT, 128]
%onnx::Conv_791[FLOAT, 64x128x1x1]
%onnx::Conv_792[FLOAT, 64]
%onnx::Conv_794[FLOAT, 64x64x1x1]
%onnx::Conv_797[FLOAT, 64x64x1x1]
%onnx::Conv_800[FLOAT, 64x64x1x1]
%onnx::Conv_803[FLOAT, 64x64x1x1]
%onnx::Conv_806[FLOAT, 64x128x1x1]
%onnx::Conv_809[FLOAT, 64x64x1x1]
%onnx::Conv_812[FLOAT, 64x64x1x1]
%onnx::Conv_815[FLOAT, 64x64x1x1]
%onnx::Conv_818[FLOAT, 64x64x1x1]
%onnx::Conv_821[FLOAT, 64x128x1x1]
%onnx::Conv_824[FLOAT, 64x64x1x1]
%onnx::Conv_827[FLOAT, 64x64x1x1]
%onnx::Conv_830[FLOAT, 64x64x1x1]
%onnx::Conv_833[FLOAT, 64x64x1x1]
%onnx::Conv_836[FLOAT, 128x128x1x1]
%onnx::Conv_839[FLOAT, 128x128x1x1]
%onnx::Conv_842[FLOAT, 128x128x1x1]
%onnx::Conv_845[FLOAT, 128x128x1x1]
%onnx::Conv_848[FLOAT, 128x128x1x1]
%onnx::Conv_851[FLOAT, 128x256x1x1]
%onnx::Conv_854[FLOAT, 128x128x1x1]
%onnx::Conv_857[FLOAT, 128x128x1x1]
%onnx::Conv_860[FLOAT, 128x128x1x1]
%onnx::Conv_863[FLOAT, 128x128x1x1]
%onnx::Conv_866[FLOAT, 128x256x1x1]
%onnx::Conv_869[FLOAT, 128x128x1x1]
%onnx::Conv_872[FLOAT, 128x128x1x1]
%onnx::Conv_875[FLOAT, 128x128x1x1]
%onnx::Conv_878[FLOAT, 128x128x1x1]
%onnx::Conv_881[FLOAT, 256x256x1x1]
%onnx::Conv_882[FLOAT, 256]
%onnx::Conv_884[FLOAT, 256x256x1x1]
%onnx::Conv_887[FLOAT, 256x256x1x1]
%onnx::Conv_890[FLOAT, 256x256x1x1]
%onnx::Conv_893[FLOAT, 256x256x1x1]
%onnx::Conv_896[FLOAT, 256x512x1x1]
%onnx::Conv_899[FLOAT, 256x256x1x1]
%onnx::Conv_902[FLOAT, 256x256x1x1]
%onnx::Conv_905[FLOAT, 256x256x1x1]
%onnx::Conv_908[FLOAT, 256x256x1x1]
%onnx::Conv_911[FLOAT, 256x512x1x1]
%onnx::Conv_914[FLOAT, 256x256x1x1]
%onnx::Conv_917[FLOAT, 256x256x1x1]
%onnx::Conv_920[FLOAT, 256x256x1x1]
%onnx::Conv_923[FLOAT, 256x256x1x1]
) {
%onnx::Conv_924 = Identity(%onnx::Conv_882)
%onnx::Conv_921 = Identity(%onnx::Conv_882)
%onnx::Conv_918 = Identity(%onnx::Conv_882)
%onnx::Conv_915 = Identity(%onnx::Conv_882)
%onnx::Conv_912 = Identity(%onnx::Conv_882)
%onnx::Conv_909 = Identity(%onnx::Conv_882)
%onnx::Conv_906 = Identity(%onnx::Conv_882)
%onnx::Conv_903 = Identity(%onnx::Conv_882)
%onnx::Conv_900 = Identity(%onnx::Conv_882)
%onnx::Conv_897 = Identity(%onnx::Conv_882)
%onnx::Conv_894 = Identity(%onnx::Conv_882)
%onnx::Conv_891 = Identity(%onnx::Conv_882)
%onnx::Conv_888 = Identity(%onnx::Conv_882)
%onnx::Conv_885 = Identity(%onnx::Conv_882)
%onnx::Conv_879 = Identity(%onnx::Conv_789)
%onnx::Conv_876 = Identity(%onnx::Conv_789)
%onnx::Conv_873 = Identity(%onnx::Conv_789)
%onnx::Conv_870 = Identity(%onnx::Conv_789)
%onnx::Conv_867 = Identity(%onnx::Conv_789)
%onnx::Conv_864 = Identity(%onnx::Conv_789)
%onnx::Conv_861 = Identity(%onnx::Conv_789)
%onnx::Conv_858 = Identity(%onnx::Conv_789)
%onnx::Conv_855 = Identity(%onnx::Conv_789)
%onnx::Conv_852 = Identity(%onnx::Conv_789)
%onnx::Conv_849 = Identity(%onnx::Conv_789)
%onnx::Conv_846 = Identity(%onnx::Conv_789)
%onnx::Conv_843 = Identity(%onnx::Conv_789)
%onnx::Conv_840 = Identity(%onnx::Conv_789)
%onnx::Conv_837 = Identity(%onnx::Conv_789)
%onnx::Conv_834 = Identity(%onnx::Conv_792)
%onnx::Conv_831 = Identity(%onnx::Conv_792)
%onnx::Conv_828 = Identity(%onnx::Conv_792)
%onnx::Conv_825 = Identity(%onnx::Conv_792)
%onnx::Conv_822 = Identity(%onnx::Conv_792)
%onnx::Conv_819 = Identity(%onnx::Conv_792)
%onnx::Conv_816 = Identity(%onnx::Conv_792)
%onnx::Conv_813 = Identity(%onnx::Conv_792)
%onnx::Conv_810 = Identity(%onnx::Conv_792)
%onnx::Conv_807 = Identity(%onnx::Conv_792)
%onnx::Conv_804 = Identity(%onnx::Conv_792)
%onnx::Conv_801 = Identity(%onnx::Conv_792)
%onnx::Conv_798 = Identity(%onnx::Conv_792)
%onnx::Conv_795 = Identity(%onnx::Conv_792)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_788, %onnx::Conv_789)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_797, %onnx::Conv_798)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_800, %onnx::Conv_801)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0)
%/layers.1/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_4_output_0)
%/layers.1/Add_5_output_0 = Add(%/layers.1/Add_4_output_0, %/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/Add_6_output_0 = Add(%/layers.1/Add_5_output_0, %/layers.1/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_6_output_0, %onnx::Conv_803, %onnx::Conv_804)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.4/maxpool/MaxPool_output_0, %/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_806, %onnx::Conv_807)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_809, %onnx::Conv_810)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_812, %onnx::Conv_813)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_815, %onnx::Conv_816)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0)
%/layers.2/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_4_output_0)
%/layers.2/Add_5_output_0 = Add(%/layers.2/Add_4_output_0, %/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/Add_6_output_0 = Add(%/layers.2/Add_5_output_0, %/layers.2/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_6_output_0, %onnx::Conv_818, %onnx::Conv_819)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.4/maxpool/MaxPool_output_0, %/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_821, %onnx::Conv_822)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_824, %onnx::Conv_825)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_827, %onnx::Conv_828)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_830, %onnx::Conv_831)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0)
%/layers.3/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_4_output_0)
%/layers.3/Add_5_output_0 = Add(%/layers.3/Add_4_output_0, %/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/Add_6_output_0 = Add(%/layers.3/Add_5_output_0, %/layers.3/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_6_output_0, %onnx::Conv_833, %onnx::Conv_834)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.4/maxpool/MaxPool_output_0, %/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_836, %onnx::Conv_837)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_839, %onnx::Conv_840)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_842, %onnx::Conv_843)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_845, %onnx::Conv_846)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0)
%/layers.5/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_4_output_0)
%/layers.5/Add_5_output_0 = Add(%/layers.5/Add_4_output_0, %/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/Add_6_output_0 = Add(%/layers.5/Add_5_output_0, %/layers.5/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_6_output_0, %onnx::Conv_848, %onnx::Conv_849)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.4/maxpool/MaxPool_output_0, %/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_851, %onnx::Conv_852)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_854, %onnx::Conv_855)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_857, %onnx::Conv_858)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0)
%/layers.6/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_4_output_0)
%/layers.6/Add_5_output_0 = Add(%/layers.6/Add_4_output_0, %/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/Add_6_output_0 = Add(%/layers.6/Add_5_output_0, %/layers.6/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_6_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.4/maxpool/MaxPool_output_0, %/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0)
%/layers.7/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_4_output_0)
%/layers.7/Add_5_output_0 = Add(%/layers.7/Add_4_output_0, %/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/Add_6_output_0 = Add(%/layers.7/Add_5_output_0, %/layers.7/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_6_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.4/maxpool/MaxPool_output_0, %/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0)
%/layers.9/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_4_output_0)
%/layers.9/Add_5_output_0 = Add(%/layers.9/Add_4_output_0, %/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/Add_6_output_0 = Add(%/layers.9/Add_5_output_0, %/layers.9/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_6_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.4/maxpool/MaxPool_output_0, %/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0)
%/layers.10/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_4_output_0)
%/layers.10/Add_5_output_0 = Add(%/layers.10/Add_4_output_0, %/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/Add_6_output_0 = Add(%/layers.10/Add_5_output_0, %/layers.10/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_6_output_0, %onnx::Conv_908, %onnx::Conv_909)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.4/maxpool/MaxPool_output_0, %/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_911, %onnx::Conv_912)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_914, %onnx::Conv_915)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_917, %onnx::Conv_918)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_920, %onnx::Conv_921)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0)
%/layers.11/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_4_output_0)
%/layers.11/Add_5_output_0 = Add(%/layers.11/Add_4_output_0, %/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/Add_6_output_0 = Add(%/layers.11/Add_5_output_0, %/layers.11/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_6_output_0, %onnx::Conv_923, %onnx::Conv_924)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.4/maxpool/MaxPool_output_0, %/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%786 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %786
}
|
val_accuracy
| 67.107373
| 458,106,880
| 1,488,650
|
{'zcp_epe_nas': 91.34936952719188, 'zcp_fisher': 2992.593017578125, 'zcp_flops': 7329710080.0, 'zcp_grad_norm': 1113.0748291015625, 'zcp_grasp': -20795.90625, 'zcp_jacov': -16.042805925673765, 'zcp_l2_norm': 798.5652465820312, 'zcp_nwot': 222.1583929387852, 'zcp_params': 1488650.0, 'zcp_plain': 0.020146172493696, 'zcp_snip': 4774.81494140625, 'zcp_synflow': 127.30215968125195, 'zcp_zen': 70.30174255371094, 'zcp_val_accuracy': 0.8829126358032221}
| |
NASBench101_263397
|
NASBench101
|
263397
|
9f84adc0f8b0d0198a2ef46fe8011538
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_869[FLOAT, 128x3x3x3]
%onnx::Conv_870[FLOAT, 128]
%onnx::Conv_872[FLOAT, 128x128x1x1]
%onnx::Conv_875[FLOAT, 128x128x1x1]
%onnx::Conv_878[FLOAT, 128x128x3x3]
%onnx::Conv_881[FLOAT, 128x128x1x1]
%onnx::Conv_884[FLOAT, 128x128x3x3]
%onnx::Conv_887[FLOAT, 128x128x1x1]
%onnx::Conv_890[FLOAT, 128x128x1x1]
%onnx::Conv_893[FLOAT, 128x128x1x1]
%onnx::Conv_896[FLOAT, 128x128x3x3]
%onnx::Conv_899[FLOAT, 128x128x1x1]
%onnx::Conv_902[FLOAT, 128x128x3x3]
%onnx::Conv_905[FLOAT, 128x128x1x1]
%onnx::Conv_908[FLOAT, 128x128x1x1]
%onnx::Conv_911[FLOAT, 128x128x1x1]
%onnx::Conv_914[FLOAT, 128x128x3x3]
%onnx::Conv_917[FLOAT, 128x128x1x1]
%onnx::Conv_920[FLOAT, 128x128x3x3]
%onnx::Conv_923[FLOAT, 128x128x1x1]
%onnx::Conv_926[FLOAT, 256x128x1x1]
%onnx::Conv_927[FLOAT, 256]
%onnx::Conv_929[FLOAT, 256x256x1x1]
%onnx::Conv_932[FLOAT, 256x256x3x3]
%onnx::Conv_935[FLOAT, 256x128x1x1]
%onnx::Conv_938[FLOAT, 256x256x3x3]
%onnx::Conv_941[FLOAT, 256x256x1x1]
%onnx::Conv_944[FLOAT, 256x256x1x1]
%onnx::Conv_947[FLOAT, 256x256x1x1]
%onnx::Conv_950[FLOAT, 256x256x3x3]
%onnx::Conv_953[FLOAT, 256x256x1x1]
%onnx::Conv_956[FLOAT, 256x256x3x3]
%onnx::Conv_959[FLOAT, 256x256x1x1]
%onnx::Conv_962[FLOAT, 256x256x1x1]
%onnx::Conv_965[FLOAT, 256x256x1x1]
%onnx::Conv_968[FLOAT, 256x256x3x3]
%onnx::Conv_971[FLOAT, 256x256x1x1]
%onnx::Conv_974[FLOAT, 256x256x3x3]
%onnx::Conv_977[FLOAT, 256x256x1x1]
%onnx::Conv_980[FLOAT, 512x256x1x1]
%onnx::Conv_981[FLOAT, 512]
%onnx::Conv_983[FLOAT, 512x512x1x1]
%onnx::Conv_986[FLOAT, 512x512x3x3]
%onnx::Conv_989[FLOAT, 512x256x1x1]
%onnx::Conv_992[FLOAT, 512x512x3x3]
%onnx::Conv_995[FLOAT, 512x512x1x1]
%onnx::Conv_998[FLOAT, 512x512x1x1]
%onnx::Conv_1001[FLOAT, 512x512x1x1]
%onnx::Conv_1004[FLOAT, 512x512x3x3]
%onnx::Conv_1007[FLOAT, 512x512x1x1]
%onnx::Conv_1010[FLOAT, 512x512x3x3]
%onnx::Conv_1013[FLOAT, 512x512x1x1]
%onnx::Conv_1016[FLOAT, 512x512x1x1]
%onnx::Conv_1019[FLOAT, 512x512x1x1]
%onnx::Conv_1022[FLOAT, 512x512x3x3]
%onnx::Conv_1025[FLOAT, 512x512x1x1]
%onnx::Conv_1028[FLOAT, 512x512x3x3]
%onnx::Conv_1031[FLOAT, 512x512x1x1]
) {
%onnx::Conv_1032 = Identity(%onnx::Conv_981)
%onnx::Conv_1029 = Identity(%onnx::Conv_981)
%onnx::Conv_1026 = Identity(%onnx::Conv_981)
%onnx::Conv_1023 = Identity(%onnx::Conv_981)
%onnx::Conv_1020 = Identity(%onnx::Conv_981)
%onnx::Conv_1017 = Identity(%onnx::Conv_981)
%onnx::Conv_1014 = Identity(%onnx::Conv_981)
%onnx::Conv_1011 = Identity(%onnx::Conv_981)
%onnx::Conv_1008 = Identity(%onnx::Conv_981)
%onnx::Conv_1005 = Identity(%onnx::Conv_981)
%onnx::Conv_1002 = Identity(%onnx::Conv_981)
%onnx::Conv_999 = Identity(%onnx::Conv_981)
%onnx::Conv_996 = Identity(%onnx::Conv_981)
%onnx::Conv_993 = Identity(%onnx::Conv_981)
%onnx::Conv_990 = Identity(%onnx::Conv_981)
%onnx::Conv_987 = Identity(%onnx::Conv_981)
%onnx::Conv_984 = Identity(%onnx::Conv_981)
%onnx::Conv_978 = Identity(%onnx::Conv_927)
%onnx::Conv_975 = Identity(%onnx::Conv_927)
%onnx::Conv_972 = Identity(%onnx::Conv_927)
%onnx::Conv_969 = Identity(%onnx::Conv_927)
%onnx::Conv_966 = Identity(%onnx::Conv_927)
%onnx::Conv_963 = Identity(%onnx::Conv_927)
%onnx::Conv_960 = Identity(%onnx::Conv_927)
%onnx::Conv_957 = Identity(%onnx::Conv_927)
%onnx::Conv_954 = Identity(%onnx::Conv_927)
%onnx::Conv_951 = Identity(%onnx::Conv_927)
%onnx::Conv_948 = Identity(%onnx::Conv_927)
%onnx::Conv_945 = Identity(%onnx::Conv_927)
%onnx::Conv_942 = Identity(%onnx::Conv_927)
%onnx::Conv_939 = Identity(%onnx::Conv_927)
%onnx::Conv_936 = Identity(%onnx::Conv_927)
%onnx::Conv_933 = Identity(%onnx::Conv_927)
%onnx::Conv_930 = Identity(%onnx::Conv_927)
%onnx::Conv_924 = Identity(%onnx::Conv_870)
%onnx::Conv_921 = Identity(%onnx::Conv_870)
%onnx::Conv_918 = Identity(%onnx::Conv_870)
%onnx::Conv_915 = Identity(%onnx::Conv_870)
%onnx::Conv_912 = Identity(%onnx::Conv_870)
%onnx::Conv_909 = Identity(%onnx::Conv_870)
%onnx::Conv_906 = Identity(%onnx::Conv_870)
%onnx::Conv_903 = Identity(%onnx::Conv_870)
%onnx::Conv_900 = Identity(%onnx::Conv_870)
%onnx::Conv_897 = Identity(%onnx::Conv_870)
%onnx::Conv_894 = Identity(%onnx::Conv_870)
%onnx::Conv_891 = Identity(%onnx::Conv_870)
%onnx::Conv_888 = Identity(%onnx::Conv_870)
%onnx::Conv_885 = Identity(%onnx::Conv_870)
%onnx::Conv_882 = Identity(%onnx::Conv_870)
%onnx::Conv_879 = Identity(%onnx::Conv_870)
%onnx::Conv_876 = Identity(%onnx::Conv_870)
%onnx::Conv_873 = Identity(%onnx::Conv_870)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_869, %onnx::Conv_870)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/Add_2_output_0, %/layers.1/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_5_output_0 = Add(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/Add_6_output_0 = Add(%/layers.1/Add_5_output_0, %/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_6_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/Add_2_output_0, %/layers.2/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_5_output_0 = Add(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/Add_6_output_0 = Add(%/layers.2/Add_5_output_0, %/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_6_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_908, %onnx::Conv_909)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_911, %onnx::Conv_912)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/Add_2_output_0, %/layers.3/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_914, %onnx::Conv_915)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_917, %onnx::Conv_918)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_4_output_0, %onnx::Conv_920, %onnx::Conv_921)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_5_output_0 = Add(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/Add_6_output_0 = Add(%/layers.3/Add_5_output_0, %/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_6_output_0, %onnx::Conv_923, %onnx::Conv_924)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_926, %onnx::Conv_927)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_929, %onnx::Conv_930)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/Add_2_output_0, %/layers.5/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_932, %onnx::Conv_933)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_935, %onnx::Conv_936)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_938, %onnx::Conv_939)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_5_output_0 = Add(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/Add_6_output_0 = Add(%/layers.5/Add_5_output_0, %/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_6_output_0, %onnx::Conv_941, %onnx::Conv_942)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_944, %onnx::Conv_945)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_947, %onnx::Conv_948)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/Add_2_output_0, %/layers.6/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_950, %onnx::Conv_951)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_953, %onnx::Conv_954)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_956, %onnx::Conv_957)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_5_output_0 = Add(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/Add_6_output_0 = Add(%/layers.6/Add_5_output_0, %/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_6_output_0, %onnx::Conv_959, %onnx::Conv_960)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_962, %onnx::Conv_963)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_965, %onnx::Conv_966)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/Add_2_output_0, %/layers.7/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_968, %onnx::Conv_969)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_4_output_0, %onnx::Conv_974, %onnx::Conv_975)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_5_output_0 = Add(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/Add_6_output_0 = Add(%/layers.7/Add_5_output_0, %/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_6_output_0, %onnx::Conv_977, %onnx::Conv_978)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/Add_2_output_0, %/layers.9/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_5_output_0 = Add(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/Add_6_output_0 = Add(%/layers.9/Add_5_output_0, %/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_6_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/Add_2_output_0, %/layers.10/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_5_output_0 = Add(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/Add_6_output_0 = Add(%/layers.10/Add_5_output_0, %/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_6_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/Add_2_output_0, %/layers.11/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1025, %onnx::Conv_1026)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_4_output_0, %onnx::Conv_1028, %onnx::Conv_1029)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_5_output_0 = Add(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/Add_6_output_0 = Add(%/layers.11/Add_5_output_0, %/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_6_output_0, %onnx::Conv_1031, %onnx::Conv_1032)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%867 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %867
}
|
val_accuracy
| 92.13742
| 6,617,835,520
| 22,421,642
|
{'zcp_epe_nas': 134.55914991256117, 'zcp_fisher': 168.1866455078125, 'zcp_flops': 105885368320.0, 'zcp_grad_norm': 222.28721618652344, 'zcp_grasp': -3.26708984375, 'zcp_jacov': -16.05687802068193, 'zcp_l2_norm': 1242.342529296875, 'zcp_nwot': 234.85042326551596, 'zcp_params': 22421642.0, 'zcp_plain': 0.005058322101831001, 'zcp_snip': 1803.6968994140625, 'zcp_synflow': 129.67580388238764, 'zcp_zen': 117.80960845947266, 'zcp_val_accuracy': 0.9120593070983881}
| |
NASBench101_45303
|
NASBench101
|
45303
|
1b7e87eba1594be7377a8fb942f64a7d
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_779[FLOAT, 128x3x3x3]
%onnx::Conv_780[FLOAT, 128]
%onnx::Conv_782[FLOAT, 64x128x1x1]
%onnx::Conv_783[FLOAT, 64]
%onnx::Conv_785[FLOAT, 64x64x3x3]
%onnx::Conv_788[FLOAT, 64x128x1x1]
%onnx::Conv_791[FLOAT, 64x64x1x1]
%onnx::Conv_794[FLOAT, 64x64x3x3]
%onnx::Conv_797[FLOAT, 64x128x1x1]
%onnx::Conv_800[FLOAT, 64x64x3x3]
%onnx::Conv_803[FLOAT, 64x128x1x1]
%onnx::Conv_806[FLOAT, 64x64x1x1]
%onnx::Conv_809[FLOAT, 64x64x3x3]
%onnx::Conv_812[FLOAT, 64x128x1x1]
%onnx::Conv_815[FLOAT, 64x64x3x3]
%onnx::Conv_818[FLOAT, 64x128x1x1]
%onnx::Conv_821[FLOAT, 64x64x1x1]
%onnx::Conv_824[FLOAT, 64x64x3x3]
%onnx::Conv_827[FLOAT, 128x128x1x1]
%onnx::Conv_830[FLOAT, 128x128x3x3]
%onnx::Conv_833[FLOAT, 128x128x1x1]
%onnx::Conv_836[FLOAT, 128x128x1x1]
%onnx::Conv_839[FLOAT, 128x128x3x3]
%onnx::Conv_842[FLOAT, 128x256x1x1]
%onnx::Conv_845[FLOAT, 128x128x3x3]
%onnx::Conv_848[FLOAT, 128x256x1x1]
%onnx::Conv_851[FLOAT, 128x128x1x1]
%onnx::Conv_854[FLOAT, 128x128x3x3]
%onnx::Conv_857[FLOAT, 128x256x1x1]
%onnx::Conv_860[FLOAT, 128x128x3x3]
%onnx::Conv_863[FLOAT, 128x256x1x1]
%onnx::Conv_866[FLOAT, 128x128x1x1]
%onnx::Conv_869[FLOAT, 128x128x3x3]
%onnx::Conv_872[FLOAT, 256x256x1x1]
%onnx::Conv_873[FLOAT, 256]
%onnx::Conv_875[FLOAT, 256x256x3x3]
%onnx::Conv_878[FLOAT, 256x256x1x1]
%onnx::Conv_881[FLOAT, 256x256x1x1]
%onnx::Conv_884[FLOAT, 256x256x3x3]
%onnx::Conv_887[FLOAT, 256x512x1x1]
%onnx::Conv_890[FLOAT, 256x256x3x3]
%onnx::Conv_893[FLOAT, 256x512x1x1]
%onnx::Conv_896[FLOAT, 256x256x1x1]
%onnx::Conv_899[FLOAT, 256x256x3x3]
%onnx::Conv_902[FLOAT, 256x512x1x1]
%onnx::Conv_905[FLOAT, 256x256x3x3]
%onnx::Conv_908[FLOAT, 256x512x1x1]
%onnx::Conv_911[FLOAT, 256x256x1x1]
%onnx::Conv_914[FLOAT, 256x256x3x3]
) {
%onnx::Conv_915 = Identity(%onnx::Conv_873)
%onnx::Conv_912 = Identity(%onnx::Conv_873)
%onnx::Conv_909 = Identity(%onnx::Conv_873)
%onnx::Conv_906 = Identity(%onnx::Conv_873)
%onnx::Conv_903 = Identity(%onnx::Conv_873)
%onnx::Conv_900 = Identity(%onnx::Conv_873)
%onnx::Conv_897 = Identity(%onnx::Conv_873)
%onnx::Conv_894 = Identity(%onnx::Conv_873)
%onnx::Conv_891 = Identity(%onnx::Conv_873)
%onnx::Conv_888 = Identity(%onnx::Conv_873)
%onnx::Conv_885 = Identity(%onnx::Conv_873)
%onnx::Conv_882 = Identity(%onnx::Conv_873)
%onnx::Conv_879 = Identity(%onnx::Conv_873)
%onnx::Conv_876 = Identity(%onnx::Conv_873)
%onnx::Conv_870 = Identity(%onnx::Conv_780)
%onnx::Conv_867 = Identity(%onnx::Conv_780)
%onnx::Conv_864 = Identity(%onnx::Conv_780)
%onnx::Conv_861 = Identity(%onnx::Conv_780)
%onnx::Conv_858 = Identity(%onnx::Conv_780)
%onnx::Conv_855 = Identity(%onnx::Conv_780)
%onnx::Conv_852 = Identity(%onnx::Conv_780)
%onnx::Conv_849 = Identity(%onnx::Conv_780)
%onnx::Conv_846 = Identity(%onnx::Conv_780)
%onnx::Conv_843 = Identity(%onnx::Conv_780)
%onnx::Conv_840 = Identity(%onnx::Conv_780)
%onnx::Conv_837 = Identity(%onnx::Conv_780)
%onnx::Conv_834 = Identity(%onnx::Conv_780)
%onnx::Conv_831 = Identity(%onnx::Conv_780)
%onnx::Conv_828 = Identity(%onnx::Conv_780)
%onnx::Conv_825 = Identity(%onnx::Conv_783)
%onnx::Conv_822 = Identity(%onnx::Conv_783)
%onnx::Conv_819 = Identity(%onnx::Conv_783)
%onnx::Conv_816 = Identity(%onnx::Conv_783)
%onnx::Conv_813 = Identity(%onnx::Conv_783)
%onnx::Conv_810 = Identity(%onnx::Conv_783)
%onnx::Conv_807 = Identity(%onnx::Conv_783)
%onnx::Conv_804 = Identity(%onnx::Conv_783)
%onnx::Conv_801 = Identity(%onnx::Conv_783)
%onnx::Conv_798 = Identity(%onnx::Conv_783)
%onnx::Conv_795 = Identity(%onnx::Conv_783)
%onnx::Conv_792 = Identity(%onnx::Conv_783)
%onnx::Conv_789 = Identity(%onnx::Conv_783)
%onnx::Conv_786 = Identity(%onnx::Conv_783)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_779, %onnx::Conv_780)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_782, %onnx::Conv_783)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_785, %onnx::Conv_786)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_788, %onnx::Conv_789)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/Add_3_output_0, %/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_5_output_0 = Add(%/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_5_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %/layers.1/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_797, %onnx::Conv_798)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_800, %onnx::Conv_801)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_803, %onnx::Conv_804)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_806, %onnx::Conv_807)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/Add_3_output_0, %/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_809, %onnx::Conv_810)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_5_output_0 = Add(%/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_5_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %/layers.2/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_812, %onnx::Conv_813)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_815, %onnx::Conv_816)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_818, %onnx::Conv_819)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_821, %onnx::Conv_822)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/Add_3_output_0, %/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_4_output_0, %onnx::Conv_824, %onnx::Conv_825)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_5_output_0 = Add(%/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_5_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %/layers.3/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_827, %onnx::Conv_828)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_830, %onnx::Conv_831)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_833, %onnx::Conv_834)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_836, %onnx::Conv_837)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/Add_3_output_0, %/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_839, %onnx::Conv_840)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_5_output_0 = Add(%/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_5_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %/layers.5/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_842, %onnx::Conv_843)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_845, %onnx::Conv_846)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_848, %onnx::Conv_849)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_851, %onnx::Conv_852)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/Add_3_output_0, %/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_854, %onnx::Conv_855)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_5_output_0 = Add(%/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_5_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %/layers.6/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_857, %onnx::Conv_858)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/Add_3_output_0, %/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_4_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_5_output_0 = Add(%/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_5_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %/layers.7/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/Add_3_output_0, %/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_5_output_0 = Add(%/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_5_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %/layers.9/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/Add_3_output_0, %/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_5_output_0 = Add(%/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_5_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %/layers.10/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_908, %onnx::Conv_909)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_911, %onnx::Conv_912)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/Add_3_output_0, %/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_4_output_0, %onnx::Conv_914, %onnx::Conv_915)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_5_output_0 = Add(%/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_5_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %/layers.11/vertex_op.5/maxpool/MaxPool_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%777 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %777
}
|
val_accuracy
| 90.234375
| 1,724,786,688
| 5,793,546
|
{'zcp_epe_nas': 151.17203518666994, 'zcp_fisher': 95.44281768798828, 'zcp_flops': 27596587008.0, 'zcp_grad_norm': 179.3262939453125, 'zcp_grasp': -260.759521484375, 'zcp_jacov': -16.04830251866308, 'zcp_l2_norm': 844.0401611328125, 'zcp_nwot': 221.76112298243396, 'zcp_params': 5793546.0, 'zcp_plain': 0.36606723070144603, 'zcp_snip': 1073.2872314453125, 'zcp_synflow': 97.01029187482975, 'zcp_zen': 90.4710922241211, 'zcp_val_accuracy': 0.929286837577819}
| |
NASBench101_239216
|
NASBench101
|
239216
|
90c30b0f6ff95f271b72b72ef3ccd4bc
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_770[FLOAT, 128x3x3x3]
%onnx::Conv_771[FLOAT, 128]
%onnx::Conv_773[FLOAT, 64x128x1x1]
%onnx::Conv_774[FLOAT, 64]
%onnx::Conv_776[FLOAT, 64x64x3x3]
%onnx::Conv_779[FLOAT, 64x64x3x3]
%onnx::Conv_782[FLOAT, 64x64x3x3]
%onnx::Conv_785[FLOAT, 64x64x1x1]
%onnx::Conv_788[FLOAT, 64x128x1x1]
%onnx::Conv_791[FLOAT, 64x64x3x3]
%onnx::Conv_794[FLOAT, 64x64x3x3]
%onnx::Conv_797[FLOAT, 64x64x3x3]
%onnx::Conv_800[FLOAT, 64x64x1x1]
%onnx::Conv_803[FLOAT, 64x128x1x1]
%onnx::Conv_806[FLOAT, 64x64x3x3]
%onnx::Conv_809[FLOAT, 64x64x3x3]
%onnx::Conv_812[FLOAT, 64x64x3x3]
%onnx::Conv_815[FLOAT, 64x64x1x1]
%onnx::Conv_818[FLOAT, 128x128x1x1]
%onnx::Conv_821[FLOAT, 128x128x3x3]
%onnx::Conv_824[FLOAT, 128x128x3x3]
%onnx::Conv_827[FLOAT, 128x128x3x3]
%onnx::Conv_830[FLOAT, 128x128x1x1]
%onnx::Conv_833[FLOAT, 128x256x1x1]
%onnx::Conv_836[FLOAT, 128x128x3x3]
%onnx::Conv_839[FLOAT, 128x128x3x3]
%onnx::Conv_842[FLOAT, 128x128x3x3]
%onnx::Conv_845[FLOAT, 128x128x1x1]
%onnx::Conv_848[FLOAT, 128x256x1x1]
%onnx::Conv_851[FLOAT, 128x128x3x3]
%onnx::Conv_854[FLOAT, 128x128x3x3]
%onnx::Conv_857[FLOAT, 128x128x3x3]
%onnx::Conv_860[FLOAT, 128x128x1x1]
%onnx::Conv_863[FLOAT, 256x256x1x1]
%onnx::Conv_864[FLOAT, 256]
%onnx::Conv_866[FLOAT, 256x256x3x3]
%onnx::Conv_869[FLOAT, 256x256x3x3]
%onnx::Conv_872[FLOAT, 256x256x3x3]
%onnx::Conv_875[FLOAT, 256x256x1x1]
%onnx::Conv_878[FLOAT, 256x512x1x1]
%onnx::Conv_881[FLOAT, 256x256x3x3]
%onnx::Conv_884[FLOAT, 256x256x3x3]
%onnx::Conv_887[FLOAT, 256x256x3x3]
%onnx::Conv_890[FLOAT, 256x256x1x1]
%onnx::Conv_893[FLOAT, 256x512x1x1]
%onnx::Conv_896[FLOAT, 256x256x3x3]
%onnx::Conv_899[FLOAT, 256x256x3x3]
%onnx::Conv_902[FLOAT, 256x256x3x3]
%onnx::Conv_905[FLOAT, 256x256x1x1]
) {
%onnx::Conv_906 = Identity(%onnx::Conv_864)
%onnx::Conv_903 = Identity(%onnx::Conv_864)
%onnx::Conv_900 = Identity(%onnx::Conv_864)
%onnx::Conv_897 = Identity(%onnx::Conv_864)
%onnx::Conv_894 = Identity(%onnx::Conv_864)
%onnx::Conv_891 = Identity(%onnx::Conv_864)
%onnx::Conv_888 = Identity(%onnx::Conv_864)
%onnx::Conv_885 = Identity(%onnx::Conv_864)
%onnx::Conv_882 = Identity(%onnx::Conv_864)
%onnx::Conv_879 = Identity(%onnx::Conv_864)
%onnx::Conv_876 = Identity(%onnx::Conv_864)
%onnx::Conv_873 = Identity(%onnx::Conv_864)
%onnx::Conv_870 = Identity(%onnx::Conv_864)
%onnx::Conv_867 = Identity(%onnx::Conv_864)
%onnx::Conv_861 = Identity(%onnx::Conv_771)
%onnx::Conv_858 = Identity(%onnx::Conv_771)
%onnx::Conv_855 = Identity(%onnx::Conv_771)
%onnx::Conv_852 = Identity(%onnx::Conv_771)
%onnx::Conv_849 = Identity(%onnx::Conv_771)
%onnx::Conv_846 = Identity(%onnx::Conv_771)
%onnx::Conv_843 = Identity(%onnx::Conv_771)
%onnx::Conv_840 = Identity(%onnx::Conv_771)
%onnx::Conv_837 = Identity(%onnx::Conv_771)
%onnx::Conv_834 = Identity(%onnx::Conv_771)
%onnx::Conv_831 = Identity(%onnx::Conv_771)
%onnx::Conv_828 = Identity(%onnx::Conv_771)
%onnx::Conv_825 = Identity(%onnx::Conv_771)
%onnx::Conv_822 = Identity(%onnx::Conv_771)
%onnx::Conv_819 = Identity(%onnx::Conv_771)
%onnx::Conv_816 = Identity(%onnx::Conv_774)
%onnx::Conv_813 = Identity(%onnx::Conv_774)
%onnx::Conv_810 = Identity(%onnx::Conv_774)
%onnx::Conv_807 = Identity(%onnx::Conv_774)
%onnx::Conv_804 = Identity(%onnx::Conv_774)
%onnx::Conv_801 = Identity(%onnx::Conv_774)
%onnx::Conv_798 = Identity(%onnx::Conv_774)
%onnx::Conv_795 = Identity(%onnx::Conv_774)
%onnx::Conv_792 = Identity(%onnx::Conv_774)
%onnx::Conv_789 = Identity(%onnx::Conv_774)
%onnx::Conv_786 = Identity(%onnx::Conv_774)
%onnx::Conv_783 = Identity(%onnx::Conv_774)
%onnx::Conv_780 = Identity(%onnx::Conv_774)
%onnx::Conv_777 = Identity(%onnx::Conv_774)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_770, %onnx::Conv_771)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_773, %onnx::Conv_774)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_776, %onnx::Conv_777)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_779, %onnx::Conv_780)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/Add_2_output_0, %/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_782, %onnx::Conv_783)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/Add_5_output_0 = Add(%/layers.1/Add_4_output_0, %/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_5_output_0, %onnx::Conv_785, %onnx::Conv_786)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_788, %onnx::Conv_789)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/Add_2_output_0, %/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_797, %onnx::Conv_798)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/Add_5_output_0 = Add(%/layers.2/Add_4_output_0, %/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_5_output_0, %onnx::Conv_800, %onnx::Conv_801)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_803, %onnx::Conv_804)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_806, %onnx::Conv_807)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_809, %onnx::Conv_810)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/Add_2_output_0, %/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_812, %onnx::Conv_813)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/Add_5_output_0 = Add(%/layers.3/Add_4_output_0, %/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_5_output_0, %onnx::Conv_815, %onnx::Conv_816)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_818, %onnx::Conv_819)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_821, %onnx::Conv_822)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_824, %onnx::Conv_825)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/Add_2_output_0, %/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_827, %onnx::Conv_828)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/Add_5_output_0 = Add(%/layers.5/Add_4_output_0, %/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_5_output_0, %onnx::Conv_830, %onnx::Conv_831)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_833, %onnx::Conv_834)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_836, %onnx::Conv_837)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_839, %onnx::Conv_840)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/Add_2_output_0, %/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_842, %onnx::Conv_843)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/Add_5_output_0 = Add(%/layers.6/Add_4_output_0, %/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_5_output_0, %onnx::Conv_845, %onnx::Conv_846)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_848, %onnx::Conv_849)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_851, %onnx::Conv_852)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_854, %onnx::Conv_855)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/Add_2_output_0, %/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_857, %onnx::Conv_858)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/Add_5_output_0 = Add(%/layers.7/Add_4_output_0, %/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_5_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/Add_2_output_0, %/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/Add_5_output_0 = Add(%/layers.9/Add_4_output_0, %/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_5_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/Add_2_output_0, %/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/Add_5_output_0 = Add(%/layers.10/Add_4_output_0, %/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_5_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/Add_2_output_0, %/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/Add_5_output_0 = Add(%/layers.11/Add_4_output_0, %/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_5_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%768 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %768
}
|
val_accuracy
| 92.257613
| 2,270,046,208
| 7,681,802
|
{'zcp_epe_nas': 142.9296008785049, 'zcp_fisher': 97.14608764648438, 'zcp_flops': 36320739328.0, 'zcp_grad_norm': 169.48231506347656, 'zcp_grasp': 77.068115234375, 'zcp_jacov': -16.05793515986079, 'zcp_l2_norm': 798.0038452148438, 'zcp_nwot': 221.37375427220633, 'zcp_params': 7681802.0, 'zcp_plain': -0.020390739664435, 'zcp_snip': 983.832763671875, 'zcp_synflow': 150.2111966354737, 'zcp_zen': 95.10018920898438, 'zcp_val_accuracy': 0.8724960088729851}
| |
NASBench101_204365
|
NASBench101
|
204365
|
7bc09e490bf0809a015978a5b2026e46
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_836[FLOAT, 128x3x3x3]
%onnx::Conv_837[FLOAT, 128]
%onnx::Conv_839[FLOAT, 43x128x1x1]
%onnx::Conv_840[FLOAT, 43]
%onnx::Conv_842[FLOAT, 43x43x1x1]
%onnx::Conv_845[FLOAT, 43x43x3x3]
%onnx::Conv_848[FLOAT, 42x42x3x3]
%onnx::Conv_849[FLOAT, 42]
%onnx::Conv_851[FLOAT, 42x42x3x3]
%onnx::Conv_854[FLOAT, 43x128x1x1]
%onnx::Conv_857[FLOAT, 43x43x1x1]
%onnx::Conv_860[FLOAT, 43x43x3x3]
%onnx::Conv_863[FLOAT, 42x42x3x3]
%onnx::Conv_866[FLOAT, 42x42x3x3]
%onnx::Conv_869[FLOAT, 43x128x1x1]
%onnx::Conv_872[FLOAT, 43x43x1x1]
%onnx::Conv_875[FLOAT, 43x43x3x3]
%onnx::Conv_878[FLOAT, 42x42x3x3]
%onnx::Conv_881[FLOAT, 42x42x3x3]
%onnx::Conv_884[FLOAT, 86x128x1x1]
%onnx::Conv_885[FLOAT, 86]
%onnx::Conv_887[FLOAT, 86x86x1x1]
%onnx::Conv_890[FLOAT, 85x85x3x3]
%onnx::Conv_891[FLOAT, 85]
%onnx::Conv_893[FLOAT, 85x85x3x3]
%onnx::Conv_896[FLOAT, 85x85x3x3]
%onnx::Conv_899[FLOAT, 86x256x1x1]
%onnx::Conv_902[FLOAT, 86x86x1x1]
%onnx::Conv_905[FLOAT, 85x85x3x3]
%onnx::Conv_908[FLOAT, 85x85x3x3]
%onnx::Conv_911[FLOAT, 85x85x3x3]
%onnx::Conv_914[FLOAT, 86x256x1x1]
%onnx::Conv_917[FLOAT, 86x86x1x1]
%onnx::Conv_920[FLOAT, 85x85x3x3]
%onnx::Conv_923[FLOAT, 85x85x3x3]
%onnx::Conv_926[FLOAT, 85x85x3x3]
%onnx::Conv_929[FLOAT, 171x256x1x1]
%onnx::Conv_930[FLOAT, 171]
%onnx::Conv_932[FLOAT, 171x171x1x1]
%onnx::Conv_935[FLOAT, 171x171x3x3]
%onnx::Conv_938[FLOAT, 170x170x3x3]
%onnx::Conv_939[FLOAT, 170]
%onnx::Conv_941[FLOAT, 170x170x3x3]
%onnx::Conv_944[FLOAT, 171x512x1x1]
%onnx::Conv_947[FLOAT, 171x171x1x1]
%onnx::Conv_950[FLOAT, 171x171x3x3]
%onnx::Conv_953[FLOAT, 170x170x3x3]
%onnx::Conv_956[FLOAT, 170x170x3x3]
%onnx::Conv_959[FLOAT, 171x512x1x1]
%onnx::Conv_962[FLOAT, 171x171x1x1]
%onnx::Conv_965[FLOAT, 171x171x3x3]
%onnx::Conv_968[FLOAT, 170x170x3x3]
%onnx::Conv_971[FLOAT, 170x170x3x3]
) {
%onnx::Conv_972 = Identity(%onnx::Conv_939)
%onnx::Conv_969 = Identity(%onnx::Conv_939)
%onnx::Conv_966 = Identity(%onnx::Conv_930)
%onnx::Conv_963 = Identity(%onnx::Conv_930)
%onnx::Conv_960 = Identity(%onnx::Conv_930)
%onnx::Conv_957 = Identity(%onnx::Conv_939)
%onnx::Conv_954 = Identity(%onnx::Conv_939)
%onnx::Conv_951 = Identity(%onnx::Conv_930)
%onnx::Conv_948 = Identity(%onnx::Conv_930)
%onnx::Conv_945 = Identity(%onnx::Conv_930)
%onnx::Conv_942 = Identity(%onnx::Conv_939)
%onnx::Conv_936 = Identity(%onnx::Conv_930)
%onnx::Conv_933 = Identity(%onnx::Conv_930)
%onnx::Conv_927 = Identity(%onnx::Conv_891)
%onnx::Conv_924 = Identity(%onnx::Conv_891)
%onnx::Conv_921 = Identity(%onnx::Conv_891)
%onnx::Conv_918 = Identity(%onnx::Conv_885)
%onnx::Conv_915 = Identity(%onnx::Conv_885)
%onnx::Conv_912 = Identity(%onnx::Conv_891)
%onnx::Conv_909 = Identity(%onnx::Conv_891)
%onnx::Conv_906 = Identity(%onnx::Conv_891)
%onnx::Conv_903 = Identity(%onnx::Conv_885)
%onnx::Conv_900 = Identity(%onnx::Conv_885)
%onnx::Conv_897 = Identity(%onnx::Conv_891)
%onnx::Conv_894 = Identity(%onnx::Conv_891)
%onnx::Conv_888 = Identity(%onnx::Conv_885)
%onnx::Conv_882 = Identity(%onnx::Conv_849)
%onnx::Conv_879 = Identity(%onnx::Conv_849)
%onnx::Conv_876 = Identity(%onnx::Conv_840)
%onnx::Conv_873 = Identity(%onnx::Conv_840)
%onnx::Conv_870 = Identity(%onnx::Conv_840)
%onnx::Conv_867 = Identity(%onnx::Conv_849)
%onnx::Conv_864 = Identity(%onnx::Conv_849)
%onnx::Conv_861 = Identity(%onnx::Conv_840)
%onnx::Conv_858 = Identity(%onnx::Conv_840)
%onnx::Conv_855 = Identity(%onnx::Conv_840)
%onnx::Conv_852 = Identity(%onnx::Conv_849)
%onnx::Conv_846 = Identity(%onnx::Conv_840)
%onnx::Conv_843 = Identity(%onnx::Conv_840)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_836, %onnx::Conv_837)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_839, %onnx::Conv_840)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_842, %onnx::Conv_843)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_845, %onnx::Conv_846)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.1/Slice_output_0 = Slice(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_4_output_0, %/layers.1/Constant_5_output_0, %/layers.1/Constant_3_output_0, %/layers.1/Constant_6_output_0)
%/layers.1/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/Slice_output_0, %/layers.1/Constant_7_output_0)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_848, %onnx::Conv_849)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_10_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_11_output_0 = Constant[value = <Tensor>]()
%/layers.1/Slice_1_output_0 = Slice(%/layers.1/vertex_op.3/maxpool/MaxPool_output_0, %/layers.1/Constant_9_output_0, %/layers.1/Constant_10_output_0, %/layers.1/Constant_8_output_0, %/layers.1/Constant_11_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/Slice_1_output_0, %/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_851, %onnx::Conv_852)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.3/maxpool/MaxPool_output_0, %/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_854, %onnx::Conv_855)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_857, %onnx::Conv_858)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.2/Slice_output_0 = Slice(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_4_output_0, %/layers.2/Constant_5_output_0, %/layers.2/Constant_3_output_0, %/layers.2/Constant_6_output_0)
%/layers.2/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/Slice_output_0, %/layers.2/Constant_7_output_0)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_10_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_11_output_0 = Constant[value = <Tensor>]()
%/layers.2/Slice_1_output_0 = Slice(%/layers.2/vertex_op.3/maxpool/MaxPool_output_0, %/layers.2/Constant_9_output_0, %/layers.2/Constant_10_output_0, %/layers.2/Constant_8_output_0, %/layers.2/Constant_11_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/Slice_1_output_0, %/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.3/maxpool/MaxPool_output_0, %/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.3/Slice_output_0 = Slice(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_4_output_0, %/layers.3/Constant_5_output_0, %/layers.3/Constant_3_output_0, %/layers.3/Constant_6_output_0)
%/layers.3/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/Slice_output_0, %/layers.3/Constant_7_output_0)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_10_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_11_output_0 = Constant[value = <Tensor>]()
%/layers.3/Slice_1_output_0 = Slice(%/layers.3/vertex_op.3/maxpool/MaxPool_output_0, %/layers.3/Constant_9_output_0, %/layers.3/Constant_10_output_0, %/layers.3/Constant_8_output_0, %/layers.3/Constant_11_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/Slice_1_output_0, %/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_4_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.3/maxpool/MaxPool_output_0, %/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.5/Slice_output_0 = Slice(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0, %/layers.5/Constant_3_output_0, %/layers.5/Constant_1_output_0, %/layers.5/Constant_4_output_0)
%/layers.5/Constant_5_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/Slice_output_0, %/layers.5/Constant_5_output_0)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_6_output_0)
%/layers.5/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0)
%/layers.5/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_7_output_0)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.3/maxpool/MaxPool_output_0, %/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.3/maxpool/MaxPool_output_0, %/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.6/Slice_output_0 = Slice(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0, %/layers.6/Constant_3_output_0, %/layers.6/Constant_1_output_0, %/layers.6/Constant_4_output_0)
%/layers.6/Constant_5_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/Slice_output_0, %/layers.6/Constant_5_output_0)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_6_output_0)
%/layers.6/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0)
%/layers.6/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_7_output_0)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_908, %onnx::Conv_909)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.3/maxpool/MaxPool_output_0, %/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_911, %onnx::Conv_912)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.3/maxpool/MaxPool_output_0, %/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_914, %onnx::Conv_915)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_917, %onnx::Conv_918)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.7/Slice_output_0 = Slice(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0, %/layers.7/Constant_3_output_0, %/layers.7/Constant_1_output_0, %/layers.7/Constant_4_output_0)
%/layers.7/Constant_5_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/Slice_output_0, %/layers.7/Constant_5_output_0)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_920, %onnx::Conv_921)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_6_output_0)
%/layers.7/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0)
%/layers.7/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_7_output_0)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_923, %onnx::Conv_924)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.3/maxpool/MaxPool_output_0, %/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_4_output_0, %onnx::Conv_926, %onnx::Conv_927)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.3/maxpool/MaxPool_output_0, %/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_929, %onnx::Conv_930)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_932, %onnx::Conv_933)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_935, %onnx::Conv_936)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.9/Slice_output_0 = Slice(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_4_output_0, %/layers.9/Constant_5_output_0, %/layers.9/Constant_3_output_0, %/layers.9/Constant_6_output_0)
%/layers.9/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/Slice_output_0, %/layers.9/Constant_7_output_0)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_938, %onnx::Conv_939)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_10_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_11_output_0 = Constant[value = <Tensor>]()
%/layers.9/Slice_1_output_0 = Slice(%/layers.9/vertex_op.3/maxpool/MaxPool_output_0, %/layers.9/Constant_9_output_0, %/layers.9/Constant_10_output_0, %/layers.9/Constant_8_output_0, %/layers.9/Constant_11_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/Slice_1_output_0, %/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_941, %onnx::Conv_942)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.3/maxpool/MaxPool_output_0, %/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_944, %onnx::Conv_945)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_947, %onnx::Conv_948)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_950, %onnx::Conv_951)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.10/Slice_output_0 = Slice(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_4_output_0, %/layers.10/Constant_5_output_0, %/layers.10/Constant_3_output_0, %/layers.10/Constant_6_output_0)
%/layers.10/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/Slice_output_0, %/layers.10/Constant_7_output_0)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_953, %onnx::Conv_954)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_10_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_11_output_0 = Constant[value = <Tensor>]()
%/layers.10/Slice_1_output_0 = Slice(%/layers.10/vertex_op.3/maxpool/MaxPool_output_0, %/layers.10/Constant_9_output_0, %/layers.10/Constant_10_output_0, %/layers.10/Constant_8_output_0, %/layers.10/Constant_11_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/Slice_1_output_0, %/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_956, %onnx::Conv_957)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.3/maxpool/MaxPool_output_0, %/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_959, %onnx::Conv_960)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_962, %onnx::Conv_963)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_965, %onnx::Conv_966)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.11/Slice_output_0 = Slice(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_4_output_0, %/layers.11/Constant_5_output_0, %/layers.11/Constant_3_output_0, %/layers.11/Constant_6_output_0)
%/layers.11/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/Slice_output_0, %/layers.11/Constant_7_output_0)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_968, %onnx::Conv_969)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_10_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_11_output_0 = Constant[value = <Tensor>]()
%/layers.11/Slice_1_output_0 = Slice(%/layers.11/vertex_op.3/maxpool/MaxPool_output_0, %/layers.11/Constant_9_output_0, %/layers.11/Constant_10_output_0, %/layers.11/Constant_8_output_0, %/layers.11/Constant_11_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/Slice_1_output_0, %/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_4_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.3/maxpool/MaxPool_output_0, %/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%834 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %834
}
|
val_accuracy
| 90.735179
| 1,039,043,328
| 3,504,201
|
{'zcp_epe_nas': 103.24018533652072, 'zcp_fisher': 168.61679077148438, 'zcp_flops': 16624693248.0, 'zcp_grad_norm': 253.93983459472656, 'zcp_grasp': 101.19921875, 'zcp_jacov': -16.067257355367012, 'zcp_l2_norm': 688.1243896484375, 'zcp_nwot': 215.84331031078486, 'zcp_params': 3504201.0, 'zcp_plain': 0.023081105202436003, 'zcp_snip': 1131.19482421875, 'zcp_synflow': 140.23396315438782, 'zcp_zen': 86.91487121582031, 'zcp_val_accuracy': 0.914763629436492}
| |
NASBench101_226723
|
NASBench101
|
226723
|
8957383c97f35fa0382139660c79dfee
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_761[FLOAT, 128x3x3x3]
%onnx::Conv_762[FLOAT, 128]
%onnx::Conv_764[FLOAT, 64x128x1x1]
%onnx::Conv_765[FLOAT, 64]
%onnx::Conv_767[FLOAT, 64x64x1x1]
%onnx::Conv_770[FLOAT, 64x128x1x1]
%onnx::Conv_773[FLOAT, 64x64x1x1]
%onnx::Conv_776[FLOAT, 64x64x1x1]
%onnx::Conv_779[FLOAT, 64x128x1x1]
%onnx::Conv_782[FLOAT, 64x64x1x1]
%onnx::Conv_785[FLOAT, 64x128x1x1]
%onnx::Conv_788[FLOAT, 64x64x1x1]
%onnx::Conv_791[FLOAT, 64x64x1x1]
%onnx::Conv_794[FLOAT, 64x128x1x1]
%onnx::Conv_797[FLOAT, 64x64x1x1]
%onnx::Conv_800[FLOAT, 64x128x1x1]
%onnx::Conv_803[FLOAT, 64x64x1x1]
%onnx::Conv_806[FLOAT, 64x64x1x1]
%onnx::Conv_809[FLOAT, 128x128x1x1]
%onnx::Conv_812[FLOAT, 128x128x1x1]
%onnx::Conv_815[FLOAT, 128x128x1x1]
%onnx::Conv_818[FLOAT, 128x128x1x1]
%onnx::Conv_821[FLOAT, 128x128x1x1]
%onnx::Conv_824[FLOAT, 128x256x1x1]
%onnx::Conv_827[FLOAT, 128x128x1x1]
%onnx::Conv_830[FLOAT, 128x256x1x1]
%onnx::Conv_833[FLOAT, 128x128x1x1]
%onnx::Conv_836[FLOAT, 128x128x1x1]
%onnx::Conv_839[FLOAT, 128x256x1x1]
%onnx::Conv_842[FLOAT, 128x128x1x1]
%onnx::Conv_845[FLOAT, 128x256x1x1]
%onnx::Conv_848[FLOAT, 128x128x1x1]
%onnx::Conv_851[FLOAT, 128x128x1x1]
%onnx::Conv_854[FLOAT, 256x256x1x1]
%onnx::Conv_855[FLOAT, 256]
%onnx::Conv_857[FLOAT, 256x256x1x1]
%onnx::Conv_860[FLOAT, 256x256x1x1]
%onnx::Conv_863[FLOAT, 256x256x1x1]
%onnx::Conv_866[FLOAT, 256x256x1x1]
%onnx::Conv_869[FLOAT, 256x512x1x1]
%onnx::Conv_872[FLOAT, 256x256x1x1]
%onnx::Conv_875[FLOAT, 256x512x1x1]
%onnx::Conv_878[FLOAT, 256x256x1x1]
%onnx::Conv_881[FLOAT, 256x256x1x1]
%onnx::Conv_884[FLOAT, 256x512x1x1]
%onnx::Conv_887[FLOAT, 256x256x1x1]
%onnx::Conv_890[FLOAT, 256x512x1x1]
%onnx::Conv_893[FLOAT, 256x256x1x1]
%onnx::Conv_896[FLOAT, 256x256x1x1]
) {
%onnx::Conv_897 = Identity(%onnx::Conv_855)
%onnx::Conv_894 = Identity(%onnx::Conv_855)
%onnx::Conv_891 = Identity(%onnx::Conv_855)
%onnx::Conv_888 = Identity(%onnx::Conv_855)
%onnx::Conv_885 = Identity(%onnx::Conv_855)
%onnx::Conv_882 = Identity(%onnx::Conv_855)
%onnx::Conv_879 = Identity(%onnx::Conv_855)
%onnx::Conv_876 = Identity(%onnx::Conv_855)
%onnx::Conv_873 = Identity(%onnx::Conv_855)
%onnx::Conv_870 = Identity(%onnx::Conv_855)
%onnx::Conv_867 = Identity(%onnx::Conv_855)
%onnx::Conv_864 = Identity(%onnx::Conv_855)
%onnx::Conv_861 = Identity(%onnx::Conv_855)
%onnx::Conv_858 = Identity(%onnx::Conv_855)
%onnx::Conv_852 = Identity(%onnx::Conv_762)
%onnx::Conv_849 = Identity(%onnx::Conv_762)
%onnx::Conv_846 = Identity(%onnx::Conv_762)
%onnx::Conv_843 = Identity(%onnx::Conv_762)
%onnx::Conv_840 = Identity(%onnx::Conv_762)
%onnx::Conv_837 = Identity(%onnx::Conv_762)
%onnx::Conv_834 = Identity(%onnx::Conv_762)
%onnx::Conv_831 = Identity(%onnx::Conv_762)
%onnx::Conv_828 = Identity(%onnx::Conv_762)
%onnx::Conv_825 = Identity(%onnx::Conv_762)
%onnx::Conv_822 = Identity(%onnx::Conv_762)
%onnx::Conv_819 = Identity(%onnx::Conv_762)
%onnx::Conv_816 = Identity(%onnx::Conv_762)
%onnx::Conv_813 = Identity(%onnx::Conv_762)
%onnx::Conv_810 = Identity(%onnx::Conv_762)
%onnx::Conv_807 = Identity(%onnx::Conv_765)
%onnx::Conv_804 = Identity(%onnx::Conv_765)
%onnx::Conv_801 = Identity(%onnx::Conv_765)
%onnx::Conv_798 = Identity(%onnx::Conv_765)
%onnx::Conv_795 = Identity(%onnx::Conv_765)
%onnx::Conv_792 = Identity(%onnx::Conv_765)
%onnx::Conv_789 = Identity(%onnx::Conv_765)
%onnx::Conv_786 = Identity(%onnx::Conv_765)
%onnx::Conv_783 = Identity(%onnx::Conv_765)
%onnx::Conv_780 = Identity(%onnx::Conv_765)
%onnx::Conv_777 = Identity(%onnx::Conv_765)
%onnx::Conv_774 = Identity(%onnx::Conv_765)
%onnx::Conv_771 = Identity(%onnx::Conv_765)
%onnx::Conv_768 = Identity(%onnx::Conv_765)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_761, %onnx::Conv_762)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_764, %onnx::Conv_765)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_767, %onnx::Conv_768)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_770, %onnx::Conv_771)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/Add_1_output_0, %/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_773, %onnx::Conv_774)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/Add_3_output_0, %/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_4_output_0)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_776, %onnx::Conv_777)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_779, %onnx::Conv_780)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_782, %onnx::Conv_783)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_785, %onnx::Conv_786)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/Add_1_output_0, %/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_788, %onnx::Conv_789)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/Add_3_output_0, %/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_4_output_0)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_797, %onnx::Conv_798)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_800, %onnx::Conv_801)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/Add_1_output_0, %/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_803, %onnx::Conv_804)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/Add_3_output_0, %/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_4_output_0)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_806, %onnx::Conv_807)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_809, %onnx::Conv_810)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_812, %onnx::Conv_813)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_815, %onnx::Conv_816)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/Add_1_output_0, %/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_818, %onnx::Conv_819)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/Add_3_output_0, %/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_4_output_0)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_821, %onnx::Conv_822)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_824, %onnx::Conv_825)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_827, %onnx::Conv_828)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_830, %onnx::Conv_831)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/Add_1_output_0, %/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_833, %onnx::Conv_834)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/Add_3_output_0, %/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_4_output_0)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_836, %onnx::Conv_837)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_839, %onnx::Conv_840)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_842, %onnx::Conv_843)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_845, %onnx::Conv_846)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/Add_1_output_0, %/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_848, %onnx::Conv_849)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/Add_3_output_0, %/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_4_output_0)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_851, %onnx::Conv_852)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_854, %onnx::Conv_855)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_857, %onnx::Conv_858)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/Add_1_output_0, %/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/Add_3_output_0, %/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_4_output_0)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/Add_1_output_0, %/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/Add_3_output_0, %/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_4_output_0)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/Add_1_output_0, %/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/Add_3_output_0, %/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_4_output_0)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%759 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %759
}
|
val_accuracy
| 87.870592
| 516,827,136
| 1,664,778
|
{'zcp_epe_nas': 140.6617354003023, 'zcp_fisher': 91.23956298828125, 'zcp_flops': 8269234176.0, 'zcp_grad_norm': 197.9735565185547, 'zcp_grasp': 8.5654296875, 'zcp_jacov': -16.046158754126722, 'zcp_l2_norm': 842.7713012695312, 'zcp_nwot': 221.7564059794626, 'zcp_params': 1664778.0, 'zcp_plain': 0.064525194466114, 'zcp_snip': 957.8521728515625, 'zcp_synflow': 99.46511796428425, 'zcp_zen': 70.62129211425781, 'zcp_val_accuracy': 0.890424668788909}
| |
NASBench101_286504
|
NASBench101
|
286504
|
ad6e24459ab41df5f5db41edc5ba1ea7
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_695[FLOAT, 128x3x3x3]
%onnx::Conv_696[FLOAT, 128]
%onnx::Conv_698[FLOAT, 43x128x1x1]
%onnx::Conv_699[FLOAT, 43]
%onnx::Conv_701[FLOAT, 43x43x3x3]
%onnx::Conv_704[FLOAT, 42x128x1x1]
%onnx::Conv_705[FLOAT, 42]
%onnx::Conv_707[FLOAT, 42x42x3x3]
%onnx::Conv_710[FLOAT, 43x128x1x1]
%onnx::Conv_713[FLOAT, 43x43x3x3]
%onnx::Conv_716[FLOAT, 42x128x1x1]
%onnx::Conv_719[FLOAT, 42x42x3x3]
%onnx::Conv_722[FLOAT, 43x128x1x1]
%onnx::Conv_725[FLOAT, 43x43x3x3]
%onnx::Conv_728[FLOAT, 42x128x1x1]
%onnx::Conv_731[FLOAT, 42x42x3x3]
%onnx::Conv_734[FLOAT, 86x128x1x1]
%onnx::Conv_735[FLOAT, 86]
%onnx::Conv_737[FLOAT, 86x86x3x3]
%onnx::Conv_740[FLOAT, 85x128x1x1]
%onnx::Conv_741[FLOAT, 85]
%onnx::Conv_743[FLOAT, 85x85x3x3]
%onnx::Conv_746[FLOAT, 86x256x1x1]
%onnx::Conv_749[FLOAT, 86x86x3x3]
%onnx::Conv_752[FLOAT, 85x256x1x1]
%onnx::Conv_755[FLOAT, 85x85x3x3]
%onnx::Conv_758[FLOAT, 86x256x1x1]
%onnx::Conv_761[FLOAT, 86x86x3x3]
%onnx::Conv_764[FLOAT, 85x256x1x1]
%onnx::Conv_767[FLOAT, 85x85x3x3]
%onnx::Conv_770[FLOAT, 171x256x1x1]
%onnx::Conv_771[FLOAT, 171]
%onnx::Conv_773[FLOAT, 171x171x3x3]
%onnx::Conv_776[FLOAT, 170x256x1x1]
%onnx::Conv_777[FLOAT, 170]
%onnx::Conv_779[FLOAT, 170x170x3x3]
%onnx::Conv_782[FLOAT, 171x512x1x1]
%onnx::Conv_785[FLOAT, 171x171x3x3]
%onnx::Conv_788[FLOAT, 170x512x1x1]
%onnx::Conv_791[FLOAT, 170x170x3x3]
%onnx::Conv_794[FLOAT, 171x512x1x1]
%onnx::Conv_797[FLOAT, 171x171x3x3]
%onnx::Conv_800[FLOAT, 170x512x1x1]
%onnx::Conv_803[FLOAT, 170x170x3x3]
) {
%onnx::Conv_804 = Identity(%onnx::Conv_777)
%onnx::Conv_801 = Identity(%onnx::Conv_777)
%onnx::Conv_798 = Identity(%onnx::Conv_771)
%onnx::Conv_795 = Identity(%onnx::Conv_771)
%onnx::Conv_792 = Identity(%onnx::Conv_777)
%onnx::Conv_789 = Identity(%onnx::Conv_777)
%onnx::Conv_786 = Identity(%onnx::Conv_771)
%onnx::Conv_783 = Identity(%onnx::Conv_771)
%onnx::Conv_780 = Identity(%onnx::Conv_777)
%onnx::Conv_774 = Identity(%onnx::Conv_771)
%onnx::Conv_768 = Identity(%onnx::Conv_741)
%onnx::Conv_765 = Identity(%onnx::Conv_741)
%onnx::Conv_762 = Identity(%onnx::Conv_735)
%onnx::Conv_759 = Identity(%onnx::Conv_735)
%onnx::Conv_756 = Identity(%onnx::Conv_741)
%onnx::Conv_753 = Identity(%onnx::Conv_741)
%onnx::Conv_750 = Identity(%onnx::Conv_735)
%onnx::Conv_747 = Identity(%onnx::Conv_735)
%onnx::Conv_744 = Identity(%onnx::Conv_741)
%onnx::Conv_738 = Identity(%onnx::Conv_735)
%onnx::Conv_732 = Identity(%onnx::Conv_705)
%onnx::Conv_729 = Identity(%onnx::Conv_705)
%onnx::Conv_726 = Identity(%onnx::Conv_699)
%onnx::Conv_723 = Identity(%onnx::Conv_699)
%onnx::Conv_720 = Identity(%onnx::Conv_705)
%onnx::Conv_717 = Identity(%onnx::Conv_705)
%onnx::Conv_714 = Identity(%onnx::Conv_699)
%onnx::Conv_711 = Identity(%onnx::Conv_699)
%onnx::Conv_708 = Identity(%onnx::Conv_705)
%onnx::Conv_702 = Identity(%onnx::Conv_699)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_695, %onnx::Conv_696)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_698, %onnx::Conv_699)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_701, %onnx::Conv_702)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.1/Slice_output_0 = Slice(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0, %/layers.1/Constant_4_output_0, %/layers.1/Constant_2_output_0, %/layers.1/Constant_5_output_0)
%/layers.1/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/Slice_output_0, %/layers.1/Constant_6_output_0)
%/layers.1/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_704, %onnx::Conv_705)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.3/maxpool/MaxPool_output_0, %/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_707, %onnx::Conv_708)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_710, %onnx::Conv_711)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_713, %onnx::Conv_714)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.2/Slice_output_0 = Slice(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0, %/layers.2/Constant_4_output_0, %/layers.2/Constant_2_output_0, %/layers.2/Constant_5_output_0)
%/layers.2/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/Slice_output_0, %/layers.2/Constant_6_output_0)
%/layers.2/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_716, %onnx::Conv_717)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.3/maxpool/MaxPool_output_0, %/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_719, %onnx::Conv_720)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_722, %onnx::Conv_723)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_725, %onnx::Conv_726)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.3/Slice_output_0 = Slice(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0, %/layers.3/Constant_4_output_0, %/layers.3/Constant_2_output_0, %/layers.3/Constant_5_output_0)
%/layers.3/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/Slice_output_0, %/layers.3/Constant_6_output_0)
%/layers.3/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_728, %onnx::Conv_729)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.3/maxpool/MaxPool_output_0, %/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_731, %onnx::Conv_732)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_734, %onnx::Conv_735)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_737, %onnx::Conv_738)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.5/Slice_output_0 = Slice(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0, %/layers.5/Constant_3_output_0, %/layers.5/Constant_1_output_0, %/layers.5/Constant_4_output_0)
%/layers.5/Constant_5_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/Slice_output_0, %/layers.5/Constant_5_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.5/Slice_1_output_0 = Slice(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_7_output_0, %/layers.5/Constant_8_output_0, %/layers.5/Constant_6_output_0, %/layers.5/Constant_9_output_0)
%/layers.5/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/Slice_1_output_0, %/layers.5/Constant_10_output_0)
%/layers.5/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_740, %onnx::Conv_741)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.3/maxpool/MaxPool_output_0, %/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_743, %onnx::Conv_744)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_746, %onnx::Conv_747)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_749, %onnx::Conv_750)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.6/Slice_output_0 = Slice(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0, %/layers.6/Constant_3_output_0, %/layers.6/Constant_1_output_0, %/layers.6/Constant_4_output_0)
%/layers.6/Constant_5_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/Slice_output_0, %/layers.6/Constant_5_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.6/Slice_1_output_0 = Slice(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_7_output_0, %/layers.6/Constant_8_output_0, %/layers.6/Constant_6_output_0, %/layers.6/Constant_9_output_0)
%/layers.6/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/Slice_1_output_0, %/layers.6/Constant_10_output_0)
%/layers.6/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_752, %onnx::Conv_753)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.3/maxpool/MaxPool_output_0, %/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_755, %onnx::Conv_756)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_758, %onnx::Conv_759)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_761, %onnx::Conv_762)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.7/Slice_output_0 = Slice(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0, %/layers.7/Constant_3_output_0, %/layers.7/Constant_1_output_0, %/layers.7/Constant_4_output_0)
%/layers.7/Constant_5_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/Slice_output_0, %/layers.7/Constant_5_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.7/Slice_1_output_0 = Slice(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_7_output_0, %/layers.7/Constant_8_output_0, %/layers.7/Constant_6_output_0, %/layers.7/Constant_9_output_0)
%/layers.7/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/Slice_1_output_0, %/layers.7/Constant_10_output_0)
%/layers.7/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_764, %onnx::Conv_765)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.3/maxpool/MaxPool_output_0, %/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_767, %onnx::Conv_768)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_770, %onnx::Conv_771)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_773, %onnx::Conv_774)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.9/Slice_output_0 = Slice(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0, %/layers.9/Constant_4_output_0, %/layers.9/Constant_2_output_0, %/layers.9/Constant_5_output_0)
%/layers.9/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/Slice_output_0, %/layers.9/Constant_6_output_0)
%/layers.9/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_776, %onnx::Conv_777)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.3/maxpool/MaxPool_output_0, %/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_779, %onnx::Conv_780)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_782, %onnx::Conv_783)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_785, %onnx::Conv_786)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.10/Slice_output_0 = Slice(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0, %/layers.10/Constant_4_output_0, %/layers.10/Constant_2_output_0, %/layers.10/Constant_5_output_0)
%/layers.10/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/Slice_output_0, %/layers.10/Constant_6_output_0)
%/layers.10/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_788, %onnx::Conv_789)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.3/maxpool/MaxPool_output_0, %/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_797, %onnx::Conv_798)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.11/Slice_output_0 = Slice(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0, %/layers.11/Constant_4_output_0, %/layers.11/Constant_2_output_0, %/layers.11/Constant_5_output_0)
%/layers.11/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/Slice_output_0, %/layers.11/Constant_6_output_0)
%/layers.11/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_800, %onnx::Conv_801)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.3/maxpool/MaxPool_output_0, %/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_803, %onnx::Conv_804)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%693 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %693
}
|
val_accuracy
| 92.247593
| 797,053,312
| 2,656,691
|
{'zcp_epe_nas': 75.41463699187122, 'zcp_fisher': 10.055777549743652, 'zcp_flops': 12752852992.0, 'zcp_grad_norm': 54.779605865478516, 'zcp_grasp': 0.5238037109375, 'zcp_jacov': -16.048697175764175, 'zcp_l2_norm': 639.0526123046875, 'zcp_nwot': 212.42708572738198, 'zcp_params': 2656691.0, 'zcp_plain': 0.021505499258637, 'zcp_snip': 286.9241943359375, 'zcp_synflow': 89.13954565968479, 'zcp_zen': 72.6868896484375, 'zcp_val_accuracy': 0.8810096383094781}
| |
NASBench101_416007
|
NASBench101
|
416007
|
fb63fae13a3756649a4079c6a0019357
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_959[FLOAT, 128x3x3x3]
%onnx::Conv_960[FLOAT, 128]
%onnx::Conv_962[FLOAT, 43x128x1x1]
%onnx::Conv_963[FLOAT, 43]
%onnx::Conv_965[FLOAT, 43x43x1x1]
%onnx::Conv_968[FLOAT, 43x43x3x3]
%onnx::Conv_971[FLOAT, 43x43x1x1]
%onnx::Conv_974[FLOAT, 42x42x3x3]
%onnx::Conv_975[FLOAT, 42]
%onnx::Conv_977[FLOAT, 42x42x3x3]
%onnx::Conv_980[FLOAT, 43x128x1x1]
%onnx::Conv_983[FLOAT, 43x43x1x1]
%onnx::Conv_986[FLOAT, 43x43x3x3]
%onnx::Conv_989[FLOAT, 43x43x1x1]
%onnx::Conv_992[FLOAT, 42x42x3x3]
%onnx::Conv_995[FLOAT, 42x42x3x3]
%onnx::Conv_998[FLOAT, 43x128x1x1]
%onnx::Conv_1001[FLOAT, 43x43x1x1]
%onnx::Conv_1004[FLOAT, 43x43x3x3]
%onnx::Conv_1007[FLOAT, 43x43x1x1]
%onnx::Conv_1010[FLOAT, 42x42x3x3]
%onnx::Conv_1013[FLOAT, 42x42x3x3]
%onnx::Conv_1016[FLOAT, 86x128x1x1]
%onnx::Conv_1017[FLOAT, 86]
%onnx::Conv_1019[FLOAT, 86x86x1x1]
%onnx::Conv_1022[FLOAT, 86x86x3x3]
%onnx::Conv_1025[FLOAT, 85x85x1x1]
%onnx::Conv_1026[FLOAT, 85]
%onnx::Conv_1028[FLOAT, 85x85x3x3]
%onnx::Conv_1031[FLOAT, 85x85x3x3]
%onnx::Conv_1034[FLOAT, 86x256x1x1]
%onnx::Conv_1037[FLOAT, 86x86x1x1]
%onnx::Conv_1040[FLOAT, 86x86x3x3]
%onnx::Conv_1043[FLOAT, 85x85x1x1]
%onnx::Conv_1046[FLOAT, 85x85x3x3]
%onnx::Conv_1049[FLOAT, 85x85x3x3]
%onnx::Conv_1052[FLOAT, 86x256x1x1]
%onnx::Conv_1055[FLOAT, 86x86x1x1]
%onnx::Conv_1058[FLOAT, 86x86x3x3]
%onnx::Conv_1061[FLOAT, 85x85x1x1]
%onnx::Conv_1064[FLOAT, 85x85x3x3]
%onnx::Conv_1067[FLOAT, 85x85x3x3]
%onnx::Conv_1070[FLOAT, 171x256x1x1]
%onnx::Conv_1071[FLOAT, 171]
%onnx::Conv_1073[FLOAT, 171x171x1x1]
%onnx::Conv_1076[FLOAT, 171x171x3x3]
%onnx::Conv_1079[FLOAT, 171x171x1x1]
%onnx::Conv_1082[FLOAT, 170x170x3x3]
%onnx::Conv_1083[FLOAT, 170]
%onnx::Conv_1085[FLOAT, 170x170x3x3]
%onnx::Conv_1088[FLOAT, 171x512x1x1]
%onnx::Conv_1091[FLOAT, 171x171x1x1]
%onnx::Conv_1094[FLOAT, 171x171x3x3]
%onnx::Conv_1097[FLOAT, 171x171x1x1]
%onnx::Conv_1100[FLOAT, 170x170x3x3]
%onnx::Conv_1103[FLOAT, 170x170x3x3]
%onnx::Conv_1106[FLOAT, 171x512x1x1]
%onnx::Conv_1109[FLOAT, 171x171x1x1]
%onnx::Conv_1112[FLOAT, 171x171x3x3]
%onnx::Conv_1115[FLOAT, 171x171x1x1]
%onnx::Conv_1118[FLOAT, 170x170x3x3]
%onnx::Conv_1121[FLOAT, 170x170x3x3]
) {
%onnx::Conv_1122 = Identity(%onnx::Conv_1083)
%onnx::Conv_1119 = Identity(%onnx::Conv_1083)
%onnx::Conv_1116 = Identity(%onnx::Conv_1071)
%onnx::Conv_1113 = Identity(%onnx::Conv_1071)
%onnx::Conv_1110 = Identity(%onnx::Conv_1071)
%onnx::Conv_1107 = Identity(%onnx::Conv_1071)
%onnx::Conv_1104 = Identity(%onnx::Conv_1083)
%onnx::Conv_1101 = Identity(%onnx::Conv_1083)
%onnx::Conv_1098 = Identity(%onnx::Conv_1071)
%onnx::Conv_1095 = Identity(%onnx::Conv_1071)
%onnx::Conv_1092 = Identity(%onnx::Conv_1071)
%onnx::Conv_1089 = Identity(%onnx::Conv_1071)
%onnx::Conv_1086 = Identity(%onnx::Conv_1083)
%onnx::Conv_1080 = Identity(%onnx::Conv_1071)
%onnx::Conv_1077 = Identity(%onnx::Conv_1071)
%onnx::Conv_1074 = Identity(%onnx::Conv_1071)
%onnx::Conv_1068 = Identity(%onnx::Conv_1026)
%onnx::Conv_1065 = Identity(%onnx::Conv_1026)
%onnx::Conv_1062 = Identity(%onnx::Conv_1026)
%onnx::Conv_1059 = Identity(%onnx::Conv_1017)
%onnx::Conv_1056 = Identity(%onnx::Conv_1017)
%onnx::Conv_1053 = Identity(%onnx::Conv_1017)
%onnx::Conv_1050 = Identity(%onnx::Conv_1026)
%onnx::Conv_1047 = Identity(%onnx::Conv_1026)
%onnx::Conv_1044 = Identity(%onnx::Conv_1026)
%onnx::Conv_1041 = Identity(%onnx::Conv_1017)
%onnx::Conv_1038 = Identity(%onnx::Conv_1017)
%onnx::Conv_1035 = Identity(%onnx::Conv_1017)
%onnx::Conv_1032 = Identity(%onnx::Conv_1026)
%onnx::Conv_1029 = Identity(%onnx::Conv_1026)
%onnx::Conv_1023 = Identity(%onnx::Conv_1017)
%onnx::Conv_1020 = Identity(%onnx::Conv_1017)
%onnx::Conv_1014 = Identity(%onnx::Conv_975)
%onnx::Conv_1011 = Identity(%onnx::Conv_975)
%onnx::Conv_1008 = Identity(%onnx::Conv_963)
%onnx::Conv_1005 = Identity(%onnx::Conv_963)
%onnx::Conv_1002 = Identity(%onnx::Conv_963)
%onnx::Conv_999 = Identity(%onnx::Conv_963)
%onnx::Conv_996 = Identity(%onnx::Conv_975)
%onnx::Conv_993 = Identity(%onnx::Conv_975)
%onnx::Conv_990 = Identity(%onnx::Conv_963)
%onnx::Conv_987 = Identity(%onnx::Conv_963)
%onnx::Conv_984 = Identity(%onnx::Conv_963)
%onnx::Conv_981 = Identity(%onnx::Conv_963)
%onnx::Conv_978 = Identity(%onnx::Conv_975)
%onnx::Conv_972 = Identity(%onnx::Conv_963)
%onnx::Conv_969 = Identity(%onnx::Conv_963)
%onnx::Conv_966 = Identity(%onnx::Conv_963)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_959, %onnx::Conv_960)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_962, %onnx::Conv_963)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_965, %onnx::Conv_966)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_968, %onnx::Conv_969)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.1/Slice_output_0 = Slice(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_4_output_0, %/layers.1/Constant_5_output_0, %/layers.1/Constant_3_output_0, %/layers.1/Constant_6_output_0)
%/layers.1/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_10_output_0 = Constant[value = <Tensor>]()
%/layers.1/Slice_1_output_0 = Slice(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_8_output_0, %/layers.1/Constant_9_output_0, %/layers.1/Constant_7_output_0, %/layers.1/Constant_10_output_0)
%/layers.1/Constant_11_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/Slice_output_0, %/layers.1/Constant_11_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/Add_3_output_0, %/layers.1/Slice_1_output_0)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_974, %onnx::Conv_975)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_12_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_5_output_0 = Add(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_12_output_0)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_5_output_0, %onnx::Conv_977, %onnx::Conv_978)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.2/Slice_output_0 = Slice(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_4_output_0, %/layers.2/Constant_5_output_0, %/layers.2/Constant_3_output_0, %/layers.2/Constant_6_output_0)
%/layers.2/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_10_output_0 = Constant[value = <Tensor>]()
%/layers.2/Slice_1_output_0 = Slice(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_8_output_0, %/layers.2/Constant_9_output_0, %/layers.2/Constant_7_output_0, %/layers.2/Constant_10_output_0)
%/layers.2/Constant_11_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/Slice_output_0, %/layers.2/Constant_11_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/Add_3_output_0, %/layers.2/Slice_1_output_0)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_12_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_5_output_0 = Add(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_12_output_0)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_5_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.3/Slice_output_0 = Slice(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_4_output_0, %/layers.3/Constant_5_output_0, %/layers.3/Constant_3_output_0, %/layers.3/Constant_6_output_0)
%/layers.3/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_10_output_0 = Constant[value = <Tensor>]()
%/layers.3/Slice_1_output_0 = Slice(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_8_output_0, %/layers.3/Constant_9_output_0, %/layers.3/Constant_7_output_0, %/layers.3/Constant_10_output_0)
%/layers.3/Constant_11_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/Slice_output_0, %/layers.3/Constant_11_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/Add_3_output_0, %/layers.3/Slice_1_output_0)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_4_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_12_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_5_output_0 = Add(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_12_output_0)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_5_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.5/Slice_output_0 = Slice(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0, %/layers.5/Constant_4_output_0, %/layers.5/Constant_2_output_0, %/layers.5/Constant_5_output_0)
%/layers.5/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/Slice_output_0, %/layers.5/Constant_6_output_0)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_1025, %onnx::Conv_1026)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_10_output_0 = Constant[value = <Tensor>]()
%/layers.5/Slice_1_output_0 = Slice(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_8_output_0, %/layers.5/Constant_9_output_0, %/layers.5/Constant_7_output_0, %/layers.5/Constant_10_output_0)
%/layers.5/Constant_11_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/Slice_1_output_0, %/layers.5/Constant_11_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/Add_3_output_0, %/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_1028, %onnx::Conv_1029)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_12_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_5_output_0 = Add(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_12_output_0)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_5_output_0, %onnx::Conv_1031, %onnx::Conv_1032)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_1034, %onnx::Conv_1035)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_1037, %onnx::Conv_1038)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_1040, %onnx::Conv_1041)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.6/Slice_output_0 = Slice(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0, %/layers.6/Constant_4_output_0, %/layers.6/Constant_2_output_0, %/layers.6/Constant_5_output_0)
%/layers.6/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/Slice_output_0, %/layers.6/Constant_6_output_0)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_1043, %onnx::Conv_1044)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_10_output_0 = Constant[value = <Tensor>]()
%/layers.6/Slice_1_output_0 = Slice(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_8_output_0, %/layers.6/Constant_9_output_0, %/layers.6/Constant_7_output_0, %/layers.6/Constant_10_output_0)
%/layers.6/Constant_11_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/Slice_1_output_0, %/layers.6/Constant_11_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/Add_3_output_0, %/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_1046, %onnx::Conv_1047)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_12_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_5_output_0 = Add(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_12_output_0)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_5_output_0, %onnx::Conv_1049, %onnx::Conv_1050)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_1052, %onnx::Conv_1053)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_1055, %onnx::Conv_1056)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_1058, %onnx::Conv_1059)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.7/Slice_output_0 = Slice(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0, %/layers.7/Constant_4_output_0, %/layers.7/Constant_2_output_0, %/layers.7/Constant_5_output_0)
%/layers.7/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/Slice_output_0, %/layers.7/Constant_6_output_0)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_1061, %onnx::Conv_1062)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_10_output_0 = Constant[value = <Tensor>]()
%/layers.7/Slice_1_output_0 = Slice(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_8_output_0, %/layers.7/Constant_9_output_0, %/layers.7/Constant_7_output_0, %/layers.7/Constant_10_output_0)
%/layers.7/Constant_11_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/Slice_1_output_0, %/layers.7/Constant_11_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/Add_3_output_0, %/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_4_output_0, %onnx::Conv_1064, %onnx::Conv_1065)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_12_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_5_output_0 = Add(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_12_output_0)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_5_output_0, %onnx::Conv_1067, %onnx::Conv_1068)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1070, %onnx::Conv_1071)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_1073, %onnx::Conv_1074)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_1076, %onnx::Conv_1077)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_1079, %onnx::Conv_1080)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.9/Slice_output_0 = Slice(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_4_output_0, %/layers.9/Constant_5_output_0, %/layers.9/Constant_3_output_0, %/layers.9/Constant_6_output_0)
%/layers.9/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_10_output_0 = Constant[value = <Tensor>]()
%/layers.9/Slice_1_output_0 = Slice(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_8_output_0, %/layers.9/Constant_9_output_0, %/layers.9/Constant_7_output_0, %/layers.9/Constant_10_output_0)
%/layers.9/Constant_11_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/Slice_output_0, %/layers.9/Constant_11_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/Add_3_output_0, %/layers.9/Slice_1_output_0)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_1082, %onnx::Conv_1083)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_12_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_5_output_0 = Add(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_12_output_0)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_5_output_0, %onnx::Conv_1085, %onnx::Conv_1086)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_1088, %onnx::Conv_1089)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_1091, %onnx::Conv_1092)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_1094, %onnx::Conv_1095)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_1097, %onnx::Conv_1098)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.10/Slice_output_0 = Slice(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_4_output_0, %/layers.10/Constant_5_output_0, %/layers.10/Constant_3_output_0, %/layers.10/Constant_6_output_0)
%/layers.10/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_10_output_0 = Constant[value = <Tensor>]()
%/layers.10/Slice_1_output_0 = Slice(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_8_output_0, %/layers.10/Constant_9_output_0, %/layers.10/Constant_7_output_0, %/layers.10/Constant_10_output_0)
%/layers.10/Constant_11_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/Slice_output_0, %/layers.10/Constant_11_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/Add_3_output_0, %/layers.10/Slice_1_output_0)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_1100, %onnx::Conv_1101)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_12_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_5_output_0 = Add(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_12_output_0)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_5_output_0, %onnx::Conv_1103, %onnx::Conv_1104)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1106, %onnx::Conv_1107)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_1109, %onnx::Conv_1110)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_1112, %onnx::Conv_1113)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_1115, %onnx::Conv_1116)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.11/Slice_output_0 = Slice(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_4_output_0, %/layers.11/Constant_5_output_0, %/layers.11/Constant_3_output_0, %/layers.11/Constant_6_output_0)
%/layers.11/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_10_output_0 = Constant[value = <Tensor>]()
%/layers.11/Slice_1_output_0 = Slice(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_8_output_0, %/layers.11/Constant_9_output_0, %/layers.11/Constant_7_output_0, %/layers.11/Constant_10_output_0)
%/layers.11/Constant_11_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/Slice_output_0, %/layers.11/Constant_11_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/Add_3_output_0, %/layers.11/Slice_1_output_0)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_4_output_0, %onnx::Conv_1118, %onnx::Conv_1119)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_12_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_5_output_0 = Add(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_12_output_0)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_5_output_0, %onnx::Conv_1121, %onnx::Conv_1122)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%957 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %957
}
|
val_accuracy
| 91.676682
| 1,076,941,440
| 3,625,563
|
{'zcp_epe_nas': 97.97334621616237, 'zcp_fisher': 141.3499755859375, 'zcp_flops': 17231063040.0, 'zcp_grad_norm': 255.17889404296875, 'zcp_grasp': 859.361328125, 'zcp_jacov': -16.056367371681862, 'zcp_l2_norm': 811.5448608398438, 'zcp_nwot': 218.52359760945723, 'zcp_params': 3625563.0, 'zcp_plain': -0.0011834832839660002, 'zcp_snip': 1091.8756103515625, 'zcp_synflow': 161.30451278516563, 'zcp_zen': 92.00989532470703, 'zcp_val_accuracy': 0.8185096383094781}
| |
NASBench101_138724
|
NASBench101
|
138724
|
53dfdaa624501190a596d4b94c93b8a0
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_743[FLOAT, 128x3x3x3]
%onnx::Conv_744[FLOAT, 128]
%onnx::Conv_746[FLOAT, 64x128x1x1]
%onnx::Conv_747[FLOAT, 64]
%onnx::Conv_749[FLOAT, 64x64x1x1]
%onnx::Conv_752[FLOAT, 64x64x3x3]
%onnx::Conv_755[FLOAT, 64x128x1x1]
%onnx::Conv_758[FLOAT, 64x64x1x1]
%onnx::Conv_761[FLOAT, 64x128x1x1]
%onnx::Conv_764[FLOAT, 64x64x1x1]
%onnx::Conv_767[FLOAT, 64x64x3x3]
%onnx::Conv_770[FLOAT, 64x128x1x1]
%onnx::Conv_773[FLOAT, 64x64x1x1]
%onnx::Conv_776[FLOAT, 64x128x1x1]
%onnx::Conv_779[FLOAT, 64x64x1x1]
%onnx::Conv_782[FLOAT, 64x64x3x3]
%onnx::Conv_785[FLOAT, 64x128x1x1]
%onnx::Conv_788[FLOAT, 64x64x1x1]
%onnx::Conv_791[FLOAT, 128x128x1x1]
%onnx::Conv_794[FLOAT, 128x128x1x1]
%onnx::Conv_797[FLOAT, 128x128x3x3]
%onnx::Conv_800[FLOAT, 128x128x1x1]
%onnx::Conv_803[FLOAT, 128x128x1x1]
%onnx::Conv_806[FLOAT, 128x256x1x1]
%onnx::Conv_809[FLOAT, 128x128x1x1]
%onnx::Conv_812[FLOAT, 128x128x3x3]
%onnx::Conv_815[FLOAT, 128x256x1x1]
%onnx::Conv_818[FLOAT, 128x128x1x1]
%onnx::Conv_821[FLOAT, 128x256x1x1]
%onnx::Conv_824[FLOAT, 128x128x1x1]
%onnx::Conv_827[FLOAT, 128x128x3x3]
%onnx::Conv_830[FLOAT, 128x256x1x1]
%onnx::Conv_833[FLOAT, 128x128x1x1]
%onnx::Conv_836[FLOAT, 256x256x1x1]
%onnx::Conv_837[FLOAT, 256]
%onnx::Conv_839[FLOAT, 256x256x1x1]
%onnx::Conv_842[FLOAT, 256x256x3x3]
%onnx::Conv_845[FLOAT, 256x256x1x1]
%onnx::Conv_848[FLOAT, 256x256x1x1]
%onnx::Conv_851[FLOAT, 256x512x1x1]
%onnx::Conv_854[FLOAT, 256x256x1x1]
%onnx::Conv_857[FLOAT, 256x256x3x3]
%onnx::Conv_860[FLOAT, 256x512x1x1]
%onnx::Conv_863[FLOAT, 256x256x1x1]
%onnx::Conv_866[FLOAT, 256x512x1x1]
%onnx::Conv_869[FLOAT, 256x256x1x1]
%onnx::Conv_872[FLOAT, 256x256x3x3]
%onnx::Conv_875[FLOAT, 256x512x1x1]
%onnx::Conv_878[FLOAT, 256x256x1x1]
) {
%onnx::Conv_879 = Identity(%onnx::Conv_837)
%onnx::Conv_876 = Identity(%onnx::Conv_837)
%onnx::Conv_873 = Identity(%onnx::Conv_837)
%onnx::Conv_870 = Identity(%onnx::Conv_837)
%onnx::Conv_867 = Identity(%onnx::Conv_837)
%onnx::Conv_864 = Identity(%onnx::Conv_837)
%onnx::Conv_861 = Identity(%onnx::Conv_837)
%onnx::Conv_858 = Identity(%onnx::Conv_837)
%onnx::Conv_855 = Identity(%onnx::Conv_837)
%onnx::Conv_852 = Identity(%onnx::Conv_837)
%onnx::Conv_849 = Identity(%onnx::Conv_837)
%onnx::Conv_846 = Identity(%onnx::Conv_837)
%onnx::Conv_843 = Identity(%onnx::Conv_837)
%onnx::Conv_840 = Identity(%onnx::Conv_837)
%onnx::Conv_834 = Identity(%onnx::Conv_744)
%onnx::Conv_831 = Identity(%onnx::Conv_744)
%onnx::Conv_828 = Identity(%onnx::Conv_744)
%onnx::Conv_825 = Identity(%onnx::Conv_744)
%onnx::Conv_822 = Identity(%onnx::Conv_744)
%onnx::Conv_819 = Identity(%onnx::Conv_744)
%onnx::Conv_816 = Identity(%onnx::Conv_744)
%onnx::Conv_813 = Identity(%onnx::Conv_744)
%onnx::Conv_810 = Identity(%onnx::Conv_744)
%onnx::Conv_807 = Identity(%onnx::Conv_744)
%onnx::Conv_804 = Identity(%onnx::Conv_744)
%onnx::Conv_801 = Identity(%onnx::Conv_744)
%onnx::Conv_798 = Identity(%onnx::Conv_744)
%onnx::Conv_795 = Identity(%onnx::Conv_744)
%onnx::Conv_792 = Identity(%onnx::Conv_744)
%onnx::Conv_789 = Identity(%onnx::Conv_747)
%onnx::Conv_786 = Identity(%onnx::Conv_747)
%onnx::Conv_783 = Identity(%onnx::Conv_747)
%onnx::Conv_780 = Identity(%onnx::Conv_747)
%onnx::Conv_777 = Identity(%onnx::Conv_747)
%onnx::Conv_774 = Identity(%onnx::Conv_747)
%onnx::Conv_771 = Identity(%onnx::Conv_747)
%onnx::Conv_768 = Identity(%onnx::Conv_747)
%onnx::Conv_765 = Identity(%onnx::Conv_747)
%onnx::Conv_762 = Identity(%onnx::Conv_747)
%onnx::Conv_759 = Identity(%onnx::Conv_747)
%onnx::Conv_756 = Identity(%onnx::Conv_747)
%onnx::Conv_753 = Identity(%onnx::Conv_747)
%onnx::Conv_750 = Identity(%onnx::Conv_747)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_743, %onnx::Conv_744)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_746, %onnx::Conv_747)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_749, %onnx::Conv_750)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_752, %onnx::Conv_753)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_755, %onnx::Conv_756)
%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.3/maxpool/MaxPool_output_0, %/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/Add_2_output_0, %/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_758, %onnx::Conv_759)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.3/maxpool/MaxPool_output_0, %/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_761, %onnx::Conv_762)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_764, %onnx::Conv_765)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_767, %onnx::Conv_768)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_770, %onnx::Conv_771)
%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.3/maxpool/MaxPool_output_0, %/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/Add_2_output_0, %/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_773, %onnx::Conv_774)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.3/maxpool/MaxPool_output_0, %/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_776, %onnx::Conv_777)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_779, %onnx::Conv_780)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_782, %onnx::Conv_783)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_785, %onnx::Conv_786)
%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.3/maxpool/MaxPool_output_0, %/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/Add_2_output_0, %/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_788, %onnx::Conv_789)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.3/maxpool/MaxPool_output_0, %/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_797, %onnx::Conv_798)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_800, %onnx::Conv_801)
%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.3/maxpool/MaxPool_output_0, %/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/Add_2_output_0, %/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_803, %onnx::Conv_804)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.3/maxpool/MaxPool_output_0, %/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_806, %onnx::Conv_807)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_809, %onnx::Conv_810)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_812, %onnx::Conv_813)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_815, %onnx::Conv_816)
%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.3/maxpool/MaxPool_output_0, %/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/Add_2_output_0, %/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_818, %onnx::Conv_819)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.3/maxpool/MaxPool_output_0, %/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_821, %onnx::Conv_822)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_824, %onnx::Conv_825)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_827, %onnx::Conv_828)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_830, %onnx::Conv_831)
%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.3/maxpool/MaxPool_output_0, %/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/Add_2_output_0, %/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_833, %onnx::Conv_834)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.3/maxpool/MaxPool_output_0, %/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_836, %onnx::Conv_837)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_839, %onnx::Conv_840)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_842, %onnx::Conv_843)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_845, %onnx::Conv_846)
%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.3/maxpool/MaxPool_output_0, %/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/Add_2_output_0, %/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_848, %onnx::Conv_849)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.3/maxpool/MaxPool_output_0, %/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_851, %onnx::Conv_852)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_854, %onnx::Conv_855)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_857, %onnx::Conv_858)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.3/maxpool/MaxPool_output_0, %/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/Add_2_output_0, %/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.3/maxpool/MaxPool_output_0, %/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.3/maxpool/MaxPool_output_0, %/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/Add_2_output_0, %/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.3/maxpool/MaxPool_output_0, %/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%741 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %741
}
|
val_accuracy
| 90.404648
| 1,120,806,912
| 3,729,162
|
{'zcp_epe_nas': 99.94561915881827, 'zcp_fisher': 63.87828826904297, 'zcp_flops': 17932910592.0, 'zcp_grad_norm': 168.6416778564453, 'zcp_grasp': -23.8798828125, 'zcp_jacov': -16.064098336236732, 'zcp_l2_norm': 844.1124877929688, 'zcp_nwot': 221.8987484862255, 'zcp_params': 3729162.0, 'zcp_plain': 0.0029769521206610004, 'zcp_snip': 891.7140502929688, 'zcp_synflow': 107.46945412478875, 'zcp_zen': 81.59292602539062, 'zcp_val_accuracy': 0.9377003312110901}
| |
NASBench101_127860
|
NASBench101
|
127860
|
4d42803bcd709fb5c217b5de806d99da
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_770[FLOAT, 128x3x3x3]
%onnx::Conv_771[FLOAT, 128]
%onnx::Conv_773[FLOAT, 64x128x1x1]
%onnx::Conv_774[FLOAT, 64]
%onnx::Conv_776[FLOAT, 64x128x1x1]
%onnx::Conv_779[FLOAT, 64x64x3x3]
%onnx::Conv_782[FLOAT, 64x128x1x1]
%onnx::Conv_785[FLOAT, 64x64x3x3]
%onnx::Conv_788[FLOAT, 64x128x1x1]
%onnx::Conv_791[FLOAT, 64x128x1x1]
%onnx::Conv_794[FLOAT, 64x64x3x3]
%onnx::Conv_797[FLOAT, 64x128x1x1]
%onnx::Conv_800[FLOAT, 64x64x3x3]
%onnx::Conv_803[FLOAT, 64x128x1x1]
%onnx::Conv_806[FLOAT, 64x128x1x1]
%onnx::Conv_809[FLOAT, 64x64x3x3]
%onnx::Conv_812[FLOAT, 64x128x1x1]
%onnx::Conv_815[FLOAT, 64x64x3x3]
%onnx::Conv_818[FLOAT, 128x128x1x1]
%onnx::Conv_821[FLOAT, 128x128x1x1]
%onnx::Conv_824[FLOAT, 128x128x3x3]
%onnx::Conv_827[FLOAT, 128x128x1x1]
%onnx::Conv_830[FLOAT, 128x128x3x3]
%onnx::Conv_833[FLOAT, 128x256x1x1]
%onnx::Conv_836[FLOAT, 128x256x1x1]
%onnx::Conv_839[FLOAT, 128x128x3x3]
%onnx::Conv_842[FLOAT, 128x256x1x1]
%onnx::Conv_845[FLOAT, 128x128x3x3]
%onnx::Conv_848[FLOAT, 128x256x1x1]
%onnx::Conv_851[FLOAT, 128x256x1x1]
%onnx::Conv_854[FLOAT, 128x128x3x3]
%onnx::Conv_857[FLOAT, 128x256x1x1]
%onnx::Conv_860[FLOAT, 128x128x3x3]
%onnx::Conv_863[FLOAT, 256x256x1x1]
%onnx::Conv_864[FLOAT, 256]
%onnx::Conv_866[FLOAT, 256x256x1x1]
%onnx::Conv_869[FLOAT, 256x256x3x3]
%onnx::Conv_872[FLOAT, 256x256x1x1]
%onnx::Conv_875[FLOAT, 256x256x3x3]
%onnx::Conv_878[FLOAT, 256x512x1x1]
%onnx::Conv_881[FLOAT, 256x512x1x1]
%onnx::Conv_884[FLOAT, 256x256x3x3]
%onnx::Conv_887[FLOAT, 256x512x1x1]
%onnx::Conv_890[FLOAT, 256x256x3x3]
%onnx::Conv_893[FLOAT, 256x512x1x1]
%onnx::Conv_896[FLOAT, 256x512x1x1]
%onnx::Conv_899[FLOAT, 256x256x3x3]
%onnx::Conv_902[FLOAT, 256x512x1x1]
%onnx::Conv_905[FLOAT, 256x256x3x3]
) {
%onnx::Conv_906 = Identity(%onnx::Conv_864)
%onnx::Conv_903 = Identity(%onnx::Conv_864)
%onnx::Conv_900 = Identity(%onnx::Conv_864)
%onnx::Conv_897 = Identity(%onnx::Conv_864)
%onnx::Conv_894 = Identity(%onnx::Conv_864)
%onnx::Conv_891 = Identity(%onnx::Conv_864)
%onnx::Conv_888 = Identity(%onnx::Conv_864)
%onnx::Conv_885 = Identity(%onnx::Conv_864)
%onnx::Conv_882 = Identity(%onnx::Conv_864)
%onnx::Conv_879 = Identity(%onnx::Conv_864)
%onnx::Conv_876 = Identity(%onnx::Conv_864)
%onnx::Conv_873 = Identity(%onnx::Conv_864)
%onnx::Conv_870 = Identity(%onnx::Conv_864)
%onnx::Conv_867 = Identity(%onnx::Conv_864)
%onnx::Conv_861 = Identity(%onnx::Conv_771)
%onnx::Conv_858 = Identity(%onnx::Conv_771)
%onnx::Conv_855 = Identity(%onnx::Conv_771)
%onnx::Conv_852 = Identity(%onnx::Conv_771)
%onnx::Conv_849 = Identity(%onnx::Conv_771)
%onnx::Conv_846 = Identity(%onnx::Conv_771)
%onnx::Conv_843 = Identity(%onnx::Conv_771)
%onnx::Conv_840 = Identity(%onnx::Conv_771)
%onnx::Conv_837 = Identity(%onnx::Conv_771)
%onnx::Conv_834 = Identity(%onnx::Conv_771)
%onnx::Conv_831 = Identity(%onnx::Conv_771)
%onnx::Conv_828 = Identity(%onnx::Conv_771)
%onnx::Conv_825 = Identity(%onnx::Conv_771)
%onnx::Conv_822 = Identity(%onnx::Conv_771)
%onnx::Conv_819 = Identity(%onnx::Conv_771)
%onnx::Conv_816 = Identity(%onnx::Conv_774)
%onnx::Conv_813 = Identity(%onnx::Conv_774)
%onnx::Conv_810 = Identity(%onnx::Conv_774)
%onnx::Conv_807 = Identity(%onnx::Conv_774)
%onnx::Conv_804 = Identity(%onnx::Conv_774)
%onnx::Conv_801 = Identity(%onnx::Conv_774)
%onnx::Conv_798 = Identity(%onnx::Conv_774)
%onnx::Conv_795 = Identity(%onnx::Conv_774)
%onnx::Conv_792 = Identity(%onnx::Conv_774)
%onnx::Conv_789 = Identity(%onnx::Conv_774)
%onnx::Conv_786 = Identity(%onnx::Conv_774)
%onnx::Conv_783 = Identity(%onnx::Conv_774)
%onnx::Conv_780 = Identity(%onnx::Conv_774)
%onnx::Conv_777 = Identity(%onnx::Conv_774)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_770, %onnx::Conv_771)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_773, %onnx::Conv_774)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_776, %onnx::Conv_777)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_779, %onnx::Conv_780)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0)
%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_782, %onnx::Conv_783)
%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.4/maxpool/MaxPool_output_0, %/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_785, %onnx::Conv_786)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_788, %onnx::Conv_789)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0)
%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_797, %onnx::Conv_798)
%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.4/maxpool/MaxPool_output_0, %/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_800, %onnx::Conv_801)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_803, %onnx::Conv_804)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_806, %onnx::Conv_807)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_809, %onnx::Conv_810)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0)
%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_812, %onnx::Conv_813)
%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.4/maxpool/MaxPool_output_0, %/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_4_output_0, %onnx::Conv_815, %onnx::Conv_816)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_818, %onnx::Conv_819)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_821, %onnx::Conv_822)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_824, %onnx::Conv_825)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0)
%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_827, %onnx::Conv_828)
%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.4/maxpool/MaxPool_output_0, %/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_830, %onnx::Conv_831)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_833, %onnx::Conv_834)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_836, %onnx::Conv_837)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_839, %onnx::Conv_840)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0)
%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_842, %onnx::Conv_843)
%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.4/maxpool/MaxPool_output_0, %/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_845, %onnx::Conv_846)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_848, %onnx::Conv_849)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_851, %onnx::Conv_852)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_854, %onnx::Conv_855)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0)
%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_857, %onnx::Conv_858)
%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.4/maxpool/MaxPool_output_0, %/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_4_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0)
%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.4/maxpool/MaxPool_output_0, %/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0)
%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.4/maxpool/MaxPool_output_0, %/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0)
%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.4/maxpool/MaxPool_output_0, %/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_4_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%768 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %768
}
|
val_accuracy
| 92.658252
| 1,783,506,944
| 5,969,674
|
{'zcp_epe_nas': 129.01192167599697, 'zcp_fisher': 5.0245361328125, 'zcp_flops': 28536111104.0, 'zcp_grad_norm': 37.93091583251953, 'zcp_grasp': -0.3690185546875, 'zcp_jacov': -16.062170778114886, 'zcp_l2_norm': 889.690673828125, 'zcp_nwot': 220.8692707220673, 'zcp_params': 5969674.0, 'zcp_plain': 0.048190940171480005, 'zcp_snip': 253.60987854003906, 'zcp_synflow': 100.50208577423358, 'zcp_zen': 93.67973327636719, 'zcp_val_accuracy': 0.8917267918586731}
| |
NASBench101_151076
|
NASBench101
|
151076
|
5b634198fb8367a95bbce9e4f55df675
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_761[FLOAT, 128x3x3x3]
%onnx::Conv_762[FLOAT, 128]
%onnx::Conv_764[FLOAT, 128x128x1x1]
%onnx::Conv_767[FLOAT, 128x128x1x1]
%onnx::Conv_770[FLOAT, 128x128x1x1]
%onnx::Conv_773[FLOAT, 128x128x3x3]
%onnx::Conv_776[FLOAT, 128x128x3x3]
%onnx::Conv_779[FLOAT, 128x128x1x1]
%onnx::Conv_782[FLOAT, 128x128x1x1]
%onnx::Conv_785[FLOAT, 128x128x1x1]
%onnx::Conv_788[FLOAT, 128x128x3x3]
%onnx::Conv_791[FLOAT, 128x128x3x3]
%onnx::Conv_794[FLOAT, 128x128x1x1]
%onnx::Conv_797[FLOAT, 128x128x1x1]
%onnx::Conv_800[FLOAT, 128x128x1x1]
%onnx::Conv_803[FLOAT, 128x128x3x3]
%onnx::Conv_806[FLOAT, 128x128x3x3]
%onnx::Conv_809[FLOAT, 256x128x1x1]
%onnx::Conv_810[FLOAT, 256]
%onnx::Conv_812[FLOAT, 256x256x1x1]
%onnx::Conv_815[FLOAT, 256x128x1x1]
%onnx::Conv_818[FLOAT, 256x256x3x3]
%onnx::Conv_821[FLOAT, 256x256x3x3]
%onnx::Conv_824[FLOAT, 256x256x1x1]
%onnx::Conv_827[FLOAT, 256x256x1x1]
%onnx::Conv_830[FLOAT, 256x256x1x1]
%onnx::Conv_833[FLOAT, 256x256x3x3]
%onnx::Conv_836[FLOAT, 256x256x3x3]
%onnx::Conv_839[FLOAT, 256x256x1x1]
%onnx::Conv_842[FLOAT, 256x256x1x1]
%onnx::Conv_845[FLOAT, 256x256x1x1]
%onnx::Conv_848[FLOAT, 256x256x3x3]
%onnx::Conv_851[FLOAT, 256x256x3x3]
%onnx::Conv_854[FLOAT, 512x256x1x1]
%onnx::Conv_855[FLOAT, 512]
%onnx::Conv_857[FLOAT, 512x512x1x1]
%onnx::Conv_860[FLOAT, 512x256x1x1]
%onnx::Conv_863[FLOAT, 512x512x3x3]
%onnx::Conv_866[FLOAT, 512x512x3x3]
%onnx::Conv_869[FLOAT, 512x512x1x1]
%onnx::Conv_872[FLOAT, 512x512x1x1]
%onnx::Conv_875[FLOAT, 512x512x1x1]
%onnx::Conv_878[FLOAT, 512x512x3x3]
%onnx::Conv_881[FLOAT, 512x512x3x3]
%onnx::Conv_884[FLOAT, 512x512x1x1]
%onnx::Conv_887[FLOAT, 512x512x1x1]
%onnx::Conv_890[FLOAT, 512x512x1x1]
%onnx::Conv_893[FLOAT, 512x512x3x3]
%onnx::Conv_896[FLOAT, 512x512x3x3]
) {
%onnx::Conv_897 = Identity(%onnx::Conv_855)
%onnx::Conv_894 = Identity(%onnx::Conv_855)
%onnx::Conv_891 = Identity(%onnx::Conv_855)
%onnx::Conv_888 = Identity(%onnx::Conv_855)
%onnx::Conv_885 = Identity(%onnx::Conv_855)
%onnx::Conv_882 = Identity(%onnx::Conv_855)
%onnx::Conv_879 = Identity(%onnx::Conv_855)
%onnx::Conv_876 = Identity(%onnx::Conv_855)
%onnx::Conv_873 = Identity(%onnx::Conv_855)
%onnx::Conv_870 = Identity(%onnx::Conv_855)
%onnx::Conv_867 = Identity(%onnx::Conv_855)
%onnx::Conv_864 = Identity(%onnx::Conv_855)
%onnx::Conv_861 = Identity(%onnx::Conv_855)
%onnx::Conv_858 = Identity(%onnx::Conv_855)
%onnx::Conv_852 = Identity(%onnx::Conv_810)
%onnx::Conv_849 = Identity(%onnx::Conv_810)
%onnx::Conv_846 = Identity(%onnx::Conv_810)
%onnx::Conv_843 = Identity(%onnx::Conv_810)
%onnx::Conv_840 = Identity(%onnx::Conv_810)
%onnx::Conv_837 = Identity(%onnx::Conv_810)
%onnx::Conv_834 = Identity(%onnx::Conv_810)
%onnx::Conv_831 = Identity(%onnx::Conv_810)
%onnx::Conv_828 = Identity(%onnx::Conv_810)
%onnx::Conv_825 = Identity(%onnx::Conv_810)
%onnx::Conv_822 = Identity(%onnx::Conv_810)
%onnx::Conv_819 = Identity(%onnx::Conv_810)
%onnx::Conv_816 = Identity(%onnx::Conv_810)
%onnx::Conv_813 = Identity(%onnx::Conv_810)
%onnx::Conv_807 = Identity(%onnx::Conv_762)
%onnx::Conv_804 = Identity(%onnx::Conv_762)
%onnx::Conv_801 = Identity(%onnx::Conv_762)
%onnx::Conv_798 = Identity(%onnx::Conv_762)
%onnx::Conv_795 = Identity(%onnx::Conv_762)
%onnx::Conv_792 = Identity(%onnx::Conv_762)
%onnx::Conv_789 = Identity(%onnx::Conv_762)
%onnx::Conv_786 = Identity(%onnx::Conv_762)
%onnx::Conv_783 = Identity(%onnx::Conv_762)
%onnx::Conv_780 = Identity(%onnx::Conv_762)
%onnx::Conv_777 = Identity(%onnx::Conv_762)
%onnx::Conv_774 = Identity(%onnx::Conv_762)
%onnx::Conv_771 = Identity(%onnx::Conv_762)
%onnx::Conv_768 = Identity(%onnx::Conv_762)
%onnx::Conv_765 = Identity(%onnx::Conv_762)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_761, %onnx::Conv_762)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_764, %onnx::Conv_765)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_767, %onnx::Conv_768)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_770, %onnx::Conv_771)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/Add_1_output_0, %/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_773, %onnx::Conv_774)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/Add_5_output_0 = Add(%/layers.1/Add_4_output_0, %/layers.1/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_5_output_0, %onnx::Conv_776, %onnx::Conv_777)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_779, %onnx::Conv_780)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_782, %onnx::Conv_783)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_785, %onnx::Conv_786)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/Add_1_output_0, %/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_788, %onnx::Conv_789)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/Add_5_output_0 = Add(%/layers.2/Add_4_output_0, %/layers.2/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_5_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_797, %onnx::Conv_798)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_800, %onnx::Conv_801)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/Add_1_output_0, %/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_803, %onnx::Conv_804)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/Add_5_output_0 = Add(%/layers.3/Add_4_output_0, %/layers.3/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_5_output_0, %onnx::Conv_806, %onnx::Conv_807)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_809, %onnx::Conv_810)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_812, %onnx::Conv_813)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_815, %onnx::Conv_816)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/Add_1_output_0, %/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_818, %onnx::Conv_819)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/Add_5_output_0 = Add(%/layers.5/Add_4_output_0, %/layers.5/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_5_output_0, %onnx::Conv_821, %onnx::Conv_822)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_824, %onnx::Conv_825)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_827, %onnx::Conv_828)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_830, %onnx::Conv_831)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/Add_1_output_0, %/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_833, %onnx::Conv_834)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/Add_5_output_0 = Add(%/layers.6/Add_4_output_0, %/layers.6/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_5_output_0, %onnx::Conv_836, %onnx::Conv_837)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_839, %onnx::Conv_840)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_842, %onnx::Conv_843)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_845, %onnx::Conv_846)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/Add_1_output_0, %/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_848, %onnx::Conv_849)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/Add_5_output_0 = Add(%/layers.7/Add_4_output_0, %/layers.7/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_5_output_0, %onnx::Conv_851, %onnx::Conv_852)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_854, %onnx::Conv_855)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_857, %onnx::Conv_858)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/Add_1_output_0, %/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/Add_5_output_0 = Add(%/layers.9/Add_4_output_0, %/layers.9/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_5_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/Add_1_output_0, %/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/Add_5_output_0 = Add(%/layers.10/Add_4_output_0, %/layers.10/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_5_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/Add_1_output_0, %/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/Add_5_output_0 = Add(%/layers.11/Add_4_output_0, %/layers.11/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_5_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%759 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %759
}
|
val_accuracy
| 91.907054
| 6,310,340,608
| 21,384,074
|
{'zcp_epe_nas': 84.89820685474571, 'zcp_fisher': 30.060224533081055, 'zcp_flops': 100965449728.0, 'zcp_grad_norm': 85.62329864501953, 'zcp_grasp': -4.55047607421875, 'zcp_jacov': -16.058831160860507, 'zcp_l2_norm': 1030.2796630859375, 'zcp_nwot': 231.09205981972, 'zcp_params': 21384074.0, 'zcp_plain': 0.033845916390419, 'zcp_snip': 744.8335571289062, 'zcp_synflow': 130.2745924511092, 'zcp_zen': 102.50752258300781, 'zcp_val_accuracy': 0.9201722741127011}
| |
NASBench101_405310
|
NASBench101
|
405310
|
f50507bd5ff7c067649c9baf1b3054c5
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_851[FLOAT, 128x3x3x3]
%onnx::Conv_852[FLOAT, 128]
%onnx::Conv_854[FLOAT, 64x128x1x1]
%onnx::Conv_855[FLOAT, 64]
%onnx::Conv_857[FLOAT, 64x64x3x3]
%onnx::Conv_860[FLOAT, 64x64x1x1]
%onnx::Conv_863[FLOAT, 64x64x3x3]
%onnx::Conv_866[FLOAT, 64x128x1x1]
%onnx::Conv_869[FLOAT, 64x64x1x1]
%onnx::Conv_872[FLOAT, 64x128x1x1]
%onnx::Conv_875[FLOAT, 64x64x3x3]
%onnx::Conv_878[FLOAT, 64x64x1x1]
%onnx::Conv_881[FLOAT, 64x64x3x3]
%onnx::Conv_884[FLOAT, 64x128x1x1]
%onnx::Conv_887[FLOAT, 64x64x1x1]
%onnx::Conv_890[FLOAT, 64x128x1x1]
%onnx::Conv_893[FLOAT, 64x64x3x3]
%onnx::Conv_896[FLOAT, 64x64x1x1]
%onnx::Conv_899[FLOAT, 64x64x3x3]
%onnx::Conv_902[FLOAT, 64x128x1x1]
%onnx::Conv_905[FLOAT, 64x64x1x1]
%onnx::Conv_908[FLOAT, 128x128x1x1]
%onnx::Conv_911[FLOAT, 128x128x3x3]
%onnx::Conv_914[FLOAT, 128x128x1x1]
%onnx::Conv_917[FLOAT, 128x128x3x3]
%onnx::Conv_920[FLOAT, 128x128x1x1]
%onnx::Conv_923[FLOAT, 128x128x1x1]
%onnx::Conv_926[FLOAT, 128x256x1x1]
%onnx::Conv_929[FLOAT, 128x128x3x3]
%onnx::Conv_932[FLOAT, 128x128x1x1]
%onnx::Conv_935[FLOAT, 128x128x3x3]
%onnx::Conv_938[FLOAT, 128x256x1x1]
%onnx::Conv_941[FLOAT, 128x128x1x1]
%onnx::Conv_944[FLOAT, 128x256x1x1]
%onnx::Conv_947[FLOAT, 128x128x3x3]
%onnx::Conv_950[FLOAT, 128x128x1x1]
%onnx::Conv_953[FLOAT, 128x128x3x3]
%onnx::Conv_956[FLOAT, 128x256x1x1]
%onnx::Conv_959[FLOAT, 128x128x1x1]
%onnx::Conv_962[FLOAT, 256x256x1x1]
%onnx::Conv_963[FLOAT, 256]
%onnx::Conv_965[FLOAT, 256x256x3x3]
%onnx::Conv_968[FLOAT, 256x256x1x1]
%onnx::Conv_971[FLOAT, 256x256x3x3]
%onnx::Conv_974[FLOAT, 256x256x1x1]
%onnx::Conv_977[FLOAT, 256x256x1x1]
%onnx::Conv_980[FLOAT, 256x512x1x1]
%onnx::Conv_983[FLOAT, 256x256x3x3]
%onnx::Conv_986[FLOAT, 256x256x1x1]
%onnx::Conv_989[FLOAT, 256x256x3x3]
%onnx::Conv_992[FLOAT, 256x512x1x1]
%onnx::Conv_995[FLOAT, 256x256x1x1]
%onnx::Conv_998[FLOAT, 256x512x1x1]
%onnx::Conv_1001[FLOAT, 256x256x3x3]
%onnx::Conv_1004[FLOAT, 256x256x1x1]
%onnx::Conv_1007[FLOAT, 256x256x3x3]
%onnx::Conv_1010[FLOAT, 256x512x1x1]
%onnx::Conv_1013[FLOAT, 256x256x1x1]
) {
%onnx::Conv_1014 = Identity(%onnx::Conv_963)
%onnx::Conv_1011 = Identity(%onnx::Conv_963)
%onnx::Conv_1008 = Identity(%onnx::Conv_963)
%onnx::Conv_1005 = Identity(%onnx::Conv_963)
%onnx::Conv_1002 = Identity(%onnx::Conv_963)
%onnx::Conv_999 = Identity(%onnx::Conv_963)
%onnx::Conv_996 = Identity(%onnx::Conv_963)
%onnx::Conv_993 = Identity(%onnx::Conv_963)
%onnx::Conv_990 = Identity(%onnx::Conv_963)
%onnx::Conv_987 = Identity(%onnx::Conv_963)
%onnx::Conv_984 = Identity(%onnx::Conv_963)
%onnx::Conv_981 = Identity(%onnx::Conv_963)
%onnx::Conv_978 = Identity(%onnx::Conv_963)
%onnx::Conv_975 = Identity(%onnx::Conv_963)
%onnx::Conv_972 = Identity(%onnx::Conv_963)
%onnx::Conv_969 = Identity(%onnx::Conv_963)
%onnx::Conv_966 = Identity(%onnx::Conv_963)
%onnx::Conv_960 = Identity(%onnx::Conv_852)
%onnx::Conv_957 = Identity(%onnx::Conv_852)
%onnx::Conv_954 = Identity(%onnx::Conv_852)
%onnx::Conv_951 = Identity(%onnx::Conv_852)
%onnx::Conv_948 = Identity(%onnx::Conv_852)
%onnx::Conv_945 = Identity(%onnx::Conv_852)
%onnx::Conv_942 = Identity(%onnx::Conv_852)
%onnx::Conv_939 = Identity(%onnx::Conv_852)
%onnx::Conv_936 = Identity(%onnx::Conv_852)
%onnx::Conv_933 = Identity(%onnx::Conv_852)
%onnx::Conv_930 = Identity(%onnx::Conv_852)
%onnx::Conv_927 = Identity(%onnx::Conv_852)
%onnx::Conv_924 = Identity(%onnx::Conv_852)
%onnx::Conv_921 = Identity(%onnx::Conv_852)
%onnx::Conv_918 = Identity(%onnx::Conv_852)
%onnx::Conv_915 = Identity(%onnx::Conv_852)
%onnx::Conv_912 = Identity(%onnx::Conv_852)
%onnx::Conv_909 = Identity(%onnx::Conv_852)
%onnx::Conv_906 = Identity(%onnx::Conv_855)
%onnx::Conv_903 = Identity(%onnx::Conv_855)
%onnx::Conv_900 = Identity(%onnx::Conv_855)
%onnx::Conv_897 = Identity(%onnx::Conv_855)
%onnx::Conv_894 = Identity(%onnx::Conv_855)
%onnx::Conv_891 = Identity(%onnx::Conv_855)
%onnx::Conv_888 = Identity(%onnx::Conv_855)
%onnx::Conv_885 = Identity(%onnx::Conv_855)
%onnx::Conv_882 = Identity(%onnx::Conv_855)
%onnx::Conv_879 = Identity(%onnx::Conv_855)
%onnx::Conv_876 = Identity(%onnx::Conv_855)
%onnx::Conv_873 = Identity(%onnx::Conv_855)
%onnx::Conv_870 = Identity(%onnx::Conv_855)
%onnx::Conv_867 = Identity(%onnx::Conv_855)
%onnx::Conv_864 = Identity(%onnx::Conv_855)
%onnx::Conv_861 = Identity(%onnx::Conv_855)
%onnx::Conv_858 = Identity(%onnx::Conv_855)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_851, %onnx::Conv_852)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_854, %onnx::Conv_855)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_857, %onnx::Conv_858)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/Add_1_output_0, %/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/Add_3_output_0, %/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/Add_1_output_0, %/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/Add_3_output_0, %/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/Add_1_output_0, %/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/Add_3_output_0, %/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_4_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_908, %onnx::Conv_909)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_911, %onnx::Conv_912)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_914, %onnx::Conv_915)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/Add_1_output_0, %/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_917, %onnx::Conv_918)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_920, %onnx::Conv_921)
%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/Add_3_output_0, %/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_923, %onnx::Conv_924)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_926, %onnx::Conv_927)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_929, %onnx::Conv_930)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_932, %onnx::Conv_933)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/Add_1_output_0, %/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_935, %onnx::Conv_936)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_938, %onnx::Conv_939)
%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/Add_3_output_0, %/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_941, %onnx::Conv_942)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_944, %onnx::Conv_945)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_947, %onnx::Conv_948)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_950, %onnx::Conv_951)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/Add_1_output_0, %/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_953, %onnx::Conv_954)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_956, %onnx::Conv_957)
%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/Add_3_output_0, %/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_4_output_0, %onnx::Conv_959, %onnx::Conv_960)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_962, %onnx::Conv_963)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_965, %onnx::Conv_966)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_968, %onnx::Conv_969)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/Add_1_output_0, %/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_974, %onnx::Conv_975)
%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/Add_3_output_0, %/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_977, %onnx::Conv_978)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/Add_1_output_0, %/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/Add_3_output_0, %/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/Add_1_output_0, %/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/Add_3_output_0, %/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_4_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%849 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %849
}
|
val_accuracy
| 92.297679
| 1,803,036,672
| 6,054,282
|
{'zcp_epe_nas': 62.05171633067493, 'zcp_fisher': 51.32865524291992, 'zcp_flops': 28848586752.0, 'zcp_grad_norm': 132.4163360595703, 'zcp_grasp': 10.8095703125, 'zcp_jacov': -16.05144442649545, 'zcp_l2_norm': 995.2034301757812, 'zcp_nwot': 224.21177133536375, 'zcp_params': 6054282.0, 'zcp_plain': 0.029885256662964002, 'zcp_snip': 788.8931274414062, 'zcp_synflow': 120.34546559225478, 'zcp_zen': 101.15592956542969, 'zcp_val_accuracy': 0.9266827106475831}
| |
NASBench101_399691
|
NASBench101
|
399691
|
f1a1d8b7815652979fb3a96f47618f36
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_935[FLOAT, 128x3x3x3]
%onnx::Conv_936[FLOAT, 128]
%onnx::Conv_938[FLOAT, 43x128x1x1]
%onnx::Conv_939[FLOAT, 43]
%onnx::Conv_941[FLOAT, 43x43x1x1]
%onnx::Conv_944[FLOAT, 43x128x1x1]
%onnx::Conv_947[FLOAT, 42x42x3x3]
%onnx::Conv_948[FLOAT, 42]
%onnx::Conv_950[FLOAT, 42x42x3x3]
%onnx::Conv_953[FLOAT, 42x42x1x1]
%onnx::Conv_956[FLOAT, 43x128x1x1]
%onnx::Conv_959[FLOAT, 43x43x1x1]
%onnx::Conv_962[FLOAT, 43x128x1x1]
%onnx::Conv_965[FLOAT, 42x42x3x3]
%onnx::Conv_968[FLOAT, 42x42x3x3]
%onnx::Conv_971[FLOAT, 42x42x1x1]
%onnx::Conv_974[FLOAT, 43x128x1x1]
%onnx::Conv_977[FLOAT, 43x43x1x1]
%onnx::Conv_980[FLOAT, 43x128x1x1]
%onnx::Conv_983[FLOAT, 42x42x3x3]
%onnx::Conv_986[FLOAT, 42x42x3x3]
%onnx::Conv_989[FLOAT, 42x42x1x1]
%onnx::Conv_992[FLOAT, 86x128x1x1]
%onnx::Conv_993[FLOAT, 86]
%onnx::Conv_995[FLOAT, 86x86x1x1]
%onnx::Conv_998[FLOAT, 85x128x1x1]
%onnx::Conv_999[FLOAT, 85]
%onnx::Conv_1001[FLOAT, 85x85x3x3]
%onnx::Conv_1004[FLOAT, 85x85x3x3]
%onnx::Conv_1007[FLOAT, 85x85x1x1]
%onnx::Conv_1010[FLOAT, 86x256x1x1]
%onnx::Conv_1013[FLOAT, 86x86x1x1]
%onnx::Conv_1016[FLOAT, 85x256x1x1]
%onnx::Conv_1019[FLOAT, 85x85x3x3]
%onnx::Conv_1022[FLOAT, 85x85x3x3]
%onnx::Conv_1025[FLOAT, 85x85x1x1]
%onnx::Conv_1028[FLOAT, 86x256x1x1]
%onnx::Conv_1031[FLOAT, 86x86x1x1]
%onnx::Conv_1034[FLOAT, 85x256x1x1]
%onnx::Conv_1037[FLOAT, 85x85x3x3]
%onnx::Conv_1040[FLOAT, 85x85x3x3]
%onnx::Conv_1043[FLOAT, 85x85x1x1]
%onnx::Conv_1046[FLOAT, 171x256x1x1]
%onnx::Conv_1047[FLOAT, 171]
%onnx::Conv_1049[FLOAT, 171x171x1x1]
%onnx::Conv_1052[FLOAT, 171x256x1x1]
%onnx::Conv_1055[FLOAT, 170x170x3x3]
%onnx::Conv_1056[FLOAT, 170]
%onnx::Conv_1058[FLOAT, 170x170x3x3]
%onnx::Conv_1061[FLOAT, 170x170x1x1]
%onnx::Conv_1064[FLOAT, 171x512x1x1]
%onnx::Conv_1067[FLOAT, 171x171x1x1]
%onnx::Conv_1070[FLOAT, 171x512x1x1]
%onnx::Conv_1073[FLOAT, 170x170x3x3]
%onnx::Conv_1076[FLOAT, 170x170x3x3]
%onnx::Conv_1079[FLOAT, 170x170x1x1]
%onnx::Conv_1082[FLOAT, 171x512x1x1]
%onnx::Conv_1085[FLOAT, 171x171x1x1]
%onnx::Conv_1088[FLOAT, 171x512x1x1]
%onnx::Conv_1091[FLOAT, 170x170x3x3]
%onnx::Conv_1094[FLOAT, 170x170x3x3]
%onnx::Conv_1097[FLOAT, 170x170x1x1]
) {
%onnx::Conv_1098 = Identity(%onnx::Conv_1056)
%onnx::Conv_1095 = Identity(%onnx::Conv_1056)
%onnx::Conv_1092 = Identity(%onnx::Conv_1056)
%onnx::Conv_1089 = Identity(%onnx::Conv_1047)
%onnx::Conv_1086 = Identity(%onnx::Conv_1047)
%onnx::Conv_1083 = Identity(%onnx::Conv_1047)
%onnx::Conv_1080 = Identity(%onnx::Conv_1056)
%onnx::Conv_1077 = Identity(%onnx::Conv_1056)
%onnx::Conv_1074 = Identity(%onnx::Conv_1056)
%onnx::Conv_1071 = Identity(%onnx::Conv_1047)
%onnx::Conv_1068 = Identity(%onnx::Conv_1047)
%onnx::Conv_1065 = Identity(%onnx::Conv_1047)
%onnx::Conv_1062 = Identity(%onnx::Conv_1056)
%onnx::Conv_1059 = Identity(%onnx::Conv_1056)
%onnx::Conv_1053 = Identity(%onnx::Conv_1047)
%onnx::Conv_1050 = Identity(%onnx::Conv_1047)
%onnx::Conv_1044 = Identity(%onnx::Conv_999)
%onnx::Conv_1041 = Identity(%onnx::Conv_999)
%onnx::Conv_1038 = Identity(%onnx::Conv_999)
%onnx::Conv_1035 = Identity(%onnx::Conv_999)
%onnx::Conv_1032 = Identity(%onnx::Conv_993)
%onnx::Conv_1029 = Identity(%onnx::Conv_993)
%onnx::Conv_1026 = Identity(%onnx::Conv_999)
%onnx::Conv_1023 = Identity(%onnx::Conv_999)
%onnx::Conv_1020 = Identity(%onnx::Conv_999)
%onnx::Conv_1017 = Identity(%onnx::Conv_999)
%onnx::Conv_1014 = Identity(%onnx::Conv_993)
%onnx::Conv_1011 = Identity(%onnx::Conv_993)
%onnx::Conv_1008 = Identity(%onnx::Conv_999)
%onnx::Conv_1005 = Identity(%onnx::Conv_999)
%onnx::Conv_1002 = Identity(%onnx::Conv_999)
%onnx::Conv_996 = Identity(%onnx::Conv_993)
%onnx::Conv_990 = Identity(%onnx::Conv_948)
%onnx::Conv_987 = Identity(%onnx::Conv_948)
%onnx::Conv_984 = Identity(%onnx::Conv_948)
%onnx::Conv_981 = Identity(%onnx::Conv_939)
%onnx::Conv_978 = Identity(%onnx::Conv_939)
%onnx::Conv_975 = Identity(%onnx::Conv_939)
%onnx::Conv_972 = Identity(%onnx::Conv_948)
%onnx::Conv_969 = Identity(%onnx::Conv_948)
%onnx::Conv_966 = Identity(%onnx::Conv_948)
%onnx::Conv_963 = Identity(%onnx::Conv_939)
%onnx::Conv_960 = Identity(%onnx::Conv_939)
%onnx::Conv_957 = Identity(%onnx::Conv_939)
%onnx::Conv_954 = Identity(%onnx::Conv_948)
%onnx::Conv_951 = Identity(%onnx::Conv_948)
%onnx::Conv_945 = Identity(%onnx::Conv_939)
%onnx::Conv_942 = Identity(%onnx::Conv_939)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_935, %onnx::Conv_936)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_938, %onnx::Conv_939)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_941, %onnx::Conv_942)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_944, %onnx::Conv_945)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.1/Slice_output_0 = Slice(%/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %/layers.1/Constant_3_output_0, %/layers.1/Constant_4_output_0, %/layers.1/Constant_2_output_0, %/layers.1/Constant_5_output_0)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Slice_output_0, %onnx::Conv_947, %onnx::Conv_948)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.1/Slice_1_output_0 = Slice(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_7_output_0, %/layers.1/Constant_8_output_0, %/layers.1/Constant_6_output_0, %/layers.1/Constant_9_output_0)
%/layers.1/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/Slice_1_output_0, %/layers.1/Constant_10_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/Add_2_output_0, %/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_950, %onnx::Conv_951)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_11_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_11_output_0)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_953, %onnx::Conv_954)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_956, %onnx::Conv_957)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_959, %onnx::Conv_960)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_962, %onnx::Conv_963)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.2/Slice_output_0 = Slice(%/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %/layers.2/Constant_3_output_0, %/layers.2/Constant_4_output_0, %/layers.2/Constant_2_output_0, %/layers.2/Constant_5_output_0)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Slice_output_0, %onnx::Conv_965, %onnx::Conv_966)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.2/Slice_1_output_0 = Slice(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_7_output_0, %/layers.2/Constant_8_output_0, %/layers.2/Constant_6_output_0, %/layers.2/Constant_9_output_0)
%/layers.2/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/Slice_1_output_0, %/layers.2/Constant_10_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/Add_2_output_0, %/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_968, %onnx::Conv_969)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_11_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_11_output_0)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_974, %onnx::Conv_975)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_977, %onnx::Conv_978)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.3/Slice_output_0 = Slice(%/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %/layers.3/Constant_3_output_0, %/layers.3/Constant_4_output_0, %/layers.3/Constant_2_output_0, %/layers.3/Constant_5_output_0)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Slice_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.3/Slice_1_output_0 = Slice(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_7_output_0, %/layers.3/Constant_8_output_0, %/layers.3/Constant_6_output_0, %/layers.3/Constant_9_output_0)
%/layers.3/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/Slice_1_output_0, %/layers.3/Constant_10_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/Add_2_output_0, %/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_11_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_11_output_0)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_4_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.5/Slice_output_0 = Slice(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0, %/layers.5/Constant_4_output_0, %/layers.5/Constant_2_output_0, %/layers.5/Constant_5_output_0)
%/layers.5/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/Slice_output_0, %/layers.5/Constant_6_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/Add_2_output_0, %/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_7_output_0)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.6/Slice_output_0 = Slice(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0, %/layers.6/Constant_4_output_0, %/layers.6/Constant_2_output_0, %/layers.6/Constant_5_output_0)
%/layers.6/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/Slice_output_0, %/layers.6/Constant_6_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/Add_2_output_0, %/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_7_output_0)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_1025, %onnx::Conv_1026)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_1028, %onnx::Conv_1029)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_1031, %onnx::Conv_1032)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_1034, %onnx::Conv_1035)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_1037, %onnx::Conv_1038)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.7/Slice_output_0 = Slice(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0, %/layers.7/Constant_4_output_0, %/layers.7/Constant_2_output_0, %/layers.7/Constant_5_output_0)
%/layers.7/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/Slice_output_0, %/layers.7/Constant_6_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/Add_2_output_0, %/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_1040, %onnx::Conv_1041)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_7_output_0)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_4_output_0, %onnx::Conv_1043, %onnx::Conv_1044)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1046, %onnx::Conv_1047)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_1049, %onnx::Conv_1050)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1052, %onnx::Conv_1053)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.9/Slice_output_0 = Slice(%/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %/layers.9/Constant_3_output_0, %/layers.9/Constant_4_output_0, %/layers.9/Constant_2_output_0, %/layers.9/Constant_5_output_0)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Slice_output_0, %onnx::Conv_1055, %onnx::Conv_1056)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.9/Slice_1_output_0 = Slice(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_7_output_0, %/layers.9/Constant_8_output_0, %/layers.9/Constant_6_output_0, %/layers.9/Constant_9_output_0)
%/layers.9/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/Slice_1_output_0, %/layers.9/Constant_10_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/Add_2_output_0, %/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_1058, %onnx::Conv_1059)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_11_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_11_output_0)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_1061, %onnx::Conv_1062)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_1064, %onnx::Conv_1065)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_1067, %onnx::Conv_1068)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_1070, %onnx::Conv_1071)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.10/Slice_output_0 = Slice(%/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %/layers.10/Constant_3_output_0, %/layers.10/Constant_4_output_0, %/layers.10/Constant_2_output_0, %/layers.10/Constant_5_output_0)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Slice_output_0, %onnx::Conv_1073, %onnx::Conv_1074)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.10/Slice_1_output_0 = Slice(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_7_output_0, %/layers.10/Constant_8_output_0, %/layers.10/Constant_6_output_0, %/layers.10/Constant_9_output_0)
%/layers.10/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/Slice_1_output_0, %/layers.10/Constant_10_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/Add_2_output_0, %/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_1076, %onnx::Conv_1077)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_11_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_11_output_0)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_1079, %onnx::Conv_1080)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1082, %onnx::Conv_1083)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_1085, %onnx::Conv_1086)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1088, %onnx::Conv_1089)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.11/Slice_output_0 = Slice(%/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %/layers.11/Constant_3_output_0, %/layers.11/Constant_4_output_0, %/layers.11/Constant_2_output_0, %/layers.11/Constant_5_output_0)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Slice_output_0, %onnx::Conv_1091, %onnx::Conv_1092)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.11/Slice_1_output_0 = Slice(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_7_output_0, %/layers.11/Constant_8_output_0, %/layers.11/Constant_6_output_0, %/layers.11/Constant_9_output_0)
%/layers.11/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/Slice_1_output_0, %/layers.11/Constant_10_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/Add_2_output_0, %/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_1094, %onnx::Conv_1095)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_11_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_11_output_0)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_4_output_0, %onnx::Conv_1097, %onnx::Conv_1098)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%933 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %933
}
|
val_accuracy
| 92.127407
| 860,400,512
| 2,874,937
|
{'zcp_epe_nas': 72.1476902293358, 'zcp_fisher': 7.183144569396973, 'zcp_flops': 13766408192.0, 'zcp_grad_norm': 53.60993194580078, 'zcp_grasp': -4.331939697265625, 'zcp_jacov': -16.063052196108433, 'zcp_l2_norm': 884.2396240234375, 'zcp_nwot': 218.274277599722, 'zcp_params': 2874937.0, 'zcp_plain': 0.029054882004857, 'zcp_snip': 281.9461669921875, 'zcp_synflow': 112.12642889318894, 'zcp_zen': 90.44986724853516, 'zcp_val_accuracy': 0.9171674847602841}
| |
NASBench101_64800
|
NASBench101
|
64800
|
2755ed3de57a124986e1eb039c463b68
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_743[FLOAT, 128x3x3x3]
%onnx::Conv_744[FLOAT, 128]
%onnx::Conv_746[FLOAT, 32x128x1x1]
%onnx::Conv_747[FLOAT, 32]
%onnx::Conv_749[FLOAT, 32x32x1x1]
%onnx::Conv_752[FLOAT, 32x32x1x1]
%onnx::Conv_755[FLOAT, 32x32x1x1]
%onnx::Conv_758[FLOAT, 32x128x1x1]
%onnx::Conv_761[FLOAT, 32x128x1x1]
%onnx::Conv_764[FLOAT, 32x32x1x1]
%onnx::Conv_767[FLOAT, 32x32x1x1]
%onnx::Conv_770[FLOAT, 32x32x1x1]
%onnx::Conv_773[FLOAT, 32x128x1x1]
%onnx::Conv_776[FLOAT, 32x128x1x1]
%onnx::Conv_779[FLOAT, 32x32x1x1]
%onnx::Conv_782[FLOAT, 32x32x1x1]
%onnx::Conv_785[FLOAT, 32x32x1x1]
%onnx::Conv_788[FLOAT, 32x128x1x1]
%onnx::Conv_791[FLOAT, 64x128x1x1]
%onnx::Conv_792[FLOAT, 64]
%onnx::Conv_794[FLOAT, 64x64x1x1]
%onnx::Conv_797[FLOAT, 64x64x1x1]
%onnx::Conv_800[FLOAT, 64x64x1x1]
%onnx::Conv_803[FLOAT, 64x128x1x1]
%onnx::Conv_806[FLOAT, 64x256x1x1]
%onnx::Conv_809[FLOAT, 64x64x1x1]
%onnx::Conv_812[FLOAT, 64x64x1x1]
%onnx::Conv_815[FLOAT, 64x64x1x1]
%onnx::Conv_818[FLOAT, 64x256x1x1]
%onnx::Conv_821[FLOAT, 64x256x1x1]
%onnx::Conv_824[FLOAT, 64x64x1x1]
%onnx::Conv_827[FLOAT, 64x64x1x1]
%onnx::Conv_830[FLOAT, 64x64x1x1]
%onnx::Conv_833[FLOAT, 64x256x1x1]
%onnx::Conv_836[FLOAT, 128x256x1x1]
%onnx::Conv_839[FLOAT, 128x128x1x1]
%onnx::Conv_842[FLOAT, 128x128x1x1]
%onnx::Conv_845[FLOAT, 128x128x1x1]
%onnx::Conv_848[FLOAT, 128x256x1x1]
%onnx::Conv_851[FLOAT, 128x512x1x1]
%onnx::Conv_854[FLOAT, 128x128x1x1]
%onnx::Conv_857[FLOAT, 128x128x1x1]
%onnx::Conv_860[FLOAT, 128x128x1x1]
%onnx::Conv_863[FLOAT, 128x512x1x1]
%onnx::Conv_866[FLOAT, 128x512x1x1]
%onnx::Conv_869[FLOAT, 128x128x1x1]
%onnx::Conv_872[FLOAT, 128x128x1x1]
%onnx::Conv_875[FLOAT, 128x128x1x1]
%onnx::Conv_878[FLOAT, 128x512x1x1]
) {
%onnx::Conv_879 = Identity(%onnx::Conv_744)
%onnx::Conv_876 = Identity(%onnx::Conv_744)
%onnx::Conv_873 = Identity(%onnx::Conv_744)
%onnx::Conv_870 = Identity(%onnx::Conv_744)
%onnx::Conv_867 = Identity(%onnx::Conv_744)
%onnx::Conv_864 = Identity(%onnx::Conv_744)
%onnx::Conv_861 = Identity(%onnx::Conv_744)
%onnx::Conv_858 = Identity(%onnx::Conv_744)
%onnx::Conv_855 = Identity(%onnx::Conv_744)
%onnx::Conv_852 = Identity(%onnx::Conv_744)
%onnx::Conv_849 = Identity(%onnx::Conv_744)
%onnx::Conv_846 = Identity(%onnx::Conv_744)
%onnx::Conv_843 = Identity(%onnx::Conv_744)
%onnx::Conv_840 = Identity(%onnx::Conv_744)
%onnx::Conv_837 = Identity(%onnx::Conv_744)
%onnx::Conv_834 = Identity(%onnx::Conv_792)
%onnx::Conv_831 = Identity(%onnx::Conv_792)
%onnx::Conv_828 = Identity(%onnx::Conv_792)
%onnx::Conv_825 = Identity(%onnx::Conv_792)
%onnx::Conv_822 = Identity(%onnx::Conv_792)
%onnx::Conv_819 = Identity(%onnx::Conv_792)
%onnx::Conv_816 = Identity(%onnx::Conv_792)
%onnx::Conv_813 = Identity(%onnx::Conv_792)
%onnx::Conv_810 = Identity(%onnx::Conv_792)
%onnx::Conv_807 = Identity(%onnx::Conv_792)
%onnx::Conv_804 = Identity(%onnx::Conv_792)
%onnx::Conv_801 = Identity(%onnx::Conv_792)
%onnx::Conv_798 = Identity(%onnx::Conv_792)
%onnx::Conv_795 = Identity(%onnx::Conv_792)
%onnx::Conv_789 = Identity(%onnx::Conv_747)
%onnx::Conv_786 = Identity(%onnx::Conv_747)
%onnx::Conv_783 = Identity(%onnx::Conv_747)
%onnx::Conv_780 = Identity(%onnx::Conv_747)
%onnx::Conv_777 = Identity(%onnx::Conv_747)
%onnx::Conv_774 = Identity(%onnx::Conv_747)
%onnx::Conv_771 = Identity(%onnx::Conv_747)
%onnx::Conv_768 = Identity(%onnx::Conv_747)
%onnx::Conv_765 = Identity(%onnx::Conv_747)
%onnx::Conv_762 = Identity(%onnx::Conv_747)
%onnx::Conv_759 = Identity(%onnx::Conv_747)
%onnx::Conv_756 = Identity(%onnx::Conv_747)
%onnx::Conv_753 = Identity(%onnx::Conv_747)
%onnx::Conv_750 = Identity(%onnx::Conv_747)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_743, %onnx::Conv_744)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_746, %onnx::Conv_747)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_749, %onnx::Conv_750)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_752, %onnx::Conv_753)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_755, %onnx::Conv_756)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_758, %onnx::Conv_759)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/Add_3_output_0, %/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_4_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_761, %onnx::Conv_762)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_764, %onnx::Conv_765)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_767, %onnx::Conv_768)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_770, %onnx::Conv_771)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_773, %onnx::Conv_774)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/Add_3_output_0, %/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_4_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_776, %onnx::Conv_777)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_779, %onnx::Conv_780)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_782, %onnx::Conv_783)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_785, %onnx::Conv_786)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_788, %onnx::Conv_789)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/Add_3_output_0, %/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_4_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_797, %onnx::Conv_798)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_800, %onnx::Conv_801)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_803, %onnx::Conv_804)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/Add_3_output_0, %/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_4_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_806, %onnx::Conv_807)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_809, %onnx::Conv_810)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_812, %onnx::Conv_813)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_815, %onnx::Conv_816)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_818, %onnx::Conv_819)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/Add_3_output_0, %/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_4_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_821, %onnx::Conv_822)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_824, %onnx::Conv_825)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_827, %onnx::Conv_828)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_830, %onnx::Conv_831)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_833, %onnx::Conv_834)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/Add_3_output_0, %/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_4_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_836, %onnx::Conv_837)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_839, %onnx::Conv_840)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_842, %onnx::Conv_843)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_845, %onnx::Conv_846)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_848, %onnx::Conv_849)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/Add_3_output_0, %/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_4_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_851, %onnx::Conv_852)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_854, %onnx::Conv_855)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_857, %onnx::Conv_858)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/Add_3_output_0, %/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_4_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/Add_3_output_0, %/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_4_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.4/maxpool/MaxPool_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%741 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %741
}
|
val_accuracy
| 86.698717
| 205,858,816
| 643,274
|
{'zcp_epe_nas': 90.24487806040896, 'zcp_fisher': 58.2216911315918, 'zcp_flops': 3293741056.0, 'zcp_grad_norm': 153.65841674804688, 'zcp_grasp': -30.77587890625, 'zcp_jacov': -16.058806370650828, 'zcp_l2_norm': 712.2608642578125, 'zcp_nwot': 211.9761826900039, 'zcp_params': 643274.0, 'zcp_plain': 0.011880172416567001, 'zcp_snip': 569.2600708007812, 'zcp_synflow': 91.70181200492873, 'zcp_zen': 57.64912414550781, 'zcp_val_accuracy': 0.929286837577819}
| |
NASBench101_42404
|
NASBench101
|
42404
|
19b48d58db2b39783384258b85fa1e5e
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_1076[FLOAT, 128x3x3x3]
%onnx::Conv_1077[FLOAT, 128]
%onnx::Conv_1079[FLOAT, 128x128x1x1]
%onnx::Conv_1082[FLOAT, 128x128x3x3]
%onnx::Conv_1085[FLOAT, 128x128x1x1]
%onnx::Conv_1088[FLOAT, 128x128x1x1]
%onnx::Conv_1091[FLOAT, 128x128x1x1]
%onnx::Conv_1094[FLOAT, 128x128x1x1]
%onnx::Conv_1097[FLOAT, 128x128x1x1]
%onnx::Conv_1100[FLOAT, 128x128x1x1]
%onnx::Conv_1103[FLOAT, 128x128x1x1]
%onnx::Conv_1106[FLOAT, 128x128x3x3]
%onnx::Conv_1109[FLOAT, 128x128x1x1]
%onnx::Conv_1112[FLOAT, 128x128x1x1]
%onnx::Conv_1115[FLOAT, 128x128x1x1]
%onnx::Conv_1118[FLOAT, 128x128x1x1]
%onnx::Conv_1121[FLOAT, 128x128x1x1]
%onnx::Conv_1124[FLOAT, 128x128x1x1]
%onnx::Conv_1127[FLOAT, 128x128x1x1]
%onnx::Conv_1130[FLOAT, 128x128x3x3]
%onnx::Conv_1133[FLOAT, 128x128x1x1]
%onnx::Conv_1136[FLOAT, 128x128x1x1]
%onnx::Conv_1139[FLOAT, 128x128x1x1]
%onnx::Conv_1142[FLOAT, 128x128x1x1]
%onnx::Conv_1145[FLOAT, 128x128x1x1]
%onnx::Conv_1148[FLOAT, 128x128x1x1]
%onnx::Conv_1151[FLOAT, 256x128x1x1]
%onnx::Conv_1152[FLOAT, 256]
%onnx::Conv_1154[FLOAT, 256x256x3x3]
%onnx::Conv_1157[FLOAT, 256x128x1x1]
%onnx::Conv_1160[FLOAT, 256x256x1x1]
%onnx::Conv_1163[FLOAT, 256x256x1x1]
%onnx::Conv_1166[FLOAT, 256x128x1x1]
%onnx::Conv_1169[FLOAT, 256x256x1x1]
%onnx::Conv_1172[FLOAT, 256x256x1x1]
%onnx::Conv_1175[FLOAT, 256x256x1x1]
%onnx::Conv_1178[FLOAT, 256x256x3x3]
%onnx::Conv_1181[FLOAT, 256x256x1x1]
%onnx::Conv_1184[FLOAT, 256x256x1x1]
%onnx::Conv_1187[FLOAT, 256x256x1x1]
%onnx::Conv_1190[FLOAT, 256x256x1x1]
%onnx::Conv_1193[FLOAT, 256x256x1x1]
%onnx::Conv_1196[FLOAT, 256x256x1x1]
%onnx::Conv_1199[FLOAT, 256x256x1x1]
%onnx::Conv_1202[FLOAT, 256x256x3x3]
%onnx::Conv_1205[FLOAT, 256x256x1x1]
%onnx::Conv_1208[FLOAT, 256x256x1x1]
%onnx::Conv_1211[FLOAT, 256x256x1x1]
%onnx::Conv_1214[FLOAT, 256x256x1x1]
%onnx::Conv_1217[FLOAT, 256x256x1x1]
%onnx::Conv_1220[FLOAT, 256x256x1x1]
%onnx::Conv_1223[FLOAT, 512x256x1x1]
%onnx::Conv_1224[FLOAT, 512]
%onnx::Conv_1226[FLOAT, 512x512x3x3]
%onnx::Conv_1229[FLOAT, 512x256x1x1]
%onnx::Conv_1232[FLOAT, 512x512x1x1]
%onnx::Conv_1235[FLOAT, 512x512x1x1]
%onnx::Conv_1238[FLOAT, 512x256x1x1]
%onnx::Conv_1241[FLOAT, 512x512x1x1]
%onnx::Conv_1244[FLOAT, 512x512x1x1]
%onnx::Conv_1247[FLOAT, 512x512x1x1]
%onnx::Conv_1250[FLOAT, 512x512x3x3]
%onnx::Conv_1253[FLOAT, 512x512x1x1]
%onnx::Conv_1256[FLOAT, 512x512x1x1]
%onnx::Conv_1259[FLOAT, 512x512x1x1]
%onnx::Conv_1262[FLOAT, 512x512x1x1]
%onnx::Conv_1265[FLOAT, 512x512x1x1]
%onnx::Conv_1268[FLOAT, 512x512x1x1]
%onnx::Conv_1271[FLOAT, 512x512x1x1]
%onnx::Conv_1274[FLOAT, 512x512x3x3]
%onnx::Conv_1277[FLOAT, 512x512x1x1]
%onnx::Conv_1280[FLOAT, 512x512x1x1]
%onnx::Conv_1283[FLOAT, 512x512x1x1]
%onnx::Conv_1286[FLOAT, 512x512x1x1]
%onnx::Conv_1289[FLOAT, 512x512x1x1]
%onnx::Conv_1292[FLOAT, 512x512x1x1]
) {
%onnx::Conv_1293 = Identity(%onnx::Conv_1224)
%onnx::Conv_1290 = Identity(%onnx::Conv_1224)
%onnx::Conv_1287 = Identity(%onnx::Conv_1224)
%onnx::Conv_1284 = Identity(%onnx::Conv_1224)
%onnx::Conv_1281 = Identity(%onnx::Conv_1224)
%onnx::Conv_1278 = Identity(%onnx::Conv_1224)
%onnx::Conv_1275 = Identity(%onnx::Conv_1224)
%onnx::Conv_1272 = Identity(%onnx::Conv_1224)
%onnx::Conv_1269 = Identity(%onnx::Conv_1224)
%onnx::Conv_1266 = Identity(%onnx::Conv_1224)
%onnx::Conv_1263 = Identity(%onnx::Conv_1224)
%onnx::Conv_1260 = Identity(%onnx::Conv_1224)
%onnx::Conv_1257 = Identity(%onnx::Conv_1224)
%onnx::Conv_1254 = Identity(%onnx::Conv_1224)
%onnx::Conv_1251 = Identity(%onnx::Conv_1224)
%onnx::Conv_1248 = Identity(%onnx::Conv_1224)
%onnx::Conv_1245 = Identity(%onnx::Conv_1224)
%onnx::Conv_1242 = Identity(%onnx::Conv_1224)
%onnx::Conv_1239 = Identity(%onnx::Conv_1224)
%onnx::Conv_1236 = Identity(%onnx::Conv_1224)
%onnx::Conv_1233 = Identity(%onnx::Conv_1224)
%onnx::Conv_1230 = Identity(%onnx::Conv_1224)
%onnx::Conv_1227 = Identity(%onnx::Conv_1224)
%onnx::Conv_1221 = Identity(%onnx::Conv_1152)
%onnx::Conv_1218 = Identity(%onnx::Conv_1152)
%onnx::Conv_1215 = Identity(%onnx::Conv_1152)
%onnx::Conv_1212 = Identity(%onnx::Conv_1152)
%onnx::Conv_1209 = Identity(%onnx::Conv_1152)
%onnx::Conv_1206 = Identity(%onnx::Conv_1152)
%onnx::Conv_1203 = Identity(%onnx::Conv_1152)
%onnx::Conv_1200 = Identity(%onnx::Conv_1152)
%onnx::Conv_1197 = Identity(%onnx::Conv_1152)
%onnx::Conv_1194 = Identity(%onnx::Conv_1152)
%onnx::Conv_1191 = Identity(%onnx::Conv_1152)
%onnx::Conv_1188 = Identity(%onnx::Conv_1152)
%onnx::Conv_1185 = Identity(%onnx::Conv_1152)
%onnx::Conv_1182 = Identity(%onnx::Conv_1152)
%onnx::Conv_1179 = Identity(%onnx::Conv_1152)
%onnx::Conv_1176 = Identity(%onnx::Conv_1152)
%onnx::Conv_1173 = Identity(%onnx::Conv_1152)
%onnx::Conv_1170 = Identity(%onnx::Conv_1152)
%onnx::Conv_1167 = Identity(%onnx::Conv_1152)
%onnx::Conv_1164 = Identity(%onnx::Conv_1152)
%onnx::Conv_1161 = Identity(%onnx::Conv_1152)
%onnx::Conv_1158 = Identity(%onnx::Conv_1152)
%onnx::Conv_1155 = Identity(%onnx::Conv_1152)
%onnx::Conv_1149 = Identity(%onnx::Conv_1077)
%onnx::Conv_1146 = Identity(%onnx::Conv_1077)
%onnx::Conv_1143 = Identity(%onnx::Conv_1077)
%onnx::Conv_1140 = Identity(%onnx::Conv_1077)
%onnx::Conv_1137 = Identity(%onnx::Conv_1077)
%onnx::Conv_1134 = Identity(%onnx::Conv_1077)
%onnx::Conv_1131 = Identity(%onnx::Conv_1077)
%onnx::Conv_1128 = Identity(%onnx::Conv_1077)
%onnx::Conv_1125 = Identity(%onnx::Conv_1077)
%onnx::Conv_1122 = Identity(%onnx::Conv_1077)
%onnx::Conv_1119 = Identity(%onnx::Conv_1077)
%onnx::Conv_1116 = Identity(%onnx::Conv_1077)
%onnx::Conv_1113 = Identity(%onnx::Conv_1077)
%onnx::Conv_1110 = Identity(%onnx::Conv_1077)
%onnx::Conv_1107 = Identity(%onnx::Conv_1077)
%onnx::Conv_1104 = Identity(%onnx::Conv_1077)
%onnx::Conv_1101 = Identity(%onnx::Conv_1077)
%onnx::Conv_1098 = Identity(%onnx::Conv_1077)
%onnx::Conv_1095 = Identity(%onnx::Conv_1077)
%onnx::Conv_1092 = Identity(%onnx::Conv_1077)
%onnx::Conv_1089 = Identity(%onnx::Conv_1077)
%onnx::Conv_1086 = Identity(%onnx::Conv_1077)
%onnx::Conv_1083 = Identity(%onnx::Conv_1077)
%onnx::Conv_1080 = Identity(%onnx::Conv_1077)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_1076, %onnx::Conv_1077)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1079, %onnx::Conv_1080)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_1082, %onnx::Conv_1083)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1085, %onnx::Conv_1086)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/Add_1_output_0, %/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_1088, %onnx::Conv_1089)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_1091, %onnx::Conv_1092)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1094, %onnx::Conv_1095)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/Add_5_output_0 = Add(%/layers.1/Add_4_output_0, %/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/Add_6_output_0 = Add(%/layers.1/Add_5_output_0, %/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_6_output_0, %onnx::Conv_1097, %onnx::Conv_1098)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_7_output_0 = Add(%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_4_output_0)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_7_output_0, %onnx::Conv_1100, %onnx::Conv_1101)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1103, %onnx::Conv_1104)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_1106, %onnx::Conv_1107)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1109, %onnx::Conv_1110)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/Add_1_output_0, %/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_1112, %onnx::Conv_1113)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_1115, %onnx::Conv_1116)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1118, %onnx::Conv_1119)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/Add_5_output_0 = Add(%/layers.2/Add_4_output_0, %/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/Add_6_output_0 = Add(%/layers.2/Add_5_output_0, %/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_6_output_0, %onnx::Conv_1121, %onnx::Conv_1122)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_7_output_0 = Add(%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_4_output_0)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_7_output_0, %onnx::Conv_1124, %onnx::Conv_1125)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1127, %onnx::Conv_1128)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_1130, %onnx::Conv_1131)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1133, %onnx::Conv_1134)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/Add_1_output_0, %/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_1136, %onnx::Conv_1137)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_1139, %onnx::Conv_1140)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1142, %onnx::Conv_1143)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/Add_5_output_0 = Add(%/layers.3/Add_4_output_0, %/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/Add_6_output_0 = Add(%/layers.3/Add_5_output_0, %/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_6_output_0, %onnx::Conv_1145, %onnx::Conv_1146)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_7_output_0 = Add(%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_4_output_0)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_7_output_0, %onnx::Conv_1148, %onnx::Conv_1149)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1151, %onnx::Conv_1152)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_1154, %onnx::Conv_1155)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1157, %onnx::Conv_1158)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/Add_1_output_0, %/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_1160, %onnx::Conv_1161)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_1163, %onnx::Conv_1164)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1166, %onnx::Conv_1167)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/Add_5_output_0 = Add(%/layers.5/Add_4_output_0, %/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/Add_6_output_0 = Add(%/layers.5/Add_5_output_0, %/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_6_output_0, %onnx::Conv_1169, %onnx::Conv_1170)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_7_output_0 = Add(%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_4_output_0)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_7_output_0, %onnx::Conv_1172, %onnx::Conv_1173)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1175, %onnx::Conv_1176)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_1178, %onnx::Conv_1179)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1181, %onnx::Conv_1182)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/Add_1_output_0, %/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_1184, %onnx::Conv_1185)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_1187, %onnx::Conv_1188)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1190, %onnx::Conv_1191)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/Add_5_output_0 = Add(%/layers.6/Add_4_output_0, %/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/Add_6_output_0 = Add(%/layers.6/Add_5_output_0, %/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_6_output_0, %onnx::Conv_1193, %onnx::Conv_1194)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_7_output_0 = Add(%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_4_output_0)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_7_output_0, %onnx::Conv_1196, %onnx::Conv_1197)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1199, %onnx::Conv_1200)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_1202, %onnx::Conv_1203)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1205, %onnx::Conv_1206)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/Add_1_output_0, %/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_1208, %onnx::Conv_1209)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_1211, %onnx::Conv_1212)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1214, %onnx::Conv_1215)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/Add_5_output_0 = Add(%/layers.7/Add_4_output_0, %/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/Add_6_output_0 = Add(%/layers.7/Add_5_output_0, %/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_6_output_0, %onnx::Conv_1217, %onnx::Conv_1218)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_7_output_0 = Add(%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_4_output_0)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_7_output_0, %onnx::Conv_1220, %onnx::Conv_1221)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1223, %onnx::Conv_1224)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_1226, %onnx::Conv_1227)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1229, %onnx::Conv_1230)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/Add_1_output_0, %/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_1232, %onnx::Conv_1233)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_1235, %onnx::Conv_1236)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1238, %onnx::Conv_1239)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/Add_5_output_0 = Add(%/layers.9/Add_4_output_0, %/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/Add_6_output_0 = Add(%/layers.9/Add_5_output_0, %/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_6_output_0, %onnx::Conv_1241, %onnx::Conv_1242)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_7_output_0 = Add(%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_4_output_0)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_7_output_0, %onnx::Conv_1244, %onnx::Conv_1245)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1247, %onnx::Conv_1248)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_1250, %onnx::Conv_1251)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1253, %onnx::Conv_1254)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/Add_1_output_0, %/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_1256, %onnx::Conv_1257)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_1259, %onnx::Conv_1260)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1262, %onnx::Conv_1263)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/Add_5_output_0 = Add(%/layers.10/Add_4_output_0, %/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/Add_6_output_0 = Add(%/layers.10/Add_5_output_0, %/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_6_output_0, %onnx::Conv_1265, %onnx::Conv_1266)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_7_output_0 = Add(%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_4_output_0)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_7_output_0, %onnx::Conv_1268, %onnx::Conv_1269)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1271, %onnx::Conv_1272)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_1274, %onnx::Conv_1275)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1277, %onnx::Conv_1278)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/Add_1_output_0, %/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_1280, %onnx::Conv_1281)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_1283, %onnx::Conv_1284)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1286, %onnx::Conv_1287)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/Add_5_output_0 = Add(%/layers.11/Add_4_output_0, %/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/Add_6_output_0 = Add(%/layers.11/Add_5_output_0, %/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_6_output_0, %onnx::Conv_1289, %onnx::Conv_1290)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_7_output_0 = Add(%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_4_output_0)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_7_output_0, %onnx::Conv_1292, %onnx::Conv_1293)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%1074 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %1074
}
|
val_accuracy
| 89.032453
| 4,783,351,808
| 16,075,402
|
{'zcp_epe_nas': 77.56432359649384, 'zcp_fisher': 411.7222595214844, 'zcp_flops': 76533628928.0, 'zcp_grad_norm': 426.89178466796875, 'zcp_grasp': -1584.830078125, 'zcp_jacov': -16.058149026313483, 'zcp_l2_norm': 1650.728759765625, 'zcp_nwot': 239.76836340760622, 'zcp_params': 16075402.0, 'zcp_plain': -0.008737961761653, 'zcp_snip': 2991.678466796875, 'zcp_synflow': 172.5643254970728, 'zcp_zen': 128.75804138183594, 'zcp_val_accuracy': 0.9192708134651181}
| |
NASBench101_397691
|
NASBench101
|
397691
|
f065dba53855e9082c7ed237c997bb8b
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_761[FLOAT, 128x3x3x3]
%onnx::Conv_762[FLOAT, 128]
%onnx::Conv_764[FLOAT, 64x128x1x1]
%onnx::Conv_765[FLOAT, 64]
%onnx::Conv_767[FLOAT, 64x64x1x1]
%onnx::Conv_770[FLOAT, 64x64x1x1]
%onnx::Conv_773[FLOAT, 64x128x1x1]
%onnx::Conv_776[FLOAT, 64x64x1x1]
%onnx::Conv_779[FLOAT, 64x128x1x1]
%onnx::Conv_782[FLOAT, 64x64x1x1]
%onnx::Conv_785[FLOAT, 64x64x1x1]
%onnx::Conv_788[FLOAT, 64x128x1x1]
%onnx::Conv_791[FLOAT, 64x64x1x1]
%onnx::Conv_794[FLOAT, 64x128x1x1]
%onnx::Conv_797[FLOAT, 64x64x1x1]
%onnx::Conv_800[FLOAT, 64x64x1x1]
%onnx::Conv_803[FLOAT, 64x128x1x1]
%onnx::Conv_806[FLOAT, 64x64x1x1]
%onnx::Conv_809[FLOAT, 128x128x1x1]
%onnx::Conv_812[FLOAT, 128x128x1x1]
%onnx::Conv_815[FLOAT, 128x128x1x1]
%onnx::Conv_818[FLOAT, 128x128x1x1]
%onnx::Conv_821[FLOAT, 128x128x1x1]
%onnx::Conv_824[FLOAT, 128x256x1x1]
%onnx::Conv_827[FLOAT, 128x128x1x1]
%onnx::Conv_830[FLOAT, 128x128x1x1]
%onnx::Conv_833[FLOAT, 128x256x1x1]
%onnx::Conv_836[FLOAT, 128x128x1x1]
%onnx::Conv_839[FLOAT, 128x256x1x1]
%onnx::Conv_842[FLOAT, 128x128x1x1]
%onnx::Conv_845[FLOAT, 128x128x1x1]
%onnx::Conv_848[FLOAT, 128x256x1x1]
%onnx::Conv_851[FLOAT, 128x128x1x1]
%onnx::Conv_854[FLOAT, 256x256x1x1]
%onnx::Conv_855[FLOAT, 256]
%onnx::Conv_857[FLOAT, 256x256x1x1]
%onnx::Conv_860[FLOAT, 256x256x1x1]
%onnx::Conv_863[FLOAT, 256x256x1x1]
%onnx::Conv_866[FLOAT, 256x256x1x1]
%onnx::Conv_869[FLOAT, 256x512x1x1]
%onnx::Conv_872[FLOAT, 256x256x1x1]
%onnx::Conv_875[FLOAT, 256x256x1x1]
%onnx::Conv_878[FLOAT, 256x512x1x1]
%onnx::Conv_881[FLOAT, 256x256x1x1]
%onnx::Conv_884[FLOAT, 256x512x1x1]
%onnx::Conv_887[FLOAT, 256x256x1x1]
%onnx::Conv_890[FLOAT, 256x256x1x1]
%onnx::Conv_893[FLOAT, 256x512x1x1]
%onnx::Conv_896[FLOAT, 256x256x1x1]
) {
%onnx::Conv_897 = Identity(%onnx::Conv_855)
%onnx::Conv_894 = Identity(%onnx::Conv_855)
%onnx::Conv_891 = Identity(%onnx::Conv_855)
%onnx::Conv_888 = Identity(%onnx::Conv_855)
%onnx::Conv_885 = Identity(%onnx::Conv_855)
%onnx::Conv_882 = Identity(%onnx::Conv_855)
%onnx::Conv_879 = Identity(%onnx::Conv_855)
%onnx::Conv_876 = Identity(%onnx::Conv_855)
%onnx::Conv_873 = Identity(%onnx::Conv_855)
%onnx::Conv_870 = Identity(%onnx::Conv_855)
%onnx::Conv_867 = Identity(%onnx::Conv_855)
%onnx::Conv_864 = Identity(%onnx::Conv_855)
%onnx::Conv_861 = Identity(%onnx::Conv_855)
%onnx::Conv_858 = Identity(%onnx::Conv_855)
%onnx::Conv_852 = Identity(%onnx::Conv_762)
%onnx::Conv_849 = Identity(%onnx::Conv_762)
%onnx::Conv_846 = Identity(%onnx::Conv_762)
%onnx::Conv_843 = Identity(%onnx::Conv_762)
%onnx::Conv_840 = Identity(%onnx::Conv_762)
%onnx::Conv_837 = Identity(%onnx::Conv_762)
%onnx::Conv_834 = Identity(%onnx::Conv_762)
%onnx::Conv_831 = Identity(%onnx::Conv_762)
%onnx::Conv_828 = Identity(%onnx::Conv_762)
%onnx::Conv_825 = Identity(%onnx::Conv_762)
%onnx::Conv_822 = Identity(%onnx::Conv_762)
%onnx::Conv_819 = Identity(%onnx::Conv_762)
%onnx::Conv_816 = Identity(%onnx::Conv_762)
%onnx::Conv_813 = Identity(%onnx::Conv_762)
%onnx::Conv_810 = Identity(%onnx::Conv_762)
%onnx::Conv_807 = Identity(%onnx::Conv_765)
%onnx::Conv_804 = Identity(%onnx::Conv_765)
%onnx::Conv_801 = Identity(%onnx::Conv_765)
%onnx::Conv_798 = Identity(%onnx::Conv_765)
%onnx::Conv_795 = Identity(%onnx::Conv_765)
%onnx::Conv_792 = Identity(%onnx::Conv_765)
%onnx::Conv_789 = Identity(%onnx::Conv_765)
%onnx::Conv_786 = Identity(%onnx::Conv_765)
%onnx::Conv_783 = Identity(%onnx::Conv_765)
%onnx::Conv_780 = Identity(%onnx::Conv_765)
%onnx::Conv_777 = Identity(%onnx::Conv_765)
%onnx::Conv_774 = Identity(%onnx::Conv_765)
%onnx::Conv_771 = Identity(%onnx::Conv_765)
%onnx::Conv_768 = Identity(%onnx::Conv_765)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_761, %onnx::Conv_762)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_764, %onnx::Conv_765)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_767, %onnx::Conv_768)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_770, %onnx::Conv_771)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_773, %onnx::Conv_774)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/Add_1_output_0, %/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/Add_3_output_0, %/layers.1/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_776, %onnx::Conv_777)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_779, %onnx::Conv_780)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_782, %onnx::Conv_783)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_785, %onnx::Conv_786)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_788, %onnx::Conv_789)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/Add_1_output_0, %/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/Add_3_output_0, %/layers.2/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_797, %onnx::Conv_798)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_800, %onnx::Conv_801)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_803, %onnx::Conv_804)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/Add_1_output_0, %/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/Add_3_output_0, %/layers.3/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_4_output_0, %onnx::Conv_806, %onnx::Conv_807)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_809, %onnx::Conv_810)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_812, %onnx::Conv_813)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_815, %onnx::Conv_816)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_818, %onnx::Conv_819)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/Add_1_output_0, %/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/Add_3_output_0, %/layers.5/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_821, %onnx::Conv_822)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_824, %onnx::Conv_825)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_827, %onnx::Conv_828)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_830, %onnx::Conv_831)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_833, %onnx::Conv_834)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/Add_1_output_0, %/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/Add_3_output_0, %/layers.6/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_836, %onnx::Conv_837)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_839, %onnx::Conv_840)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_842, %onnx::Conv_843)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_845, %onnx::Conv_846)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_848, %onnx::Conv_849)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/Add_1_output_0, %/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/Add_3_output_0, %/layers.7/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_4_output_0, %onnx::Conv_851, %onnx::Conv_852)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_854, %onnx::Conv_855)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_857, %onnx::Conv_858)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/Add_1_output_0, %/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/Add_3_output_0, %/layers.9/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/Add_1_output_0, %/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/Add_3_output_0, %/layers.10/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/Add_1_output_0, %/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/Add_3_output_0, %/layers.11/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_4_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%759 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %759
}
|
val_accuracy
| 89.523238
| 516,827,136
| 1,664,778
|
{'zcp_epe_nas': 170.13971563404047, 'zcp_fisher': 3.080749034881592, 'zcp_flops': 8269234176.0, 'zcp_grad_norm': 35.053619384765625, 'zcp_grasp': -0.48027038574218706, 'zcp_jacov': -16.04746391351742, 'zcp_l2_norm': 844.2696533203125, 'zcp_nwot': 221.9604323563242, 'zcp_params': 1664778.0, 'zcp_plain': -0.070335566997528, 'zcp_snip': 201.33880615234375, 'zcp_synflow': 80.96864669254981, 'zcp_zen': 72.61036682128906, 'zcp_val_accuracy': 0.926382184028625}
| |
NASBench101_246438
|
NASBench101
|
246438
|
95314ded8677d5560e2a7020668ec172
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_770[FLOAT, 128x3x3x3]
%onnx::Conv_771[FLOAT, 128]
%onnx::Conv_773[FLOAT, 128x128x1x1]
%onnx::Conv_776[FLOAT, 128x128x3x3]
%onnx::Conv_779[FLOAT, 128x128x3x3]
%onnx::Conv_782[FLOAT, 128x128x1x1]
%onnx::Conv_785[FLOAT, 128x128x1x1]
%onnx::Conv_788[FLOAT, 128x128x1x1]
%onnx::Conv_791[FLOAT, 128x128x3x3]
%onnx::Conv_794[FLOAT, 128x128x3x3]
%onnx::Conv_797[FLOAT, 128x128x1x1]
%onnx::Conv_800[FLOAT, 128x128x1x1]
%onnx::Conv_803[FLOAT, 128x128x1x1]
%onnx::Conv_806[FLOAT, 128x128x3x3]
%onnx::Conv_809[FLOAT, 128x128x3x3]
%onnx::Conv_812[FLOAT, 128x128x1x1]
%onnx::Conv_815[FLOAT, 128x128x1x1]
%onnx::Conv_818[FLOAT, 256x128x1x1]
%onnx::Conv_819[FLOAT, 256]
%onnx::Conv_821[FLOAT, 256x256x3x3]
%onnx::Conv_824[FLOAT, 256x256x3x3]
%onnx::Conv_827[FLOAT, 256x128x1x1]
%onnx::Conv_830[FLOAT, 256x256x1x1]
%onnx::Conv_833[FLOAT, 256x256x1x1]
%onnx::Conv_836[FLOAT, 256x256x3x3]
%onnx::Conv_839[FLOAT, 256x256x3x3]
%onnx::Conv_842[FLOAT, 256x256x1x1]
%onnx::Conv_845[FLOAT, 256x256x1x1]
%onnx::Conv_848[FLOAT, 256x256x1x1]
%onnx::Conv_851[FLOAT, 256x256x3x3]
%onnx::Conv_854[FLOAT, 256x256x3x3]
%onnx::Conv_857[FLOAT, 256x256x1x1]
%onnx::Conv_860[FLOAT, 256x256x1x1]
%onnx::Conv_863[FLOAT, 512x256x1x1]
%onnx::Conv_864[FLOAT, 512]
%onnx::Conv_866[FLOAT, 512x512x3x3]
%onnx::Conv_869[FLOAT, 512x512x3x3]
%onnx::Conv_872[FLOAT, 512x256x1x1]
%onnx::Conv_875[FLOAT, 512x512x1x1]
%onnx::Conv_878[FLOAT, 512x512x1x1]
%onnx::Conv_881[FLOAT, 512x512x3x3]
%onnx::Conv_884[FLOAT, 512x512x3x3]
%onnx::Conv_887[FLOAT, 512x512x1x1]
%onnx::Conv_890[FLOAT, 512x512x1x1]
%onnx::Conv_893[FLOAT, 512x512x1x1]
%onnx::Conv_896[FLOAT, 512x512x3x3]
%onnx::Conv_899[FLOAT, 512x512x3x3]
%onnx::Conv_902[FLOAT, 512x512x1x1]
%onnx::Conv_905[FLOAT, 512x512x1x1]
) {
%onnx::Conv_906 = Identity(%onnx::Conv_864)
%onnx::Conv_903 = Identity(%onnx::Conv_864)
%onnx::Conv_900 = Identity(%onnx::Conv_864)
%onnx::Conv_897 = Identity(%onnx::Conv_864)
%onnx::Conv_894 = Identity(%onnx::Conv_864)
%onnx::Conv_891 = Identity(%onnx::Conv_864)
%onnx::Conv_888 = Identity(%onnx::Conv_864)
%onnx::Conv_885 = Identity(%onnx::Conv_864)
%onnx::Conv_882 = Identity(%onnx::Conv_864)
%onnx::Conv_879 = Identity(%onnx::Conv_864)
%onnx::Conv_876 = Identity(%onnx::Conv_864)
%onnx::Conv_873 = Identity(%onnx::Conv_864)
%onnx::Conv_870 = Identity(%onnx::Conv_864)
%onnx::Conv_867 = Identity(%onnx::Conv_864)
%onnx::Conv_861 = Identity(%onnx::Conv_819)
%onnx::Conv_858 = Identity(%onnx::Conv_819)
%onnx::Conv_855 = Identity(%onnx::Conv_819)
%onnx::Conv_852 = Identity(%onnx::Conv_819)
%onnx::Conv_849 = Identity(%onnx::Conv_819)
%onnx::Conv_846 = Identity(%onnx::Conv_819)
%onnx::Conv_843 = Identity(%onnx::Conv_819)
%onnx::Conv_840 = Identity(%onnx::Conv_819)
%onnx::Conv_837 = Identity(%onnx::Conv_819)
%onnx::Conv_834 = Identity(%onnx::Conv_819)
%onnx::Conv_831 = Identity(%onnx::Conv_819)
%onnx::Conv_828 = Identity(%onnx::Conv_819)
%onnx::Conv_825 = Identity(%onnx::Conv_819)
%onnx::Conv_822 = Identity(%onnx::Conv_819)
%onnx::Conv_816 = Identity(%onnx::Conv_771)
%onnx::Conv_813 = Identity(%onnx::Conv_771)
%onnx::Conv_810 = Identity(%onnx::Conv_771)
%onnx::Conv_807 = Identity(%onnx::Conv_771)
%onnx::Conv_804 = Identity(%onnx::Conv_771)
%onnx::Conv_801 = Identity(%onnx::Conv_771)
%onnx::Conv_798 = Identity(%onnx::Conv_771)
%onnx::Conv_795 = Identity(%onnx::Conv_771)
%onnx::Conv_792 = Identity(%onnx::Conv_771)
%onnx::Conv_789 = Identity(%onnx::Conv_771)
%onnx::Conv_786 = Identity(%onnx::Conv_771)
%onnx::Conv_783 = Identity(%onnx::Conv_771)
%onnx::Conv_780 = Identity(%onnx::Conv_771)
%onnx::Conv_777 = Identity(%onnx::Conv_771)
%onnx::Conv_774 = Identity(%onnx::Conv_771)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_770, %onnx::Conv_771)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_773, %onnx::Conv_774)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_776, %onnx::Conv_777)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_779, %onnx::Conv_780)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0)
%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_782, %onnx::Conv_783)
%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/Add_3_output_0, %/layers.1/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.1/Add_5_output_0 = Add(%/layers.1/Add_4_output_0, %/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_5_output_0, %onnx::Conv_785, %onnx::Conv_786)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_788, %onnx::Conv_789)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0)
%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_797, %onnx::Conv_798)
%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/Add_3_output_0, %/layers.2/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.2/Add_5_output_0 = Add(%/layers.2/Add_4_output_0, %/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_5_output_0, %onnx::Conv_800, %onnx::Conv_801)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_803, %onnx::Conv_804)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_806, %onnx::Conv_807)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_809, %onnx::Conv_810)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0)
%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_812, %onnx::Conv_813)
%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/Add_3_output_0, %/layers.3/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.3/Add_5_output_0 = Add(%/layers.3/Add_4_output_0, %/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_5_output_0, %onnx::Conv_815, %onnx::Conv_816)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_818, %onnx::Conv_819)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_821, %onnx::Conv_822)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_824, %onnx::Conv_825)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0)
%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_827, %onnx::Conv_828)
%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/Add_3_output_0, %/layers.5/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.5/Add_5_output_0 = Add(%/layers.5/Add_4_output_0, %/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_5_output_0, %onnx::Conv_830, %onnx::Conv_831)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_833, %onnx::Conv_834)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_836, %onnx::Conv_837)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_839, %onnx::Conv_840)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0)
%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_842, %onnx::Conv_843)
%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/Add_3_output_0, %/layers.6/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.6/Add_5_output_0 = Add(%/layers.6/Add_4_output_0, %/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_5_output_0, %onnx::Conv_845, %onnx::Conv_846)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_848, %onnx::Conv_849)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_851, %onnx::Conv_852)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_854, %onnx::Conv_855)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0)
%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_857, %onnx::Conv_858)
%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/Add_3_output_0, %/layers.7/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.7/Add_5_output_0 = Add(%/layers.7/Add_4_output_0, %/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_5_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0)
%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/Add_3_output_0, %/layers.9/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.9/Add_5_output_0 = Add(%/layers.9/Add_4_output_0, %/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_5_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0)
%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/Add_3_output_0, %/layers.10/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.10/Add_5_output_0 = Add(%/layers.10/Add_4_output_0, %/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_5_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0)
%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/Add_3_output_0, %/layers.11/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.11/Add_5_output_0 = Add(%/layers.11/Add_4_output_0, %/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_5_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%768 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %768
}
|
val_accuracy
| 92.568111
| 6,310,340,608
| 21,384,074
|
{'zcp_epe_nas': 89.05294920314782, 'zcp_fisher': 407.013916015625, 'zcp_flops': 100965449728.0, 'zcp_grad_norm': 318.8368225097656, 'zcp_grasp': -254.8515625, 'zcp_jacov': -16.038197340844306, 'zcp_l2_norm': 1030.62646484375, 'zcp_nwot': 232.09498752381646, 'zcp_params': 21384074.0, 'zcp_plain': 0.019305501133203, 'zcp_snip': 2574.981689453125, 'zcp_synflow': 129.91318086466458, 'zcp_zen': 98.51123809814453, 'zcp_val_accuracy': 0.9305889606475831}
| |
NASBench101_339182
|
NASBench101
|
339182
|
cd1b15dd0376294d6752bdb74fe3118c
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_707[FLOAT, 128x3x3x3]
%onnx::Conv_708[FLOAT, 128]
%onnx::Conv_710[FLOAT, 43x128x1x1]
%onnx::Conv_711[FLOAT, 43]
%onnx::Conv_713[FLOAT, 43x43x1x1]
%onnx::Conv_716[FLOAT, 43x43x3x3]
%onnx::Conv_719[FLOAT, 43x128x1x1]
%onnx::Conv_722[FLOAT, 43x128x1x1]
%onnx::Conv_725[FLOAT, 43x43x1x1]
%onnx::Conv_728[FLOAT, 43x43x3x3]
%onnx::Conv_731[FLOAT, 43x128x1x1]
%onnx::Conv_734[FLOAT, 43x128x1x1]
%onnx::Conv_737[FLOAT, 43x43x1x1]
%onnx::Conv_740[FLOAT, 43x43x3x3]
%onnx::Conv_743[FLOAT, 43x128x1x1]
%onnx::Conv_746[FLOAT, 86x128x1x1]
%onnx::Conv_747[FLOAT, 86]
%onnx::Conv_749[FLOAT, 86x86x1x1]
%onnx::Conv_752[FLOAT, 86x86x3x3]
%onnx::Conv_755[FLOAT, 85x128x1x1]
%onnx::Conv_756[FLOAT, 85]
%onnx::Conv_758[FLOAT, 86x256x1x1]
%onnx::Conv_761[FLOAT, 86x86x1x1]
%onnx::Conv_764[FLOAT, 86x86x3x3]
%onnx::Conv_767[FLOAT, 85x256x1x1]
%onnx::Conv_770[FLOAT, 86x256x1x1]
%onnx::Conv_773[FLOAT, 86x86x1x1]
%onnx::Conv_776[FLOAT, 86x86x3x3]
%onnx::Conv_779[FLOAT, 85x256x1x1]
%onnx::Conv_782[FLOAT, 171x256x1x1]
%onnx::Conv_783[FLOAT, 171]
%onnx::Conv_785[FLOAT, 171x171x1x1]
%onnx::Conv_788[FLOAT, 171x171x3x3]
%onnx::Conv_791[FLOAT, 171x256x1x1]
%onnx::Conv_794[FLOAT, 171x512x1x1]
%onnx::Conv_797[FLOAT, 171x171x1x1]
%onnx::Conv_800[FLOAT, 171x171x3x3]
%onnx::Conv_803[FLOAT, 171x512x1x1]
%onnx::Conv_806[FLOAT, 171x512x1x1]
%onnx::Conv_809[FLOAT, 171x171x1x1]
%onnx::Conv_812[FLOAT, 171x171x3x3]
%onnx::Conv_815[FLOAT, 171x512x1x1]
) {
%onnx::Conv_816 = Identity(%onnx::Conv_783)
%onnx::Conv_813 = Identity(%onnx::Conv_783)
%onnx::Conv_810 = Identity(%onnx::Conv_783)
%onnx::Conv_807 = Identity(%onnx::Conv_783)
%onnx::Conv_804 = Identity(%onnx::Conv_783)
%onnx::Conv_801 = Identity(%onnx::Conv_783)
%onnx::Conv_798 = Identity(%onnx::Conv_783)
%onnx::Conv_795 = Identity(%onnx::Conv_783)
%onnx::Conv_792 = Identity(%onnx::Conv_783)
%onnx::Conv_789 = Identity(%onnx::Conv_783)
%onnx::Conv_786 = Identity(%onnx::Conv_783)
%onnx::Conv_780 = Identity(%onnx::Conv_756)
%onnx::Conv_777 = Identity(%onnx::Conv_747)
%onnx::Conv_774 = Identity(%onnx::Conv_747)
%onnx::Conv_771 = Identity(%onnx::Conv_747)
%onnx::Conv_768 = Identity(%onnx::Conv_756)
%onnx::Conv_765 = Identity(%onnx::Conv_747)
%onnx::Conv_762 = Identity(%onnx::Conv_747)
%onnx::Conv_759 = Identity(%onnx::Conv_747)
%onnx::Conv_753 = Identity(%onnx::Conv_747)
%onnx::Conv_750 = Identity(%onnx::Conv_747)
%onnx::Conv_744 = Identity(%onnx::Conv_711)
%onnx::Conv_741 = Identity(%onnx::Conv_711)
%onnx::Conv_738 = Identity(%onnx::Conv_711)
%onnx::Conv_735 = Identity(%onnx::Conv_711)
%onnx::Conv_732 = Identity(%onnx::Conv_711)
%onnx::Conv_729 = Identity(%onnx::Conv_711)
%onnx::Conv_726 = Identity(%onnx::Conv_711)
%onnx::Conv_723 = Identity(%onnx::Conv_711)
%onnx::Conv_720 = Identity(%onnx::Conv_711)
%onnx::Conv_717 = Identity(%onnx::Conv_711)
%onnx::Conv_714 = Identity(%onnx::Conv_711)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_707, %onnx::Conv_708)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_710, %onnx::Conv_711)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_713, %onnx::Conv_714)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_716, %onnx::Conv_717)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_719, %onnx::Conv_720)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.3/maxpool/MaxPool_output_0, %/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.1/Slice_output_0 = Slice(%/layers.1/vertex_op.3/maxpool/MaxPool_output_0, %/layers.1/Constant_4_output_0, %/layers.1/Constant_5_output_0, %/layers.1/Constant_3_output_0, %/layers.1/Constant_6_output_0)
%/layers.1/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Slice_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.4/maxpool/MaxPool_output_0, %/layers.1/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_722, %onnx::Conv_723)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_725, %onnx::Conv_726)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_728, %onnx::Conv_729)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_731, %onnx::Conv_732)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.3/maxpool/MaxPool_output_0, %/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.2/Slice_output_0 = Slice(%/layers.2/vertex_op.3/maxpool/MaxPool_output_0, %/layers.2/Constant_4_output_0, %/layers.2/Constant_5_output_0, %/layers.2/Constant_3_output_0, %/layers.2/Constant_6_output_0)
%/layers.2/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Slice_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.4/maxpool/MaxPool_output_0, %/layers.2/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_734, %onnx::Conv_735)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_737, %onnx::Conv_738)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_740, %onnx::Conv_741)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_743, %onnx::Conv_744)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.3/maxpool/MaxPool_output_0, %/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.3/Slice_output_0 = Slice(%/layers.3/vertex_op.3/maxpool/MaxPool_output_0, %/layers.3/Constant_4_output_0, %/layers.3/Constant_5_output_0, %/layers.3/Constant_3_output_0, %/layers.3/Constant_6_output_0)
%/layers.3/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Slice_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.4/maxpool/MaxPool_output_0, %/layers.3/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_746, %onnx::Conv_747)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_749, %onnx::Conv_750)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_752, %onnx::Conv_753)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.5/Slice_output_0 = Slice(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0, %/layers.5/Constant_4_output_0, %/layers.5/Constant_2_output_0, %/layers.5/Constant_5_output_0)
%/layers.5/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/Slice_output_0, %/layers.5/Constant_6_output_0)
%/layers.5/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_755, %onnx::Conv_756)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.3/maxpool/MaxPool_output_0, %/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0)
%/layers.5/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.4/maxpool/MaxPool_output_0, %/layers.5/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_758, %onnx::Conv_759)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_761, %onnx::Conv_762)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_764, %onnx::Conv_765)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.6/Slice_output_0 = Slice(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0, %/layers.6/Constant_4_output_0, %/layers.6/Constant_2_output_0, %/layers.6/Constant_5_output_0)
%/layers.6/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/Slice_output_0, %/layers.6/Constant_6_output_0)
%/layers.6/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_767, %onnx::Conv_768)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.3/maxpool/MaxPool_output_0, %/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0)
%/layers.6/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.4/maxpool/MaxPool_output_0, %/layers.6/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_770, %onnx::Conv_771)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_773, %onnx::Conv_774)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_776, %onnx::Conv_777)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.7/Slice_output_0 = Slice(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0, %/layers.7/Constant_4_output_0, %/layers.7/Constant_2_output_0, %/layers.7/Constant_5_output_0)
%/layers.7/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/Slice_output_0, %/layers.7/Constant_6_output_0)
%/layers.7/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_779, %onnx::Conv_780)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.3/maxpool/MaxPool_output_0, %/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0)
%/layers.7/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.4/maxpool/MaxPool_output_0, %/layers.7/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_782, %onnx::Conv_783)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_785, %onnx::Conv_786)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_788, %onnx::Conv_789)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.3/maxpool/MaxPool_output_0, %/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.9/Slice_output_0 = Slice(%/layers.9/vertex_op.3/maxpool/MaxPool_output_0, %/layers.9/Constant_4_output_0, %/layers.9/Constant_5_output_0, %/layers.9/Constant_3_output_0, %/layers.9/Constant_6_output_0)
%/layers.9/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Slice_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.4/maxpool/MaxPool_output_0, %/layers.9/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_797, %onnx::Conv_798)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_800, %onnx::Conv_801)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_803, %onnx::Conv_804)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.3/maxpool/MaxPool_output_0, %/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.10/Slice_output_0 = Slice(%/layers.10/vertex_op.3/maxpool/MaxPool_output_0, %/layers.10/Constant_4_output_0, %/layers.10/Constant_5_output_0, %/layers.10/Constant_3_output_0, %/layers.10/Constant_6_output_0)
%/layers.10/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Slice_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.4/maxpool/MaxPool_output_0, %/layers.10/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_806, %onnx::Conv_807)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_809, %onnx::Conv_810)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_812, %onnx::Conv_813)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_815, %onnx::Conv_816)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.3/maxpool/MaxPool_output_0, %/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.11/Slice_output_0 = Slice(%/layers.11/vertex_op.3/maxpool/MaxPool_output_0, %/layers.11/Constant_4_output_0, %/layers.11/Constant_5_output_0, %/layers.11/Constant_3_output_0, %/layers.11/Constant_6_output_0)
%/layers.11/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Slice_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.4/maxpool/MaxPool_output_0, %/layers.11/vertex_op.5/maxpool/MaxPool_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%705 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %705
}
|
val_accuracy
| 89.663464
| 534,712,064
| 1,750,840
|
{'zcp_epe_nas': 80.95937309814249, 'zcp_fisher': 21.441368103027344, 'zcp_flops': 8555393024.0, 'zcp_grad_norm': 83.6706771850586, 'zcp_grasp': -8.0030517578125, 'zcp_jacov': -16.043491666326183, 'zcp_l2_norm': 640.030029296875, 'zcp_nwot': 212.7605433137477, 'zcp_params': 1750840.0, 'zcp_plain': 0.052668645977973, 'zcp_snip': 415.2268371582031, 'zcp_synflow': 79.01854789041302, 'zcp_zen': 62.615928649902344, 'zcp_val_accuracy': 0.8781049847602841}
| |
NASBench101_318652
|
NASBench101
|
318652
|
c0c8504180f138588eb6ec2bf4cd937e
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_1076[FLOAT, 128x3x3x3]
%onnx::Conv_1077[FLOAT, 128]
%onnx::Conv_1079[FLOAT, 128x128x1x1]
%onnx::Conv_1082[FLOAT, 128x128x1x1]
%onnx::Conv_1085[FLOAT, 128x128x3x3]
%onnx::Conv_1088[FLOAT, 128x128x1x1]
%onnx::Conv_1091[FLOAT, 128x128x3x3]
%onnx::Conv_1094[FLOAT, 128x128x1x1]
%onnx::Conv_1097[FLOAT, 128x128x3x3]
%onnx::Conv_1100[FLOAT, 128x128x1x1]
%onnx::Conv_1103[FLOAT, 128x128x1x1]
%onnx::Conv_1106[FLOAT, 128x128x1x1]
%onnx::Conv_1109[FLOAT, 128x128x3x3]
%onnx::Conv_1112[FLOAT, 128x128x1x1]
%onnx::Conv_1115[FLOAT, 128x128x3x3]
%onnx::Conv_1118[FLOAT, 128x128x1x1]
%onnx::Conv_1121[FLOAT, 128x128x3x3]
%onnx::Conv_1124[FLOAT, 128x128x1x1]
%onnx::Conv_1127[FLOAT, 128x128x1x1]
%onnx::Conv_1130[FLOAT, 128x128x1x1]
%onnx::Conv_1133[FLOAT, 128x128x3x3]
%onnx::Conv_1136[FLOAT, 128x128x1x1]
%onnx::Conv_1139[FLOAT, 128x128x3x3]
%onnx::Conv_1142[FLOAT, 128x128x1x1]
%onnx::Conv_1145[FLOAT, 128x128x3x3]
%onnx::Conv_1148[FLOAT, 128x128x1x1]
%onnx::Conv_1151[FLOAT, 256x128x1x1]
%onnx::Conv_1152[FLOAT, 256]
%onnx::Conv_1154[FLOAT, 256x256x1x1]
%onnx::Conv_1157[FLOAT, 256x256x3x3]
%onnx::Conv_1160[FLOAT, 256x128x1x1]
%onnx::Conv_1163[FLOAT, 256x256x3x3]
%onnx::Conv_1166[FLOAT, 256x128x1x1]
%onnx::Conv_1169[FLOAT, 256x256x3x3]
%onnx::Conv_1172[FLOAT, 256x256x1x1]
%onnx::Conv_1175[FLOAT, 256x256x1x1]
%onnx::Conv_1178[FLOAT, 256x256x1x1]
%onnx::Conv_1181[FLOAT, 256x256x3x3]
%onnx::Conv_1184[FLOAT, 256x256x1x1]
%onnx::Conv_1187[FLOAT, 256x256x3x3]
%onnx::Conv_1190[FLOAT, 256x256x1x1]
%onnx::Conv_1193[FLOAT, 256x256x3x3]
%onnx::Conv_1196[FLOAT, 256x256x1x1]
%onnx::Conv_1199[FLOAT, 256x256x1x1]
%onnx::Conv_1202[FLOAT, 256x256x1x1]
%onnx::Conv_1205[FLOAT, 256x256x3x3]
%onnx::Conv_1208[FLOAT, 256x256x1x1]
%onnx::Conv_1211[FLOAT, 256x256x3x3]
%onnx::Conv_1214[FLOAT, 256x256x1x1]
%onnx::Conv_1217[FLOAT, 256x256x3x3]
%onnx::Conv_1220[FLOAT, 256x256x1x1]
%onnx::Conv_1223[FLOAT, 512x256x1x1]
%onnx::Conv_1224[FLOAT, 512]
%onnx::Conv_1226[FLOAT, 512x512x1x1]
%onnx::Conv_1229[FLOAT, 512x512x3x3]
%onnx::Conv_1232[FLOAT, 512x256x1x1]
%onnx::Conv_1235[FLOAT, 512x512x3x3]
%onnx::Conv_1238[FLOAT, 512x256x1x1]
%onnx::Conv_1241[FLOAT, 512x512x3x3]
%onnx::Conv_1244[FLOAT, 512x512x1x1]
%onnx::Conv_1247[FLOAT, 512x512x1x1]
%onnx::Conv_1250[FLOAT, 512x512x1x1]
%onnx::Conv_1253[FLOAT, 512x512x3x3]
%onnx::Conv_1256[FLOAT, 512x512x1x1]
%onnx::Conv_1259[FLOAT, 512x512x3x3]
%onnx::Conv_1262[FLOAT, 512x512x1x1]
%onnx::Conv_1265[FLOAT, 512x512x3x3]
%onnx::Conv_1268[FLOAT, 512x512x1x1]
%onnx::Conv_1271[FLOAT, 512x512x1x1]
%onnx::Conv_1274[FLOAT, 512x512x1x1]
%onnx::Conv_1277[FLOAT, 512x512x3x3]
%onnx::Conv_1280[FLOAT, 512x512x1x1]
%onnx::Conv_1283[FLOAT, 512x512x3x3]
%onnx::Conv_1286[FLOAT, 512x512x1x1]
%onnx::Conv_1289[FLOAT, 512x512x3x3]
%onnx::Conv_1292[FLOAT, 512x512x1x1]
) {
%onnx::Conv_1293 = Identity(%onnx::Conv_1224)
%onnx::Conv_1290 = Identity(%onnx::Conv_1224)
%onnx::Conv_1287 = Identity(%onnx::Conv_1224)
%onnx::Conv_1284 = Identity(%onnx::Conv_1224)
%onnx::Conv_1281 = Identity(%onnx::Conv_1224)
%onnx::Conv_1278 = Identity(%onnx::Conv_1224)
%onnx::Conv_1275 = Identity(%onnx::Conv_1224)
%onnx::Conv_1272 = Identity(%onnx::Conv_1224)
%onnx::Conv_1269 = Identity(%onnx::Conv_1224)
%onnx::Conv_1266 = Identity(%onnx::Conv_1224)
%onnx::Conv_1263 = Identity(%onnx::Conv_1224)
%onnx::Conv_1260 = Identity(%onnx::Conv_1224)
%onnx::Conv_1257 = Identity(%onnx::Conv_1224)
%onnx::Conv_1254 = Identity(%onnx::Conv_1224)
%onnx::Conv_1251 = Identity(%onnx::Conv_1224)
%onnx::Conv_1248 = Identity(%onnx::Conv_1224)
%onnx::Conv_1245 = Identity(%onnx::Conv_1224)
%onnx::Conv_1242 = Identity(%onnx::Conv_1224)
%onnx::Conv_1239 = Identity(%onnx::Conv_1224)
%onnx::Conv_1236 = Identity(%onnx::Conv_1224)
%onnx::Conv_1233 = Identity(%onnx::Conv_1224)
%onnx::Conv_1230 = Identity(%onnx::Conv_1224)
%onnx::Conv_1227 = Identity(%onnx::Conv_1224)
%onnx::Conv_1221 = Identity(%onnx::Conv_1152)
%onnx::Conv_1218 = Identity(%onnx::Conv_1152)
%onnx::Conv_1215 = Identity(%onnx::Conv_1152)
%onnx::Conv_1212 = Identity(%onnx::Conv_1152)
%onnx::Conv_1209 = Identity(%onnx::Conv_1152)
%onnx::Conv_1206 = Identity(%onnx::Conv_1152)
%onnx::Conv_1203 = Identity(%onnx::Conv_1152)
%onnx::Conv_1200 = Identity(%onnx::Conv_1152)
%onnx::Conv_1197 = Identity(%onnx::Conv_1152)
%onnx::Conv_1194 = Identity(%onnx::Conv_1152)
%onnx::Conv_1191 = Identity(%onnx::Conv_1152)
%onnx::Conv_1188 = Identity(%onnx::Conv_1152)
%onnx::Conv_1185 = Identity(%onnx::Conv_1152)
%onnx::Conv_1182 = Identity(%onnx::Conv_1152)
%onnx::Conv_1179 = Identity(%onnx::Conv_1152)
%onnx::Conv_1176 = Identity(%onnx::Conv_1152)
%onnx::Conv_1173 = Identity(%onnx::Conv_1152)
%onnx::Conv_1170 = Identity(%onnx::Conv_1152)
%onnx::Conv_1167 = Identity(%onnx::Conv_1152)
%onnx::Conv_1164 = Identity(%onnx::Conv_1152)
%onnx::Conv_1161 = Identity(%onnx::Conv_1152)
%onnx::Conv_1158 = Identity(%onnx::Conv_1152)
%onnx::Conv_1155 = Identity(%onnx::Conv_1152)
%onnx::Conv_1149 = Identity(%onnx::Conv_1077)
%onnx::Conv_1146 = Identity(%onnx::Conv_1077)
%onnx::Conv_1143 = Identity(%onnx::Conv_1077)
%onnx::Conv_1140 = Identity(%onnx::Conv_1077)
%onnx::Conv_1137 = Identity(%onnx::Conv_1077)
%onnx::Conv_1134 = Identity(%onnx::Conv_1077)
%onnx::Conv_1131 = Identity(%onnx::Conv_1077)
%onnx::Conv_1128 = Identity(%onnx::Conv_1077)
%onnx::Conv_1125 = Identity(%onnx::Conv_1077)
%onnx::Conv_1122 = Identity(%onnx::Conv_1077)
%onnx::Conv_1119 = Identity(%onnx::Conv_1077)
%onnx::Conv_1116 = Identity(%onnx::Conv_1077)
%onnx::Conv_1113 = Identity(%onnx::Conv_1077)
%onnx::Conv_1110 = Identity(%onnx::Conv_1077)
%onnx::Conv_1107 = Identity(%onnx::Conv_1077)
%onnx::Conv_1104 = Identity(%onnx::Conv_1077)
%onnx::Conv_1101 = Identity(%onnx::Conv_1077)
%onnx::Conv_1098 = Identity(%onnx::Conv_1077)
%onnx::Conv_1095 = Identity(%onnx::Conv_1077)
%onnx::Conv_1092 = Identity(%onnx::Conv_1077)
%onnx::Conv_1089 = Identity(%onnx::Conv_1077)
%onnx::Conv_1086 = Identity(%onnx::Conv_1077)
%onnx::Conv_1083 = Identity(%onnx::Conv_1077)
%onnx::Conv_1080 = Identity(%onnx::Conv_1077)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_1076, %onnx::Conv_1077)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1079, %onnx::Conv_1080)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_1082, %onnx::Conv_1083)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_1085, %onnx::Conv_1086)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1088, %onnx::Conv_1089)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_1091, %onnx::Conv_1092)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1094, %onnx::Conv_1095)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/Add_3_output_0, %/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_1097, %onnx::Conv_1098)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_5_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_4_output_0)
%/layers.1/Add_6_output_0 = Add(%/layers.1/Add_5_output_0, %/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/Add_7_output_0 = Add(%/layers.1/Add_6_output_0, %/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_7_output_0, %onnx::Conv_1100, %onnx::Conv_1101)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1103, %onnx::Conv_1104)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_1106, %onnx::Conv_1107)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_1109, %onnx::Conv_1110)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1112, %onnx::Conv_1113)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_1115, %onnx::Conv_1116)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1118, %onnx::Conv_1119)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/Add_3_output_0, %/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_1121, %onnx::Conv_1122)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_5_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_4_output_0)
%/layers.2/Add_6_output_0 = Add(%/layers.2/Add_5_output_0, %/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/Add_7_output_0 = Add(%/layers.2/Add_6_output_0, %/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_7_output_0, %onnx::Conv_1124, %onnx::Conv_1125)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1127, %onnx::Conv_1128)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_1130, %onnx::Conv_1131)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_1133, %onnx::Conv_1134)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1136, %onnx::Conv_1137)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_1139, %onnx::Conv_1140)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1142, %onnx::Conv_1143)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/Add_3_output_0, %/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_4_output_0, %onnx::Conv_1145, %onnx::Conv_1146)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_5_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_4_output_0)
%/layers.3/Add_6_output_0 = Add(%/layers.3/Add_5_output_0, %/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/Add_7_output_0 = Add(%/layers.3/Add_6_output_0, %/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_7_output_0, %onnx::Conv_1148, %onnx::Conv_1149)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1151, %onnx::Conv_1152)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_1154, %onnx::Conv_1155)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_1157, %onnx::Conv_1158)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1160, %onnx::Conv_1161)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_1163, %onnx::Conv_1164)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1166, %onnx::Conv_1167)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/Add_3_output_0, %/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_1169, %onnx::Conv_1170)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_5_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_4_output_0)
%/layers.5/Add_6_output_0 = Add(%/layers.5/Add_5_output_0, %/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/Add_7_output_0 = Add(%/layers.5/Add_6_output_0, %/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_7_output_0, %onnx::Conv_1172, %onnx::Conv_1173)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1175, %onnx::Conv_1176)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_1178, %onnx::Conv_1179)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_1181, %onnx::Conv_1182)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1184, %onnx::Conv_1185)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_1187, %onnx::Conv_1188)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1190, %onnx::Conv_1191)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/Add_3_output_0, %/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_1193, %onnx::Conv_1194)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_5_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_4_output_0)
%/layers.6/Add_6_output_0 = Add(%/layers.6/Add_5_output_0, %/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/Add_7_output_0 = Add(%/layers.6/Add_6_output_0, %/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_7_output_0, %onnx::Conv_1196, %onnx::Conv_1197)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1199, %onnx::Conv_1200)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_1202, %onnx::Conv_1203)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_1205, %onnx::Conv_1206)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1208, %onnx::Conv_1209)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_1211, %onnx::Conv_1212)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1214, %onnx::Conv_1215)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/Add_3_output_0, %/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_4_output_0, %onnx::Conv_1217, %onnx::Conv_1218)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_5_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_4_output_0)
%/layers.7/Add_6_output_0 = Add(%/layers.7/Add_5_output_0, %/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/Add_7_output_0 = Add(%/layers.7/Add_6_output_0, %/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_7_output_0, %onnx::Conv_1220, %onnx::Conv_1221)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1223, %onnx::Conv_1224)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_1226, %onnx::Conv_1227)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_1229, %onnx::Conv_1230)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1232, %onnx::Conv_1233)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_1235, %onnx::Conv_1236)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1238, %onnx::Conv_1239)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/Add_3_output_0, %/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_1241, %onnx::Conv_1242)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_5_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_4_output_0)
%/layers.9/Add_6_output_0 = Add(%/layers.9/Add_5_output_0, %/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/Add_7_output_0 = Add(%/layers.9/Add_6_output_0, %/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_7_output_0, %onnx::Conv_1244, %onnx::Conv_1245)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1247, %onnx::Conv_1248)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_1250, %onnx::Conv_1251)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_1253, %onnx::Conv_1254)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1256, %onnx::Conv_1257)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_1259, %onnx::Conv_1260)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1262, %onnx::Conv_1263)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/Add_3_output_0, %/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_1265, %onnx::Conv_1266)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_5_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_4_output_0)
%/layers.10/Add_6_output_0 = Add(%/layers.10/Add_5_output_0, %/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/Add_7_output_0 = Add(%/layers.10/Add_6_output_0, %/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_7_output_0, %onnx::Conv_1268, %onnx::Conv_1269)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1271, %onnx::Conv_1272)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_1274, %onnx::Conv_1275)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_1277, %onnx::Conv_1278)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1280, %onnx::Conv_1281)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_1283, %onnx::Conv_1284)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1286, %onnx::Conv_1287)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/Add_3_output_0, %/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_4_output_0, %onnx::Conv_1289, %onnx::Conv_1290)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_5_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_4_output_0)
%/layers.11/Add_6_output_0 = Add(%/layers.11/Add_5_output_0, %/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/Add_7_output_0 = Add(%/layers.11/Add_6_output_0, %/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_7_output_0, %onnx::Conv_1292, %onnx::Conv_1293)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%1074 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %1074
}
|
val_accuracy
| 92.948717
| 9,615,190,016
| 32,590,474
|
{'zcp_epe_nas': 63.44345519591789, 'zcp_fisher': 33.09636306762695, 'zcp_flops': 153843040256.0, 'zcp_grad_norm': 136.58865356445312, 'zcp_grasp': -18.64599609375, 'zcp_jacov': -16.057134214085977, 'zcp_l2_norm': 1650.8330078125, 'zcp_nwot': 239.51376542020262, 'zcp_params': 32590474.0, 'zcp_plain': -0.040858708322048, 'zcp_snip': 1104.6181640625, 'zcp_synflow': 131.8021744870423, 'zcp_zen': 145.50662231445312, 'zcp_val_accuracy': 0.901943087577819}
| |
NASBench101_39542
|
NASBench101
|
39542
|
17fde098dd0c887a24d540f3a5e1554c
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_824[FLOAT, 128x3x3x3]
%onnx::Conv_825[FLOAT, 128]
%onnx::Conv_827[FLOAT, 43x128x1x1]
%onnx::Conv_828[FLOAT, 43]
%onnx::Conv_830[FLOAT, 43x43x3x3]
%onnx::Conv_833[FLOAT, 43x128x1x1]
%onnx::Conv_836[FLOAT, 42x128x1x1]
%onnx::Conv_837[FLOAT, 42]
%onnx::Conv_839[FLOAT, 42x42x1x1]
%onnx::Conv_842[FLOAT, 43x128x1x1]
%onnx::Conv_845[FLOAT, 43x43x3x3]
%onnx::Conv_848[FLOAT, 43x128x1x1]
%onnx::Conv_851[FLOAT, 42x128x1x1]
%onnx::Conv_854[FLOAT, 42x42x1x1]
%onnx::Conv_857[FLOAT, 43x128x1x1]
%onnx::Conv_860[FLOAT, 43x43x3x3]
%onnx::Conv_863[FLOAT, 43x128x1x1]
%onnx::Conv_866[FLOAT, 42x128x1x1]
%onnx::Conv_869[FLOAT, 42x42x1x1]
%onnx::Conv_872[FLOAT, 86x128x1x1]
%onnx::Conv_873[FLOAT, 86]
%onnx::Conv_875[FLOAT, 86x86x3x3]
%onnx::Conv_878[FLOAT, 85x128x1x1]
%onnx::Conv_879[FLOAT, 85]
%onnx::Conv_881[FLOAT, 85x128x1x1]
%onnx::Conv_884[FLOAT, 85x85x1x1]
%onnx::Conv_887[FLOAT, 86x256x1x1]
%onnx::Conv_890[FLOAT, 86x86x3x3]
%onnx::Conv_893[FLOAT, 85x256x1x1]
%onnx::Conv_896[FLOAT, 85x256x1x1]
%onnx::Conv_899[FLOAT, 85x85x1x1]
%onnx::Conv_902[FLOAT, 86x256x1x1]
%onnx::Conv_905[FLOAT, 86x86x3x3]
%onnx::Conv_908[FLOAT, 85x256x1x1]
%onnx::Conv_911[FLOAT, 85x256x1x1]
%onnx::Conv_914[FLOAT, 85x85x1x1]
%onnx::Conv_917[FLOAT, 171x256x1x1]
%onnx::Conv_918[FLOAT, 171]
%onnx::Conv_920[FLOAT, 171x171x3x3]
%onnx::Conv_923[FLOAT, 171x256x1x1]
%onnx::Conv_926[FLOAT, 170x256x1x1]
%onnx::Conv_927[FLOAT, 170]
%onnx::Conv_929[FLOAT, 170x170x1x1]
%onnx::Conv_932[FLOAT, 171x512x1x1]
%onnx::Conv_935[FLOAT, 171x171x3x3]
%onnx::Conv_938[FLOAT, 171x512x1x1]
%onnx::Conv_941[FLOAT, 170x512x1x1]
%onnx::Conv_944[FLOAT, 170x170x1x1]
%onnx::Conv_947[FLOAT, 171x512x1x1]
%onnx::Conv_950[FLOAT, 171x171x3x3]
%onnx::Conv_953[FLOAT, 171x512x1x1]
%onnx::Conv_956[FLOAT, 170x512x1x1]
%onnx::Conv_959[FLOAT, 170x170x1x1]
) {
%onnx::Conv_960 = Identity(%onnx::Conv_927)
%onnx::Conv_957 = Identity(%onnx::Conv_927)
%onnx::Conv_954 = Identity(%onnx::Conv_918)
%onnx::Conv_951 = Identity(%onnx::Conv_918)
%onnx::Conv_948 = Identity(%onnx::Conv_918)
%onnx::Conv_945 = Identity(%onnx::Conv_927)
%onnx::Conv_942 = Identity(%onnx::Conv_927)
%onnx::Conv_939 = Identity(%onnx::Conv_918)
%onnx::Conv_936 = Identity(%onnx::Conv_918)
%onnx::Conv_933 = Identity(%onnx::Conv_918)
%onnx::Conv_930 = Identity(%onnx::Conv_927)
%onnx::Conv_924 = Identity(%onnx::Conv_918)
%onnx::Conv_921 = Identity(%onnx::Conv_918)
%onnx::Conv_915 = Identity(%onnx::Conv_879)
%onnx::Conv_912 = Identity(%onnx::Conv_879)
%onnx::Conv_909 = Identity(%onnx::Conv_879)
%onnx::Conv_906 = Identity(%onnx::Conv_873)
%onnx::Conv_903 = Identity(%onnx::Conv_873)
%onnx::Conv_900 = Identity(%onnx::Conv_879)
%onnx::Conv_897 = Identity(%onnx::Conv_879)
%onnx::Conv_894 = Identity(%onnx::Conv_879)
%onnx::Conv_891 = Identity(%onnx::Conv_873)
%onnx::Conv_888 = Identity(%onnx::Conv_873)
%onnx::Conv_885 = Identity(%onnx::Conv_879)
%onnx::Conv_882 = Identity(%onnx::Conv_879)
%onnx::Conv_876 = Identity(%onnx::Conv_873)
%onnx::Conv_870 = Identity(%onnx::Conv_837)
%onnx::Conv_867 = Identity(%onnx::Conv_837)
%onnx::Conv_864 = Identity(%onnx::Conv_828)
%onnx::Conv_861 = Identity(%onnx::Conv_828)
%onnx::Conv_858 = Identity(%onnx::Conv_828)
%onnx::Conv_855 = Identity(%onnx::Conv_837)
%onnx::Conv_852 = Identity(%onnx::Conv_837)
%onnx::Conv_849 = Identity(%onnx::Conv_828)
%onnx::Conv_846 = Identity(%onnx::Conv_828)
%onnx::Conv_843 = Identity(%onnx::Conv_828)
%onnx::Conv_840 = Identity(%onnx::Conv_837)
%onnx::Conv_834 = Identity(%onnx::Conv_828)
%onnx::Conv_831 = Identity(%onnx::Conv_828)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_824, %onnx::Conv_825)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_827, %onnx::Conv_828)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_830, %onnx::Conv_831)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_833, %onnx::Conv_834)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.1/Slice_output_0 = Slice(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0, %/layers.1/Constant_4_output_0, %/layers.1/Constant_2_output_0, %/layers.1/Constant_5_output_0)
%/layers.1/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/Slice_output_0, %/layers.1/Constant_6_output_0)
%/layers.1/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_836, %onnx::Conv_837)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_7_output_0)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_839, %onnx::Conv_840)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.3/maxpool/MaxPool_output_0, %/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_4_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %/layers.1/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_842, %onnx::Conv_843)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_845, %onnx::Conv_846)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_848, %onnx::Conv_849)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.2/Slice_output_0 = Slice(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0, %/layers.2/Constant_4_output_0, %/layers.2/Constant_2_output_0, %/layers.2/Constant_5_output_0)
%/layers.2/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/Slice_output_0, %/layers.2/Constant_6_output_0)
%/layers.2/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_851, %onnx::Conv_852)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_7_output_0)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_854, %onnx::Conv_855)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.3/maxpool/MaxPool_output_0, %/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_4_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %/layers.2/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_857, %onnx::Conv_858)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.3/Slice_output_0 = Slice(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0, %/layers.3/Constant_4_output_0, %/layers.3/Constant_2_output_0, %/layers.3/Constant_5_output_0)
%/layers.3/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/Slice_output_0, %/layers.3/Constant_6_output_0)
%/layers.3/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_7_output_0)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.3/maxpool/MaxPool_output_0, %/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_4_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %/layers.3/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.5/Slice_output_0 = Slice(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0, %/layers.5/Constant_4_output_0, %/layers.5/Constant_2_output_0, %/layers.5/Constant_5_output_0)
%/layers.5/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/Slice_output_0, %/layers.5/Constant_6_output_0)
%/layers.5/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_7_output_0)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.3/maxpool/MaxPool_output_0, %/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_4_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %/layers.5/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.6/Slice_output_0 = Slice(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0, %/layers.6/Constant_4_output_0, %/layers.6/Constant_2_output_0, %/layers.6/Constant_5_output_0)
%/layers.6/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/Slice_output_0, %/layers.6/Constant_6_output_0)
%/layers.6/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_7_output_0)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.3/maxpool/MaxPool_output_0, %/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_4_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %/layers.6/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_908, %onnx::Conv_909)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.7/Slice_output_0 = Slice(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0, %/layers.7/Constant_4_output_0, %/layers.7/Constant_2_output_0, %/layers.7/Constant_5_output_0)
%/layers.7/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/Slice_output_0, %/layers.7/Constant_6_output_0)
%/layers.7/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_911, %onnx::Conv_912)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_7_output_0)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_914, %onnx::Conv_915)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.3/maxpool/MaxPool_output_0, %/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_4_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %/layers.7/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_917, %onnx::Conv_918)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_920, %onnx::Conv_921)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_923, %onnx::Conv_924)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.9/Slice_output_0 = Slice(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0, %/layers.9/Constant_4_output_0, %/layers.9/Constant_2_output_0, %/layers.9/Constant_5_output_0)
%/layers.9/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/Slice_output_0, %/layers.9/Constant_6_output_0)
%/layers.9/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_926, %onnx::Conv_927)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_7_output_0)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_929, %onnx::Conv_930)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.3/maxpool/MaxPool_output_0, %/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_4_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %/layers.9/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_932, %onnx::Conv_933)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_935, %onnx::Conv_936)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_938, %onnx::Conv_939)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.10/Slice_output_0 = Slice(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0, %/layers.10/Constant_4_output_0, %/layers.10/Constant_2_output_0, %/layers.10/Constant_5_output_0)
%/layers.10/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/Slice_output_0, %/layers.10/Constant_6_output_0)
%/layers.10/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_941, %onnx::Conv_942)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_7_output_0)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_944, %onnx::Conv_945)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.3/maxpool/MaxPool_output_0, %/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_4_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %/layers.10/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_947, %onnx::Conv_948)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_950, %onnx::Conv_951)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_953, %onnx::Conv_954)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.11/Slice_output_0 = Slice(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0, %/layers.11/Constant_4_output_0, %/layers.11/Constant_2_output_0, %/layers.11/Constant_5_output_0)
%/layers.11/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/Slice_output_0, %/layers.11/Constant_6_output_0)
%/layers.11/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_956, %onnx::Conv_957)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_7_output_0)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_959, %onnx::Conv_960)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.3/maxpool/MaxPool_output_0, %/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_4_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %/layers.11/vertex_op.5/maxpool/MaxPool_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%822 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %822
}
|
val_accuracy
| 91.446316
| 624,315,264
| 2,038,941
|
{'zcp_epe_nas': 84.14145386295604, 'zcp_fisher': 10.204981803894043, 'zcp_flops': 9989044224.0, 'zcp_grad_norm': 61.918460845947266, 'zcp_grasp': -6.647186279296875, 'zcp_jacov': -16.063475026495926, 'zcp_l2_norm': 835.1504516601562, 'zcp_nwot': 215.699889655845, 'zcp_params': 2038941.0, 'zcp_plain': 0.069465935230255, 'zcp_snip': 322.7650146484375, 'zcp_synflow': 64.89651284195361, 'zcp_zen': 76.90886688232422, 'zcp_val_accuracy': 0.925480782985687}
| |
NASBench101_376211
|
NASBench101
|
376211
|
e3720b316c5c49f999778cfda552fe76
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_860[FLOAT, 128x3x3x3]
%onnx::Conv_861[FLOAT, 128]
%onnx::Conv_863[FLOAT, 128x128x1x1]
%onnx::Conv_866[FLOAT, 128x128x1x1]
%onnx::Conv_869[FLOAT, 128x128x3x3]
%onnx::Conv_872[FLOAT, 128x128x3x3]
%onnx::Conv_875[FLOAT, 128x128x3x3]
%onnx::Conv_878[FLOAT, 128x128x1x1]
%onnx::Conv_881[FLOAT, 128x128x1x1]
%onnx::Conv_884[FLOAT, 128x128x1x1]
%onnx::Conv_887[FLOAT, 128x128x3x3]
%onnx::Conv_890[FLOAT, 128x128x3x3]
%onnx::Conv_893[FLOAT, 128x128x3x3]
%onnx::Conv_896[FLOAT, 128x128x1x1]
%onnx::Conv_899[FLOAT, 128x128x1x1]
%onnx::Conv_902[FLOAT, 128x128x1x1]
%onnx::Conv_905[FLOAT, 128x128x3x3]
%onnx::Conv_908[FLOAT, 128x128x3x3]
%onnx::Conv_911[FLOAT, 128x128x3x3]
%onnx::Conv_914[FLOAT, 128x128x1x1]
%onnx::Conv_917[FLOAT, 256x128x1x1]
%onnx::Conv_918[FLOAT, 256]
%onnx::Conv_920[FLOAT, 256x128x1x1]
%onnx::Conv_923[FLOAT, 256x256x3x3]
%onnx::Conv_926[FLOAT, 256x256x3x3]
%onnx::Conv_929[FLOAT, 256x256x3x3]
%onnx::Conv_932[FLOAT, 256x128x1x1]
%onnx::Conv_935[FLOAT, 256x256x1x1]
%onnx::Conv_938[FLOAT, 256x256x1x1]
%onnx::Conv_941[FLOAT, 256x256x3x3]
%onnx::Conv_944[FLOAT, 256x256x3x3]
%onnx::Conv_947[FLOAT, 256x256x3x3]
%onnx::Conv_950[FLOAT, 256x256x1x1]
%onnx::Conv_953[FLOAT, 256x256x1x1]
%onnx::Conv_956[FLOAT, 256x256x1x1]
%onnx::Conv_959[FLOAT, 256x256x3x3]
%onnx::Conv_962[FLOAT, 256x256x3x3]
%onnx::Conv_965[FLOAT, 256x256x3x3]
%onnx::Conv_968[FLOAT, 256x256x1x1]
%onnx::Conv_971[FLOAT, 512x256x1x1]
%onnx::Conv_972[FLOAT, 512]
%onnx::Conv_974[FLOAT, 512x256x1x1]
%onnx::Conv_977[FLOAT, 512x512x3x3]
%onnx::Conv_980[FLOAT, 512x512x3x3]
%onnx::Conv_983[FLOAT, 512x512x3x3]
%onnx::Conv_986[FLOAT, 512x256x1x1]
%onnx::Conv_989[FLOAT, 512x512x1x1]
%onnx::Conv_992[FLOAT, 512x512x1x1]
%onnx::Conv_995[FLOAT, 512x512x3x3]
%onnx::Conv_998[FLOAT, 512x512x3x3]
%onnx::Conv_1001[FLOAT, 512x512x3x3]
%onnx::Conv_1004[FLOAT, 512x512x1x1]
%onnx::Conv_1007[FLOAT, 512x512x1x1]
%onnx::Conv_1010[FLOAT, 512x512x1x1]
%onnx::Conv_1013[FLOAT, 512x512x3x3]
%onnx::Conv_1016[FLOAT, 512x512x3x3]
%onnx::Conv_1019[FLOAT, 512x512x3x3]
%onnx::Conv_1022[FLOAT, 512x512x1x1]
) {
%onnx::Conv_1023 = Identity(%onnx::Conv_972)
%onnx::Conv_1020 = Identity(%onnx::Conv_972)
%onnx::Conv_1017 = Identity(%onnx::Conv_972)
%onnx::Conv_1014 = Identity(%onnx::Conv_972)
%onnx::Conv_1011 = Identity(%onnx::Conv_972)
%onnx::Conv_1008 = Identity(%onnx::Conv_972)
%onnx::Conv_1005 = Identity(%onnx::Conv_972)
%onnx::Conv_1002 = Identity(%onnx::Conv_972)
%onnx::Conv_999 = Identity(%onnx::Conv_972)
%onnx::Conv_996 = Identity(%onnx::Conv_972)
%onnx::Conv_993 = Identity(%onnx::Conv_972)
%onnx::Conv_990 = Identity(%onnx::Conv_972)
%onnx::Conv_987 = Identity(%onnx::Conv_972)
%onnx::Conv_984 = Identity(%onnx::Conv_972)
%onnx::Conv_981 = Identity(%onnx::Conv_972)
%onnx::Conv_978 = Identity(%onnx::Conv_972)
%onnx::Conv_975 = Identity(%onnx::Conv_972)
%onnx::Conv_969 = Identity(%onnx::Conv_918)
%onnx::Conv_966 = Identity(%onnx::Conv_918)
%onnx::Conv_963 = Identity(%onnx::Conv_918)
%onnx::Conv_960 = Identity(%onnx::Conv_918)
%onnx::Conv_957 = Identity(%onnx::Conv_918)
%onnx::Conv_954 = Identity(%onnx::Conv_918)
%onnx::Conv_951 = Identity(%onnx::Conv_918)
%onnx::Conv_948 = Identity(%onnx::Conv_918)
%onnx::Conv_945 = Identity(%onnx::Conv_918)
%onnx::Conv_942 = Identity(%onnx::Conv_918)
%onnx::Conv_939 = Identity(%onnx::Conv_918)
%onnx::Conv_936 = Identity(%onnx::Conv_918)
%onnx::Conv_933 = Identity(%onnx::Conv_918)
%onnx::Conv_930 = Identity(%onnx::Conv_918)
%onnx::Conv_927 = Identity(%onnx::Conv_918)
%onnx::Conv_924 = Identity(%onnx::Conv_918)
%onnx::Conv_921 = Identity(%onnx::Conv_918)
%onnx::Conv_915 = Identity(%onnx::Conv_861)
%onnx::Conv_912 = Identity(%onnx::Conv_861)
%onnx::Conv_909 = Identity(%onnx::Conv_861)
%onnx::Conv_906 = Identity(%onnx::Conv_861)
%onnx::Conv_903 = Identity(%onnx::Conv_861)
%onnx::Conv_900 = Identity(%onnx::Conv_861)
%onnx::Conv_897 = Identity(%onnx::Conv_861)
%onnx::Conv_894 = Identity(%onnx::Conv_861)
%onnx::Conv_891 = Identity(%onnx::Conv_861)
%onnx::Conv_888 = Identity(%onnx::Conv_861)
%onnx::Conv_885 = Identity(%onnx::Conv_861)
%onnx::Conv_882 = Identity(%onnx::Conv_861)
%onnx::Conv_879 = Identity(%onnx::Conv_861)
%onnx::Conv_876 = Identity(%onnx::Conv_861)
%onnx::Conv_873 = Identity(%onnx::Conv_861)
%onnx::Conv_870 = Identity(%onnx::Conv_861)
%onnx::Conv_867 = Identity(%onnx::Conv_861)
%onnx::Conv_864 = Identity(%onnx::Conv_861)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_860, %onnx::Conv_861)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_5_output_0 = Add(%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_5_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_5_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_5_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_5_output_0 = Add(%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_5_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_5_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_908, %onnx::Conv_909)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_4_output_0, %onnx::Conv_911, %onnx::Conv_912)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_5_output_0, %onnx::Conv_914, %onnx::Conv_915)
%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_5_output_0 = Add(%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Add_5_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_917, %onnx::Conv_918)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_920, %onnx::Conv_921)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_923, %onnx::Conv_924)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_926, %onnx::Conv_927)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_929, %onnx::Conv_930)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_932, %onnx::Conv_933)
%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_5_output_0 = Add(%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_5_output_0, %onnx::Conv_935, %onnx::Conv_936)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_5_output_0, %onnx::Conv_938, %onnx::Conv_939)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_941, %onnx::Conv_942)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_944, %onnx::Conv_945)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_947, %onnx::Conv_948)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_5_output_0, %onnx::Conv_950, %onnx::Conv_951)
%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_5_output_0 = Add(%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_5_output_0, %onnx::Conv_953, %onnx::Conv_954)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_5_output_0, %onnx::Conv_956, %onnx::Conv_957)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_959, %onnx::Conv_960)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_962, %onnx::Conv_963)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_4_output_0, %onnx::Conv_965, %onnx::Conv_966)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_5_output_0, %onnx::Conv_968, %onnx::Conv_969)
%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_5_output_0 = Add(%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Add_5_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_974, %onnx::Conv_975)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_977, %onnx::Conv_978)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_5_output_0 = Add(%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_5_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_5_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_5_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_5_output_0 = Add(%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_5_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_5_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_4_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_5_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_5_output_0 = Add(%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Add_5_output_0)
%858 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %858
}
|
val_accuracy
| 94.120592
| 9,000,200,192
| 30,515,338
|
{'zcp_epe_nas': 74.56114462016662, 'zcp_fisher': 24.5480899810791, 'zcp_flops': 144003203072.0, 'zcp_grad_norm': 90.43429565429688, 'zcp_grasp': -1.25927734375, 'zcp_jacov': -16.056420470061962, 'zcp_l2_norm': 1226.35009765625, 'zcp_nwot': 234.39776099140403, 'zcp_params': 30515338.0, 'zcp_plain': 0.013440074399113001, 'zcp_snip': 837.2932739257812, 'zcp_synflow': 145.61014679690373, 'zcp_zen': 130.31919860839844, 'zcp_val_accuracy': 0.8969351053237911}
| |
NASBench101_367486
|
NASBench101
|
367486
|
de2ff9d9a74ea0b03c8b027f6e4d34e3
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_554[FLOAT, 128x3x3x3]
%onnx::Conv_555[FLOAT, 128]
%onnx::Conv_557[FLOAT, 64x128x1x1]
%onnx::Conv_558[FLOAT, 64]
%onnx::Conv_560[FLOAT, 64x64x1x1]
%onnx::Conv_563[FLOAT, 64x64x3x3]
%onnx::Conv_566[FLOAT, 64x128x1x1]
%onnx::Conv_569[FLOAT, 64x64x1x1]
%onnx::Conv_572[FLOAT, 64x64x3x3]
%onnx::Conv_575[FLOAT, 64x128x1x1]
%onnx::Conv_578[FLOAT, 64x64x1x1]
%onnx::Conv_581[FLOAT, 64x64x3x3]
%onnx::Conv_584[FLOAT, 128x128x1x1]
%onnx::Conv_587[FLOAT, 128x128x1x1]
%onnx::Conv_590[FLOAT, 128x128x3x3]
%onnx::Conv_593[FLOAT, 128x256x1x1]
%onnx::Conv_596[FLOAT, 128x128x1x1]
%onnx::Conv_599[FLOAT, 128x128x3x3]
%onnx::Conv_602[FLOAT, 128x256x1x1]
%onnx::Conv_605[FLOAT, 128x128x1x1]
%onnx::Conv_608[FLOAT, 128x128x3x3]
%onnx::Conv_611[FLOAT, 256x256x1x1]
%onnx::Conv_612[FLOAT, 256]
%onnx::Conv_614[FLOAT, 256x256x1x1]
%onnx::Conv_617[FLOAT, 256x256x3x3]
%onnx::Conv_620[FLOAT, 256x512x1x1]
%onnx::Conv_623[FLOAT, 256x256x1x1]
%onnx::Conv_626[FLOAT, 256x256x3x3]
%onnx::Conv_629[FLOAT, 256x512x1x1]
%onnx::Conv_632[FLOAT, 256x256x1x1]
%onnx::Conv_635[FLOAT, 256x256x3x3]
) {
%onnx::Conv_636 = Identity(%onnx::Conv_612)
%onnx::Conv_633 = Identity(%onnx::Conv_612)
%onnx::Conv_630 = Identity(%onnx::Conv_612)
%onnx::Conv_627 = Identity(%onnx::Conv_612)
%onnx::Conv_624 = Identity(%onnx::Conv_612)
%onnx::Conv_621 = Identity(%onnx::Conv_612)
%onnx::Conv_618 = Identity(%onnx::Conv_612)
%onnx::Conv_615 = Identity(%onnx::Conv_612)
%onnx::Conv_609 = Identity(%onnx::Conv_555)
%onnx::Conv_606 = Identity(%onnx::Conv_555)
%onnx::Conv_603 = Identity(%onnx::Conv_555)
%onnx::Conv_600 = Identity(%onnx::Conv_555)
%onnx::Conv_597 = Identity(%onnx::Conv_555)
%onnx::Conv_594 = Identity(%onnx::Conv_555)
%onnx::Conv_591 = Identity(%onnx::Conv_555)
%onnx::Conv_588 = Identity(%onnx::Conv_555)
%onnx::Conv_585 = Identity(%onnx::Conv_555)
%onnx::Conv_582 = Identity(%onnx::Conv_558)
%onnx::Conv_579 = Identity(%onnx::Conv_558)
%onnx::Conv_576 = Identity(%onnx::Conv_558)
%onnx::Conv_573 = Identity(%onnx::Conv_558)
%onnx::Conv_570 = Identity(%onnx::Conv_558)
%onnx::Conv_567 = Identity(%onnx::Conv_558)
%onnx::Conv_564 = Identity(%onnx::Conv_558)
%onnx::Conv_561 = Identity(%onnx::Conv_558)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_554, %onnx::Conv_555)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_557, %onnx::Conv_558)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_560, %onnx::Conv_561)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_563, %onnx::Conv_564)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0)
%/layers.1/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_566, %onnx::Conv_567)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_569, %onnx::Conv_570)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_572, %onnx::Conv_573)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0)
%/layers.2/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_575, %onnx::Conv_576)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_578, %onnx::Conv_579)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_581, %onnx::Conv_582)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0)
%/layers.3/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_584, %onnx::Conv_585)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_587, %onnx::Conv_588)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_590, %onnx::Conv_591)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0)
%/layers.5/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_593, %onnx::Conv_594)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_596, %onnx::Conv_597)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_599, %onnx::Conv_600)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0)
%/layers.6/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_602, %onnx::Conv_603)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_605, %onnx::Conv_606)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_608, %onnx::Conv_609)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0)
%/layers.7/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_611, %onnx::Conv_612)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_614, %onnx::Conv_615)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_617, %onnx::Conv_618)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0)
%/layers.9/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_620, %onnx::Conv_621)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_623, %onnx::Conv_624)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_626, %onnx::Conv_627)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0)
%/layers.10/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_629, %onnx::Conv_630)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_632, %onnx::Conv_633)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_635, %onnx::Conv_636)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0)
%/layers.11/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.5/maxpool/MaxPool_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%552 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %552
}
|
val_accuracy
| 87.620193
| 905,586,688
| 3,031,562
|
{'zcp_epe_nas': 68.52487264969923, 'zcp_fisher': 373.6503601074219, 'zcp_flops': 14489387008.0, 'zcp_grad_norm': 262.9383239746094, 'zcp_grasp': 18.5673828125, 'zcp_jacov': -16.063190046875512, 'zcp_l2_norm': 498.2822570800781, 'zcp_nwot': 214.18104778607037, 'zcp_params': 3031562.0, 'zcp_plain': 0.08527436107397, 'zcp_snip': 1428.4779052734375, 'zcp_synflow': 90.36559756219499, 'zcp_zen': 52.9733772277832, 'zcp_val_accuracy': 0.9096554517745971}
| |
NASBench101_143759
|
NASBench101
|
143759
|
56f72f657ad46091cdf5e6e8e1d65ca3
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_968[FLOAT, 128x3x3x3]
%onnx::Conv_969[FLOAT, 128]
%onnx::Conv_971[FLOAT, 64x128x1x1]
%onnx::Conv_972[FLOAT, 64]
%onnx::Conv_974[FLOAT, 64x64x3x3]
%onnx::Conv_977[FLOAT, 64x64x1x1]
%onnx::Conv_980[FLOAT, 64x128x1x1]
%onnx::Conv_983[FLOAT, 64x64x1x1]
%onnx::Conv_986[FLOAT, 64x64x1x1]
%onnx::Conv_989[FLOAT, 128x128x1x1]
%onnx::Conv_992[FLOAT, 64x128x1x1]
%onnx::Conv_995[FLOAT, 64x64x3x3]
%onnx::Conv_998[FLOAT, 64x64x1x1]
%onnx::Conv_1001[FLOAT, 64x128x1x1]
%onnx::Conv_1004[FLOAT, 64x64x1x1]
%onnx::Conv_1007[FLOAT, 64x64x1x1]
%onnx::Conv_1010[FLOAT, 128x128x1x1]
%onnx::Conv_1013[FLOAT, 64x128x1x1]
%onnx::Conv_1016[FLOAT, 64x64x3x3]
%onnx::Conv_1019[FLOAT, 64x64x1x1]
%onnx::Conv_1022[FLOAT, 64x128x1x1]
%onnx::Conv_1025[FLOAT, 64x64x1x1]
%onnx::Conv_1028[FLOAT, 64x64x1x1]
%onnx::Conv_1031[FLOAT, 128x128x1x1]
%onnx::Conv_1034[FLOAT, 128x128x1x1]
%onnx::Conv_1037[FLOAT, 128x128x3x3]
%onnx::Conv_1040[FLOAT, 128x128x1x1]
%onnx::Conv_1043[FLOAT, 128x128x1x1]
%onnx::Conv_1046[FLOAT, 128x128x1x1]
%onnx::Conv_1049[FLOAT, 128x128x1x1]
%onnx::Conv_1052[FLOAT, 256x128x1x1]
%onnx::Conv_1053[FLOAT, 256]
%onnx::Conv_1055[FLOAT, 128x256x1x1]
%onnx::Conv_1058[FLOAT, 128x128x3x3]
%onnx::Conv_1061[FLOAT, 128x128x1x1]
%onnx::Conv_1064[FLOAT, 128x256x1x1]
%onnx::Conv_1067[FLOAT, 128x128x1x1]
%onnx::Conv_1070[FLOAT, 128x128x1x1]
%onnx::Conv_1073[FLOAT, 256x256x1x1]
%onnx::Conv_1076[FLOAT, 128x256x1x1]
%onnx::Conv_1079[FLOAT, 128x128x3x3]
%onnx::Conv_1082[FLOAT, 128x128x1x1]
%onnx::Conv_1085[FLOAT, 128x256x1x1]
%onnx::Conv_1088[FLOAT, 128x128x1x1]
%onnx::Conv_1091[FLOAT, 128x128x1x1]
%onnx::Conv_1094[FLOAT, 256x256x1x1]
%onnx::Conv_1097[FLOAT, 256x256x1x1]
%onnx::Conv_1100[FLOAT, 256x256x3x3]
%onnx::Conv_1103[FLOAT, 256x256x1x1]
%onnx::Conv_1106[FLOAT, 256x256x1x1]
%onnx::Conv_1109[FLOAT, 256x256x1x1]
%onnx::Conv_1112[FLOAT, 256x256x1x1]
%onnx::Conv_1115[FLOAT, 512x256x1x1]
%onnx::Conv_1116[FLOAT, 512]
%onnx::Conv_1118[FLOAT, 256x512x1x1]
%onnx::Conv_1121[FLOAT, 256x256x3x3]
%onnx::Conv_1124[FLOAT, 256x256x1x1]
%onnx::Conv_1127[FLOAT, 256x512x1x1]
%onnx::Conv_1130[FLOAT, 256x256x1x1]
%onnx::Conv_1133[FLOAT, 256x256x1x1]
%onnx::Conv_1136[FLOAT, 512x512x1x1]
%onnx::Conv_1139[FLOAT, 256x512x1x1]
%onnx::Conv_1142[FLOAT, 256x256x3x3]
%onnx::Conv_1145[FLOAT, 256x256x1x1]
%onnx::Conv_1148[FLOAT, 256x512x1x1]
%onnx::Conv_1151[FLOAT, 256x256x1x1]
%onnx::Conv_1154[FLOAT, 256x256x1x1]
%onnx::Conv_1157[FLOAT, 512x512x1x1]
) {
%onnx::Conv_1158 = Identity(%onnx::Conv_1116)
%onnx::Conv_1155 = Identity(%onnx::Conv_1053)
%onnx::Conv_1152 = Identity(%onnx::Conv_1053)
%onnx::Conv_1149 = Identity(%onnx::Conv_1053)
%onnx::Conv_1146 = Identity(%onnx::Conv_1053)
%onnx::Conv_1143 = Identity(%onnx::Conv_1053)
%onnx::Conv_1140 = Identity(%onnx::Conv_1053)
%onnx::Conv_1137 = Identity(%onnx::Conv_1116)
%onnx::Conv_1134 = Identity(%onnx::Conv_1053)
%onnx::Conv_1131 = Identity(%onnx::Conv_1053)
%onnx::Conv_1128 = Identity(%onnx::Conv_1053)
%onnx::Conv_1125 = Identity(%onnx::Conv_1053)
%onnx::Conv_1122 = Identity(%onnx::Conv_1053)
%onnx::Conv_1119 = Identity(%onnx::Conv_1053)
%onnx::Conv_1113 = Identity(%onnx::Conv_1053)
%onnx::Conv_1110 = Identity(%onnx::Conv_1053)
%onnx::Conv_1107 = Identity(%onnx::Conv_1053)
%onnx::Conv_1104 = Identity(%onnx::Conv_1053)
%onnx::Conv_1101 = Identity(%onnx::Conv_1053)
%onnx::Conv_1098 = Identity(%onnx::Conv_1053)
%onnx::Conv_1095 = Identity(%onnx::Conv_1053)
%onnx::Conv_1092 = Identity(%onnx::Conv_969)
%onnx::Conv_1089 = Identity(%onnx::Conv_969)
%onnx::Conv_1086 = Identity(%onnx::Conv_969)
%onnx::Conv_1083 = Identity(%onnx::Conv_969)
%onnx::Conv_1080 = Identity(%onnx::Conv_969)
%onnx::Conv_1077 = Identity(%onnx::Conv_969)
%onnx::Conv_1074 = Identity(%onnx::Conv_1053)
%onnx::Conv_1071 = Identity(%onnx::Conv_969)
%onnx::Conv_1068 = Identity(%onnx::Conv_969)
%onnx::Conv_1065 = Identity(%onnx::Conv_969)
%onnx::Conv_1062 = Identity(%onnx::Conv_969)
%onnx::Conv_1059 = Identity(%onnx::Conv_969)
%onnx::Conv_1056 = Identity(%onnx::Conv_969)
%onnx::Conv_1050 = Identity(%onnx::Conv_969)
%onnx::Conv_1047 = Identity(%onnx::Conv_969)
%onnx::Conv_1044 = Identity(%onnx::Conv_969)
%onnx::Conv_1041 = Identity(%onnx::Conv_969)
%onnx::Conv_1038 = Identity(%onnx::Conv_969)
%onnx::Conv_1035 = Identity(%onnx::Conv_969)
%onnx::Conv_1032 = Identity(%onnx::Conv_969)
%onnx::Conv_1029 = Identity(%onnx::Conv_972)
%onnx::Conv_1026 = Identity(%onnx::Conv_972)
%onnx::Conv_1023 = Identity(%onnx::Conv_972)
%onnx::Conv_1020 = Identity(%onnx::Conv_972)
%onnx::Conv_1017 = Identity(%onnx::Conv_972)
%onnx::Conv_1014 = Identity(%onnx::Conv_972)
%onnx::Conv_1011 = Identity(%onnx::Conv_969)
%onnx::Conv_1008 = Identity(%onnx::Conv_972)
%onnx::Conv_1005 = Identity(%onnx::Conv_972)
%onnx::Conv_1002 = Identity(%onnx::Conv_972)
%onnx::Conv_999 = Identity(%onnx::Conv_972)
%onnx::Conv_996 = Identity(%onnx::Conv_972)
%onnx::Conv_993 = Identity(%onnx::Conv_972)
%onnx::Conv_990 = Identity(%onnx::Conv_969)
%onnx::Conv_987 = Identity(%onnx::Conv_972)
%onnx::Conv_984 = Identity(%onnx::Conv_972)
%onnx::Conv_981 = Identity(%onnx::Conv_972)
%onnx::Conv_978 = Identity(%onnx::Conv_972)
%onnx::Conv_975 = Identity(%onnx::Conv_972)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_968, %onnx::Conv_969)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_974, %onnx::Conv_975)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_977, %onnx::Conv_978)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/Add_3_output_0, %/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_5_output_0 = Add(%/layers.1/Concat_output_0, %/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_5_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_5_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/Add_3_output_0, %/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_5_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_5_output_0 = Add(%/layers.2/Concat_output_0, %/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_5_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_5_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_1025, %onnx::Conv_1026)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/Add_3_output_0, %/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_4_output_0, %onnx::Conv_1028, %onnx::Conv_1029)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_5_output_0, %onnx::Conv_1031, %onnx::Conv_1032)
%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_5_output_0 = Add(%/layers.3/Concat_output_0, %/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Add_5_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1034, %onnx::Conv_1035)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_1037, %onnx::Conv_1038)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_1040, %onnx::Conv_1041)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1043, %onnx::Conv_1044)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_1046, %onnx::Conv_1047)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/Add_3_output_0, %/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_1049, %onnx::Conv_1050)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1052, %onnx::Conv_1053)
%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_5_output_0 = Add(%/layers.5/Concat_output_0, %/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_5_output_0, %onnx::Conv_1055, %onnx::Conv_1056)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_1058, %onnx::Conv_1059)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_1061, %onnx::Conv_1062)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_5_output_0, %onnx::Conv_1064, %onnx::Conv_1065)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_1067, %onnx::Conv_1068)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/Add_3_output_0, %/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_1070, %onnx::Conv_1071)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_5_output_0, %onnx::Conv_1073, %onnx::Conv_1074)
%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_5_output_0 = Add(%/layers.6/Concat_output_0, %/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_5_output_0, %onnx::Conv_1076, %onnx::Conv_1077)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_1079, %onnx::Conv_1080)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_1082, %onnx::Conv_1083)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_5_output_0, %onnx::Conv_1085, %onnx::Conv_1086)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_1088, %onnx::Conv_1089)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/Add_3_output_0, %/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_4_output_0, %onnx::Conv_1091, %onnx::Conv_1092)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_5_output_0, %onnx::Conv_1094, %onnx::Conv_1095)
%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_5_output_0 = Add(%/layers.7/Concat_output_0, %/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Add_5_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1097, %onnx::Conv_1098)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_1100, %onnx::Conv_1101)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_1103, %onnx::Conv_1104)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1106, %onnx::Conv_1107)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_1109, %onnx::Conv_1110)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/Add_3_output_0, %/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_1112, %onnx::Conv_1113)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1115, %onnx::Conv_1116)
%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_5_output_0 = Add(%/layers.9/Concat_output_0, %/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_5_output_0, %onnx::Conv_1118, %onnx::Conv_1119)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_1121, %onnx::Conv_1122)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_1124, %onnx::Conv_1125)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_5_output_0, %onnx::Conv_1127, %onnx::Conv_1128)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_1130, %onnx::Conv_1131)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/Add_3_output_0, %/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_1133, %onnx::Conv_1134)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_5_output_0, %onnx::Conv_1136, %onnx::Conv_1137)
%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_5_output_0 = Add(%/layers.10/Concat_output_0, %/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_5_output_0, %onnx::Conv_1139, %onnx::Conv_1140)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_1142, %onnx::Conv_1143)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_1145, %onnx::Conv_1146)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_5_output_0, %onnx::Conv_1148, %onnx::Conv_1149)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_1151, %onnx::Conv_1152)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/Add_3_output_0, %/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_4_output_0, %onnx::Conv_1154, %onnx::Conv_1155)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_5_output_0, %onnx::Conv_1157, %onnx::Conv_1158)
%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_5_output_0 = Add(%/layers.11/Concat_output_0, %/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Add_5_output_0)
%966 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %966
}
|
val_accuracy
| 92.738384
| 1,472,997,376
| 4,863,626
|
{'zcp_epe_nas': 101.59218193492121, 'zcp_fisher': 10.162630081176758, 'zcp_flops': 23567958016.0, 'zcp_grad_norm': 66.76217651367188, 'zcp_grasp': -7.981903076171875, 'zcp_jacov': -16.053870565326555, 'zcp_l2_norm': 1189.9390869140625, 'zcp_nwot': 228.74931584813473, 'zcp_params': 4863626.0, 'zcp_plain': 0.17084254324436102, 'zcp_snip': 434.3724365234375, 'zcp_synflow': 107.75858438148785, 'zcp_zen': 107.04884338378906, 'zcp_val_accuracy': 0.9239783883094781}
| |
NASBench101_201339
|
NASBench101
|
201339
|
79e62073c60ab73a4927fe11e2b5fbd2
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_986[FLOAT, 128x3x3x3]
%onnx::Conv_987[FLOAT, 128]
%onnx::Conv_989[FLOAT, 64x128x1x1]
%onnx::Conv_990[FLOAT, 64]
%onnx::Conv_992[FLOAT, 64x64x1x1]
%onnx::Conv_995[FLOAT, 64x128x1x1]
%onnx::Conv_998[FLOAT, 64x64x3x3]
%onnx::Conv_1001[FLOAT, 64x64x1x1]
%onnx::Conv_1004[FLOAT, 64x64x1x1]
%onnx::Conv_1007[FLOAT, 128x128x1x1]
%onnx::Conv_1010[FLOAT, 64x128x1x1]
%onnx::Conv_1013[FLOAT, 64x64x1x1]
%onnx::Conv_1016[FLOAT, 64x128x1x1]
%onnx::Conv_1019[FLOAT, 64x64x3x3]
%onnx::Conv_1022[FLOAT, 64x64x1x1]
%onnx::Conv_1025[FLOAT, 64x64x1x1]
%onnx::Conv_1028[FLOAT, 128x128x1x1]
%onnx::Conv_1031[FLOAT, 64x128x1x1]
%onnx::Conv_1034[FLOAT, 64x64x1x1]
%onnx::Conv_1037[FLOAT, 64x128x1x1]
%onnx::Conv_1040[FLOAT, 64x64x3x3]
%onnx::Conv_1043[FLOAT, 64x64x1x1]
%onnx::Conv_1046[FLOAT, 64x64x1x1]
%onnx::Conv_1049[FLOAT, 128x128x1x1]
%onnx::Conv_1052[FLOAT, 128x128x1x1]
%onnx::Conv_1055[FLOAT, 128x128x1x1]
%onnx::Conv_1058[FLOAT, 128x128x1x1]
%onnx::Conv_1061[FLOAT, 128x128x3x3]
%onnx::Conv_1064[FLOAT, 128x128x1x1]
%onnx::Conv_1067[FLOAT, 128x128x1x1]
%onnx::Conv_1070[FLOAT, 256x128x1x1]
%onnx::Conv_1071[FLOAT, 256]
%onnx::Conv_1073[FLOAT, 128x256x1x1]
%onnx::Conv_1076[FLOAT, 128x128x1x1]
%onnx::Conv_1079[FLOAT, 128x256x1x1]
%onnx::Conv_1082[FLOAT, 128x128x3x3]
%onnx::Conv_1085[FLOAT, 128x128x1x1]
%onnx::Conv_1088[FLOAT, 128x128x1x1]
%onnx::Conv_1091[FLOAT, 256x256x1x1]
%onnx::Conv_1094[FLOAT, 128x256x1x1]
%onnx::Conv_1097[FLOAT, 128x128x1x1]
%onnx::Conv_1100[FLOAT, 128x256x1x1]
%onnx::Conv_1103[FLOAT, 128x128x3x3]
%onnx::Conv_1106[FLOAT, 128x128x1x1]
%onnx::Conv_1109[FLOAT, 128x128x1x1]
%onnx::Conv_1112[FLOAT, 256x256x1x1]
%onnx::Conv_1115[FLOAT, 256x256x1x1]
%onnx::Conv_1118[FLOAT, 256x256x1x1]
%onnx::Conv_1121[FLOAT, 256x256x1x1]
%onnx::Conv_1124[FLOAT, 256x256x3x3]
%onnx::Conv_1127[FLOAT, 256x256x1x1]
%onnx::Conv_1130[FLOAT, 256x256x1x1]
%onnx::Conv_1133[FLOAT, 512x256x1x1]
%onnx::Conv_1134[FLOAT, 512]
%onnx::Conv_1136[FLOAT, 256x512x1x1]
%onnx::Conv_1139[FLOAT, 256x256x1x1]
%onnx::Conv_1142[FLOAT, 256x512x1x1]
%onnx::Conv_1145[FLOAT, 256x256x3x3]
%onnx::Conv_1148[FLOAT, 256x256x1x1]
%onnx::Conv_1151[FLOAT, 256x256x1x1]
%onnx::Conv_1154[FLOAT, 512x512x1x1]
%onnx::Conv_1157[FLOAT, 256x512x1x1]
%onnx::Conv_1160[FLOAT, 256x256x1x1]
%onnx::Conv_1163[FLOAT, 256x512x1x1]
%onnx::Conv_1166[FLOAT, 256x256x3x3]
%onnx::Conv_1169[FLOAT, 256x256x1x1]
%onnx::Conv_1172[FLOAT, 256x256x1x1]
%onnx::Conv_1175[FLOAT, 512x512x1x1]
) {
%onnx::Conv_1176 = Identity(%onnx::Conv_1134)
%onnx::Conv_1173 = Identity(%onnx::Conv_1071)
%onnx::Conv_1170 = Identity(%onnx::Conv_1071)
%onnx::Conv_1167 = Identity(%onnx::Conv_1071)
%onnx::Conv_1164 = Identity(%onnx::Conv_1071)
%onnx::Conv_1161 = Identity(%onnx::Conv_1071)
%onnx::Conv_1158 = Identity(%onnx::Conv_1071)
%onnx::Conv_1155 = Identity(%onnx::Conv_1134)
%onnx::Conv_1152 = Identity(%onnx::Conv_1071)
%onnx::Conv_1149 = Identity(%onnx::Conv_1071)
%onnx::Conv_1146 = Identity(%onnx::Conv_1071)
%onnx::Conv_1143 = Identity(%onnx::Conv_1071)
%onnx::Conv_1140 = Identity(%onnx::Conv_1071)
%onnx::Conv_1137 = Identity(%onnx::Conv_1071)
%onnx::Conv_1131 = Identity(%onnx::Conv_1071)
%onnx::Conv_1128 = Identity(%onnx::Conv_1071)
%onnx::Conv_1125 = Identity(%onnx::Conv_1071)
%onnx::Conv_1122 = Identity(%onnx::Conv_1071)
%onnx::Conv_1119 = Identity(%onnx::Conv_1071)
%onnx::Conv_1116 = Identity(%onnx::Conv_1071)
%onnx::Conv_1113 = Identity(%onnx::Conv_1071)
%onnx::Conv_1110 = Identity(%onnx::Conv_987)
%onnx::Conv_1107 = Identity(%onnx::Conv_987)
%onnx::Conv_1104 = Identity(%onnx::Conv_987)
%onnx::Conv_1101 = Identity(%onnx::Conv_987)
%onnx::Conv_1098 = Identity(%onnx::Conv_987)
%onnx::Conv_1095 = Identity(%onnx::Conv_987)
%onnx::Conv_1092 = Identity(%onnx::Conv_1071)
%onnx::Conv_1089 = Identity(%onnx::Conv_987)
%onnx::Conv_1086 = Identity(%onnx::Conv_987)
%onnx::Conv_1083 = Identity(%onnx::Conv_987)
%onnx::Conv_1080 = Identity(%onnx::Conv_987)
%onnx::Conv_1077 = Identity(%onnx::Conv_987)
%onnx::Conv_1074 = Identity(%onnx::Conv_987)
%onnx::Conv_1068 = Identity(%onnx::Conv_987)
%onnx::Conv_1065 = Identity(%onnx::Conv_987)
%onnx::Conv_1062 = Identity(%onnx::Conv_987)
%onnx::Conv_1059 = Identity(%onnx::Conv_987)
%onnx::Conv_1056 = Identity(%onnx::Conv_987)
%onnx::Conv_1053 = Identity(%onnx::Conv_987)
%onnx::Conv_1050 = Identity(%onnx::Conv_987)
%onnx::Conv_1047 = Identity(%onnx::Conv_990)
%onnx::Conv_1044 = Identity(%onnx::Conv_990)
%onnx::Conv_1041 = Identity(%onnx::Conv_990)
%onnx::Conv_1038 = Identity(%onnx::Conv_990)
%onnx::Conv_1035 = Identity(%onnx::Conv_990)
%onnx::Conv_1032 = Identity(%onnx::Conv_990)
%onnx::Conv_1029 = Identity(%onnx::Conv_987)
%onnx::Conv_1026 = Identity(%onnx::Conv_990)
%onnx::Conv_1023 = Identity(%onnx::Conv_990)
%onnx::Conv_1020 = Identity(%onnx::Conv_990)
%onnx::Conv_1017 = Identity(%onnx::Conv_990)
%onnx::Conv_1014 = Identity(%onnx::Conv_990)
%onnx::Conv_1011 = Identity(%onnx::Conv_990)
%onnx::Conv_1008 = Identity(%onnx::Conv_987)
%onnx::Conv_1005 = Identity(%onnx::Conv_990)
%onnx::Conv_1002 = Identity(%onnx::Conv_990)
%onnx::Conv_999 = Identity(%onnx::Conv_990)
%onnx::Conv_996 = Identity(%onnx::Conv_990)
%onnx::Conv_993 = Identity(%onnx::Conv_990)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_986, %onnx::Conv_987)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/Add_2_output_0, %/layers.1/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_5_output_0 = Add(%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_4_output_0)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_5_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_6_output_0 = Add(%/layers.1/Concat_output_0, %/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_6_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_6_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/Add_2_output_0, %/layers.2/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_5_output_0 = Add(%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_4_output_0)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_5_output_0, %onnx::Conv_1025, %onnx::Conv_1026)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_6_output_0, %onnx::Conv_1028, %onnx::Conv_1029)
%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_6_output_0 = Add(%/layers.2/Concat_output_0, %/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_6_output_0, %onnx::Conv_1031, %onnx::Conv_1032)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_1034, %onnx::Conv_1035)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_6_output_0, %onnx::Conv_1037, %onnx::Conv_1038)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/Add_2_output_0, %/layers.3/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_1040, %onnx::Conv_1041)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_4_output_0, %onnx::Conv_1043, %onnx::Conv_1044)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_5_output_0 = Add(%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_4_output_0)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_5_output_0, %onnx::Conv_1046, %onnx::Conv_1047)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_6_output_0, %onnx::Conv_1049, %onnx::Conv_1050)
%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_6_output_0 = Add(%/layers.3/Concat_output_0, %/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Add_6_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1052, %onnx::Conv_1053)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_1055, %onnx::Conv_1056)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1058, %onnx::Conv_1059)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/Add_2_output_0, %/layers.5/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_1061, %onnx::Conv_1062)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_1064, %onnx::Conv_1065)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_5_output_0 = Add(%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_4_output_0)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_5_output_0, %onnx::Conv_1067, %onnx::Conv_1068)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1070, %onnx::Conv_1071)
%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_6_output_0 = Add(%/layers.5/Concat_output_0, %/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_6_output_0, %onnx::Conv_1073, %onnx::Conv_1074)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_1076, %onnx::Conv_1077)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_6_output_0, %onnx::Conv_1079, %onnx::Conv_1080)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/Add_2_output_0, %/layers.6/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_1082, %onnx::Conv_1083)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_1085, %onnx::Conv_1086)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_5_output_0 = Add(%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_4_output_0)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_5_output_0, %onnx::Conv_1088, %onnx::Conv_1089)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_6_output_0, %onnx::Conv_1091, %onnx::Conv_1092)
%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_6_output_0 = Add(%/layers.6/Concat_output_0, %/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_6_output_0, %onnx::Conv_1094, %onnx::Conv_1095)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_1097, %onnx::Conv_1098)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_6_output_0, %onnx::Conv_1100, %onnx::Conv_1101)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/Add_2_output_0, %/layers.7/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_1103, %onnx::Conv_1104)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_4_output_0, %onnx::Conv_1106, %onnx::Conv_1107)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_5_output_0 = Add(%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_4_output_0)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_5_output_0, %onnx::Conv_1109, %onnx::Conv_1110)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_6_output_0, %onnx::Conv_1112, %onnx::Conv_1113)
%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_6_output_0 = Add(%/layers.7/Concat_output_0, %/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Add_6_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1115, %onnx::Conv_1116)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_1118, %onnx::Conv_1119)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1121, %onnx::Conv_1122)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/Add_2_output_0, %/layers.9/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_1124, %onnx::Conv_1125)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_1127, %onnx::Conv_1128)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_5_output_0 = Add(%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_4_output_0)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_5_output_0, %onnx::Conv_1130, %onnx::Conv_1131)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1133, %onnx::Conv_1134)
%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_6_output_0 = Add(%/layers.9/Concat_output_0, %/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_6_output_0, %onnx::Conv_1136, %onnx::Conv_1137)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_1139, %onnx::Conv_1140)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_6_output_0, %onnx::Conv_1142, %onnx::Conv_1143)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/Add_2_output_0, %/layers.10/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_1145, %onnx::Conv_1146)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_1148, %onnx::Conv_1149)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_5_output_0 = Add(%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_4_output_0)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_5_output_0, %onnx::Conv_1151, %onnx::Conv_1152)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_6_output_0, %onnx::Conv_1154, %onnx::Conv_1155)
%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_6_output_0 = Add(%/layers.10/Concat_output_0, %/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_6_output_0, %onnx::Conv_1157, %onnx::Conv_1158)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_1160, %onnx::Conv_1161)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_6_output_0, %onnx::Conv_1163, %onnx::Conv_1164)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/Add_2_output_0, %/layers.11/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_1166, %onnx::Conv_1167)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_4_output_0, %onnx::Conv_1169, %onnx::Conv_1170)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_5_output_0 = Add(%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_4_output_0)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_5_output_0, %onnx::Conv_1172, %onnx::Conv_1173)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_6_output_0, %onnx::Conv_1175, %onnx::Conv_1176)
%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_6_output_0 = Add(%/layers.11/Concat_output_0, %/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Add_6_output_0)
%984 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %984
}
|
val_accuracy
| 93.189102
| 1,472,997,376
| 4,863,626
|
{'zcp_epe_nas': 83.41516271686173, 'zcp_fisher': 13.879354476928711, 'zcp_flops': 23567958016.0, 'zcp_grad_norm': 73.52190399169922, 'zcp_grasp': -1.36566162109375, 'zcp_jacov': -16.047145124897835, 'zcp_l2_norm': 1190.8138427734375, 'zcp_nwot': 228.61735728668592, 'zcp_params': 4863626.0, 'zcp_plain': 0.021755265071988, 'zcp_snip': 443.58660888671875, 'zcp_synflow': 131.48775557419054, 'zcp_zen': 106.26858520507812, 'zcp_val_accuracy': 0.9183694124221801}
| |
NASBench101_196715
|
NASBench101
|
196715
|
7709e51cb493cbbfe5d19f08ff48f8d6
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_788[FLOAT, 128x3x3x3]
%onnx::Conv_789[FLOAT, 128]
%onnx::Conv_791[FLOAT, 64x128x1x1]
%onnx::Conv_792[FLOAT, 64]
%onnx::Conv_794[FLOAT, 64x128x1x1]
%onnx::Conv_797[FLOAT, 64x64x3x3]
%onnx::Conv_800[FLOAT, 64x64x3x3]
%onnx::Conv_803[FLOAT, 64x128x1x1]
%onnx::Conv_806[FLOAT, 64x128x1x1]
%onnx::Conv_809[FLOAT, 64x128x1x1]
%onnx::Conv_812[FLOAT, 64x64x3x3]
%onnx::Conv_815[FLOAT, 64x64x3x3]
%onnx::Conv_818[FLOAT, 64x128x1x1]
%onnx::Conv_821[FLOAT, 64x128x1x1]
%onnx::Conv_824[FLOAT, 64x128x1x1]
%onnx::Conv_827[FLOAT, 64x64x3x3]
%onnx::Conv_830[FLOAT, 64x64x3x3]
%onnx::Conv_833[FLOAT, 64x128x1x1]
%onnx::Conv_836[FLOAT, 128x128x1x1]
%onnx::Conv_839[FLOAT, 128x128x1x1]
%onnx::Conv_842[FLOAT, 128x128x3x3]
%onnx::Conv_845[FLOAT, 128x128x3x3]
%onnx::Conv_848[FLOAT, 128x128x1x1]
%onnx::Conv_851[FLOAT, 128x256x1x1]
%onnx::Conv_854[FLOAT, 128x256x1x1]
%onnx::Conv_857[FLOAT, 128x128x3x3]
%onnx::Conv_860[FLOAT, 128x128x3x3]
%onnx::Conv_863[FLOAT, 128x256x1x1]
%onnx::Conv_866[FLOAT, 128x256x1x1]
%onnx::Conv_869[FLOAT, 128x256x1x1]
%onnx::Conv_872[FLOAT, 128x128x3x3]
%onnx::Conv_875[FLOAT, 128x128x3x3]
%onnx::Conv_878[FLOAT, 128x256x1x1]
%onnx::Conv_881[FLOAT, 256x256x1x1]
%onnx::Conv_882[FLOAT, 256]
%onnx::Conv_884[FLOAT, 256x256x1x1]
%onnx::Conv_887[FLOAT, 256x256x3x3]
%onnx::Conv_890[FLOAT, 256x256x3x3]
%onnx::Conv_893[FLOAT, 256x256x1x1]
%onnx::Conv_896[FLOAT, 256x512x1x1]
%onnx::Conv_899[FLOAT, 256x512x1x1]
%onnx::Conv_902[FLOAT, 256x256x3x3]
%onnx::Conv_905[FLOAT, 256x256x3x3]
%onnx::Conv_908[FLOAT, 256x512x1x1]
%onnx::Conv_911[FLOAT, 256x512x1x1]
%onnx::Conv_914[FLOAT, 256x512x1x1]
%onnx::Conv_917[FLOAT, 256x256x3x3]
%onnx::Conv_920[FLOAT, 256x256x3x3]
%onnx::Conv_923[FLOAT, 256x512x1x1]
) {
%onnx::Conv_924 = Identity(%onnx::Conv_882)
%onnx::Conv_921 = Identity(%onnx::Conv_882)
%onnx::Conv_918 = Identity(%onnx::Conv_882)
%onnx::Conv_915 = Identity(%onnx::Conv_882)
%onnx::Conv_912 = Identity(%onnx::Conv_882)
%onnx::Conv_909 = Identity(%onnx::Conv_882)
%onnx::Conv_906 = Identity(%onnx::Conv_882)
%onnx::Conv_903 = Identity(%onnx::Conv_882)
%onnx::Conv_900 = Identity(%onnx::Conv_882)
%onnx::Conv_897 = Identity(%onnx::Conv_882)
%onnx::Conv_894 = Identity(%onnx::Conv_882)
%onnx::Conv_891 = Identity(%onnx::Conv_882)
%onnx::Conv_888 = Identity(%onnx::Conv_882)
%onnx::Conv_885 = Identity(%onnx::Conv_882)
%onnx::Conv_879 = Identity(%onnx::Conv_789)
%onnx::Conv_876 = Identity(%onnx::Conv_789)
%onnx::Conv_873 = Identity(%onnx::Conv_789)
%onnx::Conv_870 = Identity(%onnx::Conv_789)
%onnx::Conv_867 = Identity(%onnx::Conv_789)
%onnx::Conv_864 = Identity(%onnx::Conv_789)
%onnx::Conv_861 = Identity(%onnx::Conv_789)
%onnx::Conv_858 = Identity(%onnx::Conv_789)
%onnx::Conv_855 = Identity(%onnx::Conv_789)
%onnx::Conv_852 = Identity(%onnx::Conv_789)
%onnx::Conv_849 = Identity(%onnx::Conv_789)
%onnx::Conv_846 = Identity(%onnx::Conv_789)
%onnx::Conv_843 = Identity(%onnx::Conv_789)
%onnx::Conv_840 = Identity(%onnx::Conv_789)
%onnx::Conv_837 = Identity(%onnx::Conv_789)
%onnx::Conv_834 = Identity(%onnx::Conv_792)
%onnx::Conv_831 = Identity(%onnx::Conv_792)
%onnx::Conv_828 = Identity(%onnx::Conv_792)
%onnx::Conv_825 = Identity(%onnx::Conv_792)
%onnx::Conv_822 = Identity(%onnx::Conv_792)
%onnx::Conv_819 = Identity(%onnx::Conv_792)
%onnx::Conv_816 = Identity(%onnx::Conv_792)
%onnx::Conv_813 = Identity(%onnx::Conv_792)
%onnx::Conv_810 = Identity(%onnx::Conv_792)
%onnx::Conv_807 = Identity(%onnx::Conv_792)
%onnx::Conv_804 = Identity(%onnx::Conv_792)
%onnx::Conv_801 = Identity(%onnx::Conv_792)
%onnx::Conv_798 = Identity(%onnx::Conv_792)
%onnx::Conv_795 = Identity(%onnx::Conv_792)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_788, %onnx::Conv_789)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_797, %onnx::Conv_798)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_800, %onnx::Conv_801)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0)
%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_803, %onnx::Conv_804)
%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/Add_3_output_0, %/layers.1/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.1/Add_5_output_0 = Add(%/layers.1/Add_4_output_0, %/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_5_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_806, %onnx::Conv_807)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_809, %onnx::Conv_810)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_812, %onnx::Conv_813)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_815, %onnx::Conv_816)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0)
%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_818, %onnx::Conv_819)
%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/Add_3_output_0, %/layers.2/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.2/Add_5_output_0 = Add(%/layers.2/Add_4_output_0, %/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_5_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_821, %onnx::Conv_822)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_824, %onnx::Conv_825)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_827, %onnx::Conv_828)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_830, %onnx::Conv_831)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0)
%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_833, %onnx::Conv_834)
%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/Add_3_output_0, %/layers.3/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.3/Add_5_output_0 = Add(%/layers.3/Add_4_output_0, %/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_5_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_836, %onnx::Conv_837)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_839, %onnx::Conv_840)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_842, %onnx::Conv_843)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_845, %onnx::Conv_846)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0)
%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_848, %onnx::Conv_849)
%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/Add_3_output_0, %/layers.5/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.5/Add_5_output_0 = Add(%/layers.5/Add_4_output_0, %/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_5_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_851, %onnx::Conv_852)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_854, %onnx::Conv_855)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_857, %onnx::Conv_858)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0)
%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/Add_3_output_0, %/layers.6/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.6/Add_5_output_0 = Add(%/layers.6/Add_4_output_0, %/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_5_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0)
%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/Add_3_output_0, %/layers.7/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.7/Add_5_output_0 = Add(%/layers.7/Add_4_output_0, %/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_5_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0)
%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/Add_3_output_0, %/layers.9/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.9/Add_5_output_0 = Add(%/layers.9/Add_4_output_0, %/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_5_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0)
%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_908, %onnx::Conv_909)
%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/Add_3_output_0, %/layers.10/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.10/Add_5_output_0 = Add(%/layers.10/Add_4_output_0, %/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_5_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_911, %onnx::Conv_912)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_914, %onnx::Conv_915)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_917, %onnx::Conv_918)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_920, %onnx::Conv_921)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0)
%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_923, %onnx::Conv_924)
%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/Add_3_output_0, %/layers.11/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.11/Add_5_output_0 = Add(%/layers.11/Add_4_output_0, %/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_5_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/vertex_op.5/maxpool/MaxPool_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%786 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %786
}
|
val_accuracy
| 91.606569
| 1,783,506,944
| 5,969,674
|
{'zcp_epe_nas': 146.4966019425841, 'zcp_fisher': 10.748007774353027, 'zcp_flops': 28536111104.0, 'zcp_grad_norm': 74.18268585205078, 'zcp_grasp': -31.65936279296875, 'zcp_jacov': -16.061239819856883, 'zcp_l2_norm': 889.8037719726562, 'zcp_nwot': 221.41377550476196, 'zcp_params': 5969674.0, 'zcp_plain': 0.38364890217781, 'zcp_snip': 463.39459228515625, 'zcp_synflow': 67.39827269042195, 'zcp_zen': 96.12655639648438, 'zcp_val_accuracy': 0.91796875}
| |
NASBench101_45360
|
NASBench101
|
45360
|
1b86f6d51236f0fe9434338559003f0d
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_662[FLOAT, 128x3x3x3]
%onnx::Conv_663[FLOAT, 128]
%onnx::Conv_665[FLOAT, 64x128x1x1]
%onnx::Conv_666[FLOAT, 64]
%onnx::Conv_668[FLOAT, 64x128x1x1]
%onnx::Conv_671[FLOAT, 64x128x1x1]
%onnx::Conv_674[FLOAT, 64x64x3x3]
%onnx::Conv_677[FLOAT, 64x128x1x1]
%onnx::Conv_680[FLOAT, 64x128x1x1]
%onnx::Conv_683[FLOAT, 64x128x1x1]
%onnx::Conv_686[FLOAT, 64x64x3x3]
%onnx::Conv_689[FLOAT, 64x128x1x1]
%onnx::Conv_692[FLOAT, 64x128x1x1]
%onnx::Conv_695[FLOAT, 64x128x1x1]
%onnx::Conv_698[FLOAT, 64x64x3x3]
%onnx::Conv_701[FLOAT, 128x128x1x1]
%onnx::Conv_704[FLOAT, 128x128x1x1]
%onnx::Conv_707[FLOAT, 128x128x1x1]
%onnx::Conv_710[FLOAT, 128x128x3x3]
%onnx::Conv_713[FLOAT, 128x256x1x1]
%onnx::Conv_716[FLOAT, 128x256x1x1]
%onnx::Conv_719[FLOAT, 128x256x1x1]
%onnx::Conv_722[FLOAT, 128x128x3x3]
%onnx::Conv_725[FLOAT, 128x256x1x1]
%onnx::Conv_728[FLOAT, 128x256x1x1]
%onnx::Conv_731[FLOAT, 128x256x1x1]
%onnx::Conv_734[FLOAT, 128x128x3x3]
%onnx::Conv_737[FLOAT, 256x256x1x1]
%onnx::Conv_738[FLOAT, 256]
%onnx::Conv_740[FLOAT, 256x256x1x1]
%onnx::Conv_743[FLOAT, 256x256x1x1]
%onnx::Conv_746[FLOAT, 256x256x3x3]
%onnx::Conv_749[FLOAT, 256x512x1x1]
%onnx::Conv_752[FLOAT, 256x512x1x1]
%onnx::Conv_755[FLOAT, 256x512x1x1]
%onnx::Conv_758[FLOAT, 256x256x3x3]
%onnx::Conv_761[FLOAT, 256x512x1x1]
%onnx::Conv_764[FLOAT, 256x512x1x1]
%onnx::Conv_767[FLOAT, 256x512x1x1]
%onnx::Conv_770[FLOAT, 256x256x3x3]
) {
%onnx::Conv_771 = Identity(%onnx::Conv_738)
%onnx::Conv_768 = Identity(%onnx::Conv_738)
%onnx::Conv_765 = Identity(%onnx::Conv_738)
%onnx::Conv_762 = Identity(%onnx::Conv_738)
%onnx::Conv_759 = Identity(%onnx::Conv_738)
%onnx::Conv_756 = Identity(%onnx::Conv_738)
%onnx::Conv_753 = Identity(%onnx::Conv_738)
%onnx::Conv_750 = Identity(%onnx::Conv_738)
%onnx::Conv_747 = Identity(%onnx::Conv_738)
%onnx::Conv_744 = Identity(%onnx::Conv_738)
%onnx::Conv_741 = Identity(%onnx::Conv_738)
%onnx::Conv_735 = Identity(%onnx::Conv_663)
%onnx::Conv_732 = Identity(%onnx::Conv_663)
%onnx::Conv_729 = Identity(%onnx::Conv_663)
%onnx::Conv_726 = Identity(%onnx::Conv_663)
%onnx::Conv_723 = Identity(%onnx::Conv_663)
%onnx::Conv_720 = Identity(%onnx::Conv_663)
%onnx::Conv_717 = Identity(%onnx::Conv_663)
%onnx::Conv_714 = Identity(%onnx::Conv_663)
%onnx::Conv_711 = Identity(%onnx::Conv_663)
%onnx::Conv_708 = Identity(%onnx::Conv_663)
%onnx::Conv_705 = Identity(%onnx::Conv_663)
%onnx::Conv_702 = Identity(%onnx::Conv_663)
%onnx::Conv_699 = Identity(%onnx::Conv_666)
%onnx::Conv_696 = Identity(%onnx::Conv_666)
%onnx::Conv_693 = Identity(%onnx::Conv_666)
%onnx::Conv_690 = Identity(%onnx::Conv_666)
%onnx::Conv_687 = Identity(%onnx::Conv_666)
%onnx::Conv_684 = Identity(%onnx::Conv_666)
%onnx::Conv_681 = Identity(%onnx::Conv_666)
%onnx::Conv_678 = Identity(%onnx::Conv_666)
%onnx::Conv_675 = Identity(%onnx::Conv_666)
%onnx::Conv_672 = Identity(%onnx::Conv_666)
%onnx::Conv_669 = Identity(%onnx::Conv_666)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_662, %onnx::Conv_663)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_665, %onnx::Conv_666)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_668, %onnx::Conv_669)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_671, %onnx::Conv_672)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.3/maxpool/MaxPool_output_0, %/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_674, %onnx::Conv_675)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_677, %onnx::Conv_678)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_680, %onnx::Conv_681)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_683, %onnx::Conv_684)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.3/maxpool/MaxPool_output_0, %/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_686, %onnx::Conv_687)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_689, %onnx::Conv_690)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_692, %onnx::Conv_693)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_695, %onnx::Conv_696)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.3/maxpool/MaxPool_output_0, %/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_698, %onnx::Conv_699)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_701, %onnx::Conv_702)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_704, %onnx::Conv_705)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_707, %onnx::Conv_708)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.3/maxpool/MaxPool_output_0, %/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_710, %onnx::Conv_711)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_713, %onnx::Conv_714)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_716, %onnx::Conv_717)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_719, %onnx::Conv_720)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.3/maxpool/MaxPool_output_0, %/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_722, %onnx::Conv_723)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_725, %onnx::Conv_726)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_728, %onnx::Conv_729)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_731, %onnx::Conv_732)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.3/maxpool/MaxPool_output_0, %/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_734, %onnx::Conv_735)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_737, %onnx::Conv_738)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_740, %onnx::Conv_741)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_743, %onnx::Conv_744)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.3/maxpool/MaxPool_output_0, %/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_746, %onnx::Conv_747)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_749, %onnx::Conv_750)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_752, %onnx::Conv_753)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_755, %onnx::Conv_756)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.3/maxpool/MaxPool_output_0, %/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_758, %onnx::Conv_759)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_761, %onnx::Conv_762)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_764, %onnx::Conv_765)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_767, %onnx::Conv_768)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.3/maxpool/MaxPool_output_0, %/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_770, %onnx::Conv_771)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%660 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %660
}
|
val_accuracy
| 92.207533
| 1,101,277,184
| 3,644,554
|
{'zcp_epe_nas': 87.49104712638199, 'zcp_fisher': 1.616672992706298, 'zcp_flops': 17620434944.0, 'zcp_grad_norm': 20.00773048400879, 'zcp_grasp': 0.19603347778320301, 'zcp_jacov': -16.050709532696207, 'zcp_l2_norm': 740.2786254882812, 'zcp_nwot': 217.75061419431157, 'zcp_params': 3644554.0, 'zcp_plain': -0.016322985291481, 'zcp_snip': 133.5330352783203, 'zcp_synflow': 70.694965072497, 'zcp_zen': 73.77278137207031, 'zcp_val_accuracy': 0.922375798225402}
| |
NASBench101_236849
|
NASBench101
|
236849
|
8f4f7484ae970b95a15c3c6c0660c702
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_860[FLOAT, 128x3x3x3]
%onnx::Conv_861[FLOAT, 128]
%onnx::Conv_863[FLOAT, 64x128x1x1]
%onnx::Conv_864[FLOAT, 64]
%onnx::Conv_866[FLOAT, 64x64x1x1]
%onnx::Conv_869[FLOAT, 64x128x1x1]
%onnx::Conv_872[FLOAT, 64x64x1x1]
%onnx::Conv_875[FLOAT, 64x64x3x3]
%onnx::Conv_878[FLOAT, 64x64x1x1]
%onnx::Conv_881[FLOAT, 64x128x1x1]
%onnx::Conv_884[FLOAT, 64x64x1x1]
%onnx::Conv_887[FLOAT, 64x128x1x1]
%onnx::Conv_890[FLOAT, 64x64x1x1]
%onnx::Conv_893[FLOAT, 64x64x3x3]
%onnx::Conv_896[FLOAT, 64x64x1x1]
%onnx::Conv_899[FLOAT, 64x128x1x1]
%onnx::Conv_902[FLOAT, 64x64x1x1]
%onnx::Conv_905[FLOAT, 64x128x1x1]
%onnx::Conv_908[FLOAT, 64x64x1x1]
%onnx::Conv_911[FLOAT, 64x64x3x3]
%onnx::Conv_914[FLOAT, 64x64x1x1]
%onnx::Conv_917[FLOAT, 128x128x1x1]
%onnx::Conv_920[FLOAT, 128x128x1x1]
%onnx::Conv_923[FLOAT, 128x128x1x1]
%onnx::Conv_926[FLOAT, 128x128x1x1]
%onnx::Conv_929[FLOAT, 128x128x3x3]
%onnx::Conv_932[FLOAT, 128x128x1x1]
%onnx::Conv_935[FLOAT, 128x256x1x1]
%onnx::Conv_938[FLOAT, 128x128x1x1]
%onnx::Conv_941[FLOAT, 128x256x1x1]
%onnx::Conv_944[FLOAT, 128x128x1x1]
%onnx::Conv_947[FLOAT, 128x128x3x3]
%onnx::Conv_950[FLOAT, 128x128x1x1]
%onnx::Conv_953[FLOAT, 128x256x1x1]
%onnx::Conv_956[FLOAT, 128x128x1x1]
%onnx::Conv_959[FLOAT, 128x256x1x1]
%onnx::Conv_962[FLOAT, 128x128x1x1]
%onnx::Conv_965[FLOAT, 128x128x3x3]
%onnx::Conv_968[FLOAT, 128x128x1x1]
%onnx::Conv_971[FLOAT, 256x256x1x1]
%onnx::Conv_972[FLOAT, 256]
%onnx::Conv_974[FLOAT, 256x256x1x1]
%onnx::Conv_977[FLOAT, 256x256x1x1]
%onnx::Conv_980[FLOAT, 256x256x1x1]
%onnx::Conv_983[FLOAT, 256x256x3x3]
%onnx::Conv_986[FLOAT, 256x256x1x1]
%onnx::Conv_989[FLOAT, 256x512x1x1]
%onnx::Conv_992[FLOAT, 256x256x1x1]
%onnx::Conv_995[FLOAT, 256x512x1x1]
%onnx::Conv_998[FLOAT, 256x256x1x1]
%onnx::Conv_1001[FLOAT, 256x256x3x3]
%onnx::Conv_1004[FLOAT, 256x256x1x1]
%onnx::Conv_1007[FLOAT, 256x512x1x1]
%onnx::Conv_1010[FLOAT, 256x256x1x1]
%onnx::Conv_1013[FLOAT, 256x512x1x1]
%onnx::Conv_1016[FLOAT, 256x256x1x1]
%onnx::Conv_1019[FLOAT, 256x256x3x3]
%onnx::Conv_1022[FLOAT, 256x256x1x1]
) {
%onnx::Conv_1023 = Identity(%onnx::Conv_972)
%onnx::Conv_1020 = Identity(%onnx::Conv_972)
%onnx::Conv_1017 = Identity(%onnx::Conv_972)
%onnx::Conv_1014 = Identity(%onnx::Conv_972)
%onnx::Conv_1011 = Identity(%onnx::Conv_972)
%onnx::Conv_1008 = Identity(%onnx::Conv_972)
%onnx::Conv_1005 = Identity(%onnx::Conv_972)
%onnx::Conv_1002 = Identity(%onnx::Conv_972)
%onnx::Conv_999 = Identity(%onnx::Conv_972)
%onnx::Conv_996 = Identity(%onnx::Conv_972)
%onnx::Conv_993 = Identity(%onnx::Conv_972)
%onnx::Conv_990 = Identity(%onnx::Conv_972)
%onnx::Conv_987 = Identity(%onnx::Conv_972)
%onnx::Conv_984 = Identity(%onnx::Conv_972)
%onnx::Conv_981 = Identity(%onnx::Conv_972)
%onnx::Conv_978 = Identity(%onnx::Conv_972)
%onnx::Conv_975 = Identity(%onnx::Conv_972)
%onnx::Conv_969 = Identity(%onnx::Conv_861)
%onnx::Conv_966 = Identity(%onnx::Conv_861)
%onnx::Conv_963 = Identity(%onnx::Conv_861)
%onnx::Conv_960 = Identity(%onnx::Conv_861)
%onnx::Conv_957 = Identity(%onnx::Conv_861)
%onnx::Conv_954 = Identity(%onnx::Conv_861)
%onnx::Conv_951 = Identity(%onnx::Conv_861)
%onnx::Conv_948 = Identity(%onnx::Conv_861)
%onnx::Conv_945 = Identity(%onnx::Conv_861)
%onnx::Conv_942 = Identity(%onnx::Conv_861)
%onnx::Conv_939 = Identity(%onnx::Conv_861)
%onnx::Conv_936 = Identity(%onnx::Conv_861)
%onnx::Conv_933 = Identity(%onnx::Conv_861)
%onnx::Conv_930 = Identity(%onnx::Conv_861)
%onnx::Conv_927 = Identity(%onnx::Conv_861)
%onnx::Conv_924 = Identity(%onnx::Conv_861)
%onnx::Conv_921 = Identity(%onnx::Conv_861)
%onnx::Conv_918 = Identity(%onnx::Conv_861)
%onnx::Conv_915 = Identity(%onnx::Conv_864)
%onnx::Conv_912 = Identity(%onnx::Conv_864)
%onnx::Conv_909 = Identity(%onnx::Conv_864)
%onnx::Conv_906 = Identity(%onnx::Conv_864)
%onnx::Conv_903 = Identity(%onnx::Conv_864)
%onnx::Conv_900 = Identity(%onnx::Conv_864)
%onnx::Conv_897 = Identity(%onnx::Conv_864)
%onnx::Conv_894 = Identity(%onnx::Conv_864)
%onnx::Conv_891 = Identity(%onnx::Conv_864)
%onnx::Conv_888 = Identity(%onnx::Conv_864)
%onnx::Conv_885 = Identity(%onnx::Conv_864)
%onnx::Conv_882 = Identity(%onnx::Conv_864)
%onnx::Conv_879 = Identity(%onnx::Conv_864)
%onnx::Conv_876 = Identity(%onnx::Conv_864)
%onnx::Conv_873 = Identity(%onnx::Conv_864)
%onnx::Conv_870 = Identity(%onnx::Conv_864)
%onnx::Conv_867 = Identity(%onnx::Conv_864)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_860, %onnx::Conv_861)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/Add_1_output_0, %/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/Add_1_output_0, %/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/Add_1_output_0, %/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_908, %onnx::Conv_909)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_911, %onnx::Conv_912)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_4_output_0, %onnx::Conv_914, %onnx::Conv_915)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_917, %onnx::Conv_918)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_920, %onnx::Conv_921)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_923, %onnx::Conv_924)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/Add_1_output_0, %/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_926, %onnx::Conv_927)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_929, %onnx::Conv_930)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_932, %onnx::Conv_933)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_935, %onnx::Conv_936)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_938, %onnx::Conv_939)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_941, %onnx::Conv_942)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/Add_1_output_0, %/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_944, %onnx::Conv_945)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_947, %onnx::Conv_948)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_950, %onnx::Conv_951)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_953, %onnx::Conv_954)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_956, %onnx::Conv_957)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_959, %onnx::Conv_960)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/Add_1_output_0, %/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_962, %onnx::Conv_963)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_965, %onnx::Conv_966)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_4_output_0, %onnx::Conv_968, %onnx::Conv_969)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_974, %onnx::Conv_975)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_977, %onnx::Conv_978)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/Add_1_output_0, %/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/Add_1_output_0, %/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/Add_1_output_0, %/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_4_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%858 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %858
}
|
val_accuracy
| 91.566509
| 1,199,056,896
| 3,989,898
|
{'zcp_epe_nas': 92.71660632757015, 'zcp_fisher': 146.53533935546875, 'zcp_flops': 19184910336.0, 'zcp_grad_norm': 240.74276733398438, 'zcp_grasp': -301.1162109375, 'zcp_jacov': -16.061274608964062, 'zcp_l2_norm': 994.8514404296875, 'zcp_nwot': 224.94638362684358, 'zcp_params': 3989898.0, 'zcp_plain': 0.026919210329651003, 'zcp_snip': 1255.90576171875, 'zcp_synflow': 110.54397544616208, 'zcp_zen': 87.28681945800781, 'zcp_val_accuracy': 0.9202724099159241}
| |
NASBench101_374339
|
NASBench101
|
374339
|
e24aa690dc6c8e8929fc15d350e45cdc
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_893[FLOAT, 128x3x3x3]
%onnx::Conv_894[FLOAT, 128]
%onnx::Conv_896[FLOAT, 43x128x1x1]
%onnx::Conv_897[FLOAT, 43]
%onnx::Conv_899[FLOAT, 43x43x1x1]
%onnx::Conv_902[FLOAT, 43x128x1x1]
%onnx::Conv_905[FLOAT, 43x43x3x3]
%onnx::Conv_908[FLOAT, 43x43x1x1]
%onnx::Conv_911[FLOAT, 42x42x1x1]
%onnx::Conv_912[FLOAT, 42]
%onnx::Conv_914[FLOAT, 43x128x1x1]
%onnx::Conv_917[FLOAT, 43x43x1x1]
%onnx::Conv_920[FLOAT, 43x128x1x1]
%onnx::Conv_923[FLOAT, 43x43x3x3]
%onnx::Conv_926[FLOAT, 43x43x1x1]
%onnx::Conv_929[FLOAT, 42x42x1x1]
%onnx::Conv_932[FLOAT, 43x128x1x1]
%onnx::Conv_935[FLOAT, 43x43x1x1]
%onnx::Conv_938[FLOAT, 43x128x1x1]
%onnx::Conv_941[FLOAT, 43x43x3x3]
%onnx::Conv_944[FLOAT, 43x43x1x1]
%onnx::Conv_947[FLOAT, 42x42x1x1]
%onnx::Conv_950[FLOAT, 86x128x1x1]
%onnx::Conv_951[FLOAT, 86]
%onnx::Conv_953[FLOAT, 86x86x1x1]
%onnx::Conv_956[FLOAT, 86x128x1x1]
%onnx::Conv_959[FLOAT, 86x86x3x3]
%onnx::Conv_962[FLOAT, 85x85x1x1]
%onnx::Conv_963[FLOAT, 85]
%onnx::Conv_965[FLOAT, 85x85x1x1]
%onnx::Conv_968[FLOAT, 86x256x1x1]
%onnx::Conv_971[FLOAT, 86x86x1x1]
%onnx::Conv_974[FLOAT, 86x256x1x1]
%onnx::Conv_977[FLOAT, 86x86x3x3]
%onnx::Conv_980[FLOAT, 85x85x1x1]
%onnx::Conv_983[FLOAT, 85x85x1x1]
%onnx::Conv_986[FLOAT, 86x256x1x1]
%onnx::Conv_989[FLOAT, 86x86x1x1]
%onnx::Conv_992[FLOAT, 86x256x1x1]
%onnx::Conv_995[FLOAT, 86x86x3x3]
%onnx::Conv_998[FLOAT, 85x85x1x1]
%onnx::Conv_1001[FLOAT, 85x85x1x1]
%onnx::Conv_1004[FLOAT, 171x256x1x1]
%onnx::Conv_1005[FLOAT, 171]
%onnx::Conv_1007[FLOAT, 171x171x1x1]
%onnx::Conv_1010[FLOAT, 171x256x1x1]
%onnx::Conv_1013[FLOAT, 171x171x3x3]
%onnx::Conv_1016[FLOAT, 171x171x1x1]
%onnx::Conv_1019[FLOAT, 170x170x1x1]
%onnx::Conv_1020[FLOAT, 170]
%onnx::Conv_1022[FLOAT, 171x512x1x1]
%onnx::Conv_1025[FLOAT, 171x171x1x1]
%onnx::Conv_1028[FLOAT, 171x512x1x1]
%onnx::Conv_1031[FLOAT, 171x171x3x3]
%onnx::Conv_1034[FLOAT, 171x171x1x1]
%onnx::Conv_1037[FLOAT, 170x170x1x1]
%onnx::Conv_1040[FLOAT, 171x512x1x1]
%onnx::Conv_1043[FLOAT, 171x171x1x1]
%onnx::Conv_1046[FLOAT, 171x512x1x1]
%onnx::Conv_1049[FLOAT, 171x171x3x3]
%onnx::Conv_1052[FLOAT, 171x171x1x1]
%onnx::Conv_1055[FLOAT, 170x170x1x1]
) {
%onnx::Conv_1056 = Identity(%onnx::Conv_1020)
%onnx::Conv_1053 = Identity(%onnx::Conv_1005)
%onnx::Conv_1050 = Identity(%onnx::Conv_1005)
%onnx::Conv_1047 = Identity(%onnx::Conv_1005)
%onnx::Conv_1044 = Identity(%onnx::Conv_1005)
%onnx::Conv_1041 = Identity(%onnx::Conv_1005)
%onnx::Conv_1038 = Identity(%onnx::Conv_1020)
%onnx::Conv_1035 = Identity(%onnx::Conv_1005)
%onnx::Conv_1032 = Identity(%onnx::Conv_1005)
%onnx::Conv_1029 = Identity(%onnx::Conv_1005)
%onnx::Conv_1026 = Identity(%onnx::Conv_1005)
%onnx::Conv_1023 = Identity(%onnx::Conv_1005)
%onnx::Conv_1017 = Identity(%onnx::Conv_1005)
%onnx::Conv_1014 = Identity(%onnx::Conv_1005)
%onnx::Conv_1011 = Identity(%onnx::Conv_1005)
%onnx::Conv_1008 = Identity(%onnx::Conv_1005)
%onnx::Conv_1002 = Identity(%onnx::Conv_963)
%onnx::Conv_999 = Identity(%onnx::Conv_963)
%onnx::Conv_996 = Identity(%onnx::Conv_951)
%onnx::Conv_993 = Identity(%onnx::Conv_951)
%onnx::Conv_990 = Identity(%onnx::Conv_951)
%onnx::Conv_987 = Identity(%onnx::Conv_951)
%onnx::Conv_984 = Identity(%onnx::Conv_963)
%onnx::Conv_981 = Identity(%onnx::Conv_963)
%onnx::Conv_978 = Identity(%onnx::Conv_951)
%onnx::Conv_975 = Identity(%onnx::Conv_951)
%onnx::Conv_972 = Identity(%onnx::Conv_951)
%onnx::Conv_969 = Identity(%onnx::Conv_951)
%onnx::Conv_966 = Identity(%onnx::Conv_963)
%onnx::Conv_960 = Identity(%onnx::Conv_951)
%onnx::Conv_957 = Identity(%onnx::Conv_951)
%onnx::Conv_954 = Identity(%onnx::Conv_951)
%onnx::Conv_948 = Identity(%onnx::Conv_912)
%onnx::Conv_945 = Identity(%onnx::Conv_897)
%onnx::Conv_942 = Identity(%onnx::Conv_897)
%onnx::Conv_939 = Identity(%onnx::Conv_897)
%onnx::Conv_936 = Identity(%onnx::Conv_897)
%onnx::Conv_933 = Identity(%onnx::Conv_897)
%onnx::Conv_930 = Identity(%onnx::Conv_912)
%onnx::Conv_927 = Identity(%onnx::Conv_897)
%onnx::Conv_924 = Identity(%onnx::Conv_897)
%onnx::Conv_921 = Identity(%onnx::Conv_897)
%onnx::Conv_918 = Identity(%onnx::Conv_897)
%onnx::Conv_915 = Identity(%onnx::Conv_897)
%onnx::Conv_909 = Identity(%onnx::Conv_897)
%onnx::Conv_906 = Identity(%onnx::Conv_897)
%onnx::Conv_903 = Identity(%onnx::Conv_897)
%onnx::Conv_900 = Identity(%onnx::Conv_897)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_893, %onnx::Conv_894)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/Add_1_output_0, %/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_908, %onnx::Conv_909)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.1/Slice_output_0 = Slice(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_4_output_0, %/layers.1/Constant_5_output_0, %/layers.1/Constant_3_output_0, %/layers.1/Constant_6_output_0)
%/layers.1/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_4_output_0 = Add(%/layers.1/Slice_output_0, %/layers.1/Constant_7_output_0)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_911, %onnx::Conv_912)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_914, %onnx::Conv_915)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_917, %onnx::Conv_918)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_920, %onnx::Conv_921)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/Add_1_output_0, %/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_923, %onnx::Conv_924)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_926, %onnx::Conv_927)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.2/Slice_output_0 = Slice(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_4_output_0, %/layers.2/Constant_5_output_0, %/layers.2/Constant_3_output_0, %/layers.2/Constant_6_output_0)
%/layers.2/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_4_output_0 = Add(%/layers.2/Slice_output_0, %/layers.2/Constant_7_output_0)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_929, %onnx::Conv_930)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_932, %onnx::Conv_933)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_935, %onnx::Conv_936)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_938, %onnx::Conv_939)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/Add_1_output_0, %/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_941, %onnx::Conv_942)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_944, %onnx::Conv_945)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.3/Slice_output_0 = Slice(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_4_output_0, %/layers.3/Constant_5_output_0, %/layers.3/Constant_3_output_0, %/layers.3/Constant_6_output_0)
%/layers.3/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_4_output_0 = Add(%/layers.3/Slice_output_0, %/layers.3/Constant_7_output_0)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_4_output_0, %onnx::Conv_947, %onnx::Conv_948)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_950, %onnx::Conv_951)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_953, %onnx::Conv_954)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_956, %onnx::Conv_957)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/Add_1_output_0, %/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_959, %onnx::Conv_960)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.5/Slice_output_0 = Slice(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0, %/layers.5/Constant_4_output_0, %/layers.5/Constant_2_output_0, %/layers.5/Constant_5_output_0)
%/layers.5/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/Slice_output_0, %/layers.5/Constant_6_output_0)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_962, %onnx::Conv_963)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_10_output_0 = Constant[value = <Tensor>]()
%/layers.5/Slice_1_output_0 = Slice(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_8_output_0, %/layers.5/Constant_9_output_0, %/layers.5/Constant_7_output_0, %/layers.5/Constant_10_output_0)
%/layers.5/Constant_11_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_4_output_0 = Add(%/layers.5/Slice_1_output_0, %/layers.5/Constant_11_output_0)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_965, %onnx::Conv_966)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_968, %onnx::Conv_969)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_974, %onnx::Conv_975)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/Add_1_output_0, %/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_977, %onnx::Conv_978)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.6/Slice_output_0 = Slice(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0, %/layers.6/Constant_4_output_0, %/layers.6/Constant_2_output_0, %/layers.6/Constant_5_output_0)
%/layers.6/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/Slice_output_0, %/layers.6/Constant_6_output_0)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_10_output_0 = Constant[value = <Tensor>]()
%/layers.6/Slice_1_output_0 = Slice(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_8_output_0, %/layers.6/Constant_9_output_0, %/layers.6/Constant_7_output_0, %/layers.6/Constant_10_output_0)
%/layers.6/Constant_11_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_4_output_0 = Add(%/layers.6/Slice_1_output_0, %/layers.6/Constant_11_output_0)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/Add_1_output_0, %/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.7/Slice_output_0 = Slice(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0, %/layers.7/Constant_4_output_0, %/layers.7/Constant_2_output_0, %/layers.7/Constant_5_output_0)
%/layers.7/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/Slice_output_0, %/layers.7/Constant_6_output_0)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_10_output_0 = Constant[value = <Tensor>]()
%/layers.7/Slice_1_output_0 = Slice(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_8_output_0, %/layers.7/Constant_9_output_0, %/layers.7/Constant_7_output_0, %/layers.7/Constant_10_output_0)
%/layers.7/Constant_11_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_4_output_0 = Add(%/layers.7/Slice_1_output_0, %/layers.7/Constant_11_output_0)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_4_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/Add_1_output_0, %/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.9/Slice_output_0 = Slice(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_4_output_0, %/layers.9/Constant_5_output_0, %/layers.9/Constant_3_output_0, %/layers.9/Constant_6_output_0)
%/layers.9/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_4_output_0 = Add(%/layers.9/Slice_output_0, %/layers.9/Constant_7_output_0)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_1025, %onnx::Conv_1026)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_1028, %onnx::Conv_1029)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/Add_1_output_0, %/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_1031, %onnx::Conv_1032)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_1034, %onnx::Conv_1035)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.10/Slice_output_0 = Slice(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_4_output_0, %/layers.10/Constant_5_output_0, %/layers.10/Constant_3_output_0, %/layers.10/Constant_6_output_0)
%/layers.10/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_4_output_0 = Add(%/layers.10/Slice_output_0, %/layers.10/Constant_7_output_0)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_1037, %onnx::Conv_1038)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1040, %onnx::Conv_1041)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_1043, %onnx::Conv_1044)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1046, %onnx::Conv_1047)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/Add_1_output_0, %/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_1049, %onnx::Conv_1050)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_1052, %onnx::Conv_1053)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.11/Slice_output_0 = Slice(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_4_output_0, %/layers.11/Constant_5_output_0, %/layers.11/Constant_3_output_0, %/layers.11/Constant_6_output_0)
%/layers.11/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_4_output_0 = Add(%/layers.11/Slice_output_0, %/layers.11/Constant_7_output_0)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_4_output_0, %onnx::Conv_1055, %onnx::Conv_1056)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%891 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %891
}
|
val_accuracy
| 92.127407
| 605,422,720
| 1,983,674
|
{'zcp_epe_nas': 75.47377737133054, 'zcp_fisher': 5.097987651824951, 'zcp_flops': 9686763520.0, 'zcp_grad_norm': 53.162715911865234, 'zcp_grasp': 3.66656494140625, 'zcp_jacov': -16.051669338987743, 'zcp_l2_norm': 883.6741943359375, 'zcp_nwot': 218.64112674926557, 'zcp_params': 1983674.0, 'zcp_plain': 0.0158785097301, 'zcp_snip': 244.57447814941406, 'zcp_synflow': 101.59167262089068, 'zcp_zen': 77.50706481933594, 'zcp_val_accuracy': 0.936298072338104}
| |
NASBench101_285718
|
NASBench101
|
285718
|
acf3e1af58990fc3cc271928d942480d
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_878[FLOAT, 128x3x3x3]
%onnx::Conv_879[FLOAT, 128]
%onnx::Conv_881[FLOAT, 64x128x1x1]
%onnx::Conv_882[FLOAT, 64]
%onnx::Conv_884[FLOAT, 64x64x3x3]
%onnx::Conv_887[FLOAT, 64x128x1x1]
%onnx::Conv_890[FLOAT, 64x64x3x3]
%onnx::Conv_893[FLOAT, 64x64x1x1]
%onnx::Conv_896[FLOAT, 64x64x3x3]
%onnx::Conv_899[FLOAT, 64x128x1x1]
%onnx::Conv_902[FLOAT, 64x64x3x3]
%onnx::Conv_905[FLOAT, 64x128x1x1]
%onnx::Conv_908[FLOAT, 64x64x3x3]
%onnx::Conv_911[FLOAT, 64x64x1x1]
%onnx::Conv_914[FLOAT, 64x64x3x3]
%onnx::Conv_917[FLOAT, 64x128x1x1]
%onnx::Conv_920[FLOAT, 64x64x3x3]
%onnx::Conv_923[FLOAT, 64x128x1x1]
%onnx::Conv_926[FLOAT, 64x64x3x3]
%onnx::Conv_929[FLOAT, 64x64x1x1]
%onnx::Conv_932[FLOAT, 64x64x3x3]
%onnx::Conv_935[FLOAT, 128x128x1x1]
%onnx::Conv_938[FLOAT, 128x128x3x3]
%onnx::Conv_941[FLOAT, 128x128x1x1]
%onnx::Conv_944[FLOAT, 128x128x3x3]
%onnx::Conv_947[FLOAT, 128x128x1x1]
%onnx::Conv_950[FLOAT, 128x128x3x3]
%onnx::Conv_953[FLOAT, 128x256x1x1]
%onnx::Conv_956[FLOAT, 128x128x3x3]
%onnx::Conv_959[FLOAT, 128x256x1x1]
%onnx::Conv_962[FLOAT, 128x128x3x3]
%onnx::Conv_965[FLOAT, 128x128x1x1]
%onnx::Conv_968[FLOAT, 128x128x3x3]
%onnx::Conv_971[FLOAT, 128x256x1x1]
%onnx::Conv_974[FLOAT, 128x128x3x3]
%onnx::Conv_977[FLOAT, 128x256x1x1]
%onnx::Conv_980[FLOAT, 128x128x3x3]
%onnx::Conv_983[FLOAT, 128x128x1x1]
%onnx::Conv_986[FLOAT, 128x128x3x3]
%onnx::Conv_989[FLOAT, 256x256x1x1]
%onnx::Conv_990[FLOAT, 256]
%onnx::Conv_992[FLOAT, 256x256x3x3]
%onnx::Conv_995[FLOAT, 256x256x1x1]
%onnx::Conv_998[FLOAT, 256x256x3x3]
%onnx::Conv_1001[FLOAT, 256x256x1x1]
%onnx::Conv_1004[FLOAT, 256x256x3x3]
%onnx::Conv_1007[FLOAT, 256x512x1x1]
%onnx::Conv_1010[FLOAT, 256x256x3x3]
%onnx::Conv_1013[FLOAT, 256x512x1x1]
%onnx::Conv_1016[FLOAT, 256x256x3x3]
%onnx::Conv_1019[FLOAT, 256x256x1x1]
%onnx::Conv_1022[FLOAT, 256x256x3x3]
%onnx::Conv_1025[FLOAT, 256x512x1x1]
%onnx::Conv_1028[FLOAT, 256x256x3x3]
%onnx::Conv_1031[FLOAT, 256x512x1x1]
%onnx::Conv_1034[FLOAT, 256x256x3x3]
%onnx::Conv_1037[FLOAT, 256x256x1x1]
%onnx::Conv_1040[FLOAT, 256x256x3x3]
) {
%onnx::Conv_1041 = Identity(%onnx::Conv_990)
%onnx::Conv_1038 = Identity(%onnx::Conv_990)
%onnx::Conv_1035 = Identity(%onnx::Conv_990)
%onnx::Conv_1032 = Identity(%onnx::Conv_990)
%onnx::Conv_1029 = Identity(%onnx::Conv_990)
%onnx::Conv_1026 = Identity(%onnx::Conv_990)
%onnx::Conv_1023 = Identity(%onnx::Conv_990)
%onnx::Conv_1020 = Identity(%onnx::Conv_990)
%onnx::Conv_1017 = Identity(%onnx::Conv_990)
%onnx::Conv_1014 = Identity(%onnx::Conv_990)
%onnx::Conv_1011 = Identity(%onnx::Conv_990)
%onnx::Conv_1008 = Identity(%onnx::Conv_990)
%onnx::Conv_1005 = Identity(%onnx::Conv_990)
%onnx::Conv_1002 = Identity(%onnx::Conv_990)
%onnx::Conv_999 = Identity(%onnx::Conv_990)
%onnx::Conv_996 = Identity(%onnx::Conv_990)
%onnx::Conv_993 = Identity(%onnx::Conv_990)
%onnx::Conv_987 = Identity(%onnx::Conv_879)
%onnx::Conv_984 = Identity(%onnx::Conv_879)
%onnx::Conv_981 = Identity(%onnx::Conv_879)
%onnx::Conv_978 = Identity(%onnx::Conv_879)
%onnx::Conv_975 = Identity(%onnx::Conv_879)
%onnx::Conv_972 = Identity(%onnx::Conv_879)
%onnx::Conv_969 = Identity(%onnx::Conv_879)
%onnx::Conv_966 = Identity(%onnx::Conv_879)
%onnx::Conv_963 = Identity(%onnx::Conv_879)
%onnx::Conv_960 = Identity(%onnx::Conv_879)
%onnx::Conv_957 = Identity(%onnx::Conv_879)
%onnx::Conv_954 = Identity(%onnx::Conv_879)
%onnx::Conv_951 = Identity(%onnx::Conv_879)
%onnx::Conv_948 = Identity(%onnx::Conv_879)
%onnx::Conv_945 = Identity(%onnx::Conv_879)
%onnx::Conv_942 = Identity(%onnx::Conv_879)
%onnx::Conv_939 = Identity(%onnx::Conv_879)
%onnx::Conv_936 = Identity(%onnx::Conv_879)
%onnx::Conv_933 = Identity(%onnx::Conv_882)
%onnx::Conv_930 = Identity(%onnx::Conv_882)
%onnx::Conv_927 = Identity(%onnx::Conv_882)
%onnx::Conv_924 = Identity(%onnx::Conv_882)
%onnx::Conv_921 = Identity(%onnx::Conv_882)
%onnx::Conv_918 = Identity(%onnx::Conv_882)
%onnx::Conv_915 = Identity(%onnx::Conv_882)
%onnx::Conv_912 = Identity(%onnx::Conv_882)
%onnx::Conv_909 = Identity(%onnx::Conv_882)
%onnx::Conv_906 = Identity(%onnx::Conv_882)
%onnx::Conv_903 = Identity(%onnx::Conv_882)
%onnx::Conv_900 = Identity(%onnx::Conv_882)
%onnx::Conv_897 = Identity(%onnx::Conv_882)
%onnx::Conv_894 = Identity(%onnx::Conv_882)
%onnx::Conv_891 = Identity(%onnx::Conv_882)
%onnx::Conv_888 = Identity(%onnx::Conv_882)
%onnx::Conv_885 = Identity(%onnx::Conv_882)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_878, %onnx::Conv_879)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/Add_3_output_0, %/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_5_output_0 = Add(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_4_output_0)
%/layers.1/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_5_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_908, %onnx::Conv_909)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_911, %onnx::Conv_912)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/Add_3_output_0, %/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_914, %onnx::Conv_915)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_5_output_0 = Add(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_4_output_0)
%/layers.2/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_5_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_917, %onnx::Conv_918)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_920, %onnx::Conv_921)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_923, %onnx::Conv_924)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_926, %onnx::Conv_927)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_929, %onnx::Conv_930)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/Add_3_output_0, %/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_4_output_0, %onnx::Conv_932, %onnx::Conv_933)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_5_output_0 = Add(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_4_output_0)
%/layers.3/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_5_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_935, %onnx::Conv_936)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_938, %onnx::Conv_939)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_941, %onnx::Conv_942)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_944, %onnx::Conv_945)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_947, %onnx::Conv_948)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/Add_3_output_0, %/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_950, %onnx::Conv_951)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_5_output_0 = Add(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_4_output_0)
%/layers.5/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_5_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_953, %onnx::Conv_954)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_956, %onnx::Conv_957)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_959, %onnx::Conv_960)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_962, %onnx::Conv_963)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_965, %onnx::Conv_966)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/Add_3_output_0, %/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_968, %onnx::Conv_969)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_5_output_0 = Add(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_4_output_0)
%/layers.6/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_5_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_974, %onnx::Conv_975)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_977, %onnx::Conv_978)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/Add_3_output_0, %/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_4_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_5_output_0 = Add(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_4_output_0)
%/layers.7/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_5_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/Add_3_output_0, %/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_5_output_0 = Add(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_4_output_0)
%/layers.9/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_5_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/Add_3_output_0, %/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_5_output_0 = Add(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_4_output_0)
%/layers.10/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_5_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1025, %onnx::Conv_1026)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_1028, %onnx::Conv_1029)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1031, %onnx::Conv_1032)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_1034, %onnx::Conv_1035)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_1037, %onnx::Conv_1038)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/Add_3_output_0, %/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_4_output_0, %onnx::Conv_1040, %onnx::Conv_1041)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_5_output_0 = Add(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_4_output_0)
%/layers.11/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_5_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.5/maxpool/MaxPool_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%876 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %876
}
|
val_accuracy
| 91.40625
| 2,407,016,448
| 8,118,666
|
{'zcp_epe_nas': 142.66293978575112, 'zcp_fisher': 593.5601806640625, 'zcp_flops': 38512263168.0, 'zcp_grad_norm': 440.42022705078125, 'zcp_grasp': -579.634765625, 'zcp_jacov': -16.055924352491985, 'zcp_l2_norm': 993.5210571289062, 'zcp_nwot': 224.3416338243335, 'zcp_params': 8118666.0, 'zcp_plain': 0.08270738273859, 'zcp_snip': 2712.9033203125, 'zcp_synflow': 123.84951029055162, 'zcp_zen': 104.7122802734375, 'zcp_val_accuracy': 0.904046475887298}
| |
NASBench101_89968
|
NASBench101
|
89968
|
367482fc5a1a385df34917892a0f6804
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_419[FLOAT, 128x3x3x3]
%onnx::Conv_420[FLOAT, 128]
%onnx::Conv_422[FLOAT, 64x128x1x1]
%onnx::Conv_423[FLOAT, 64]
%onnx::Conv_425[FLOAT, 64x64x3x3]
%onnx::Conv_428[FLOAT, 64x128x1x1]
%onnx::Conv_431[FLOAT, 64x64x3x3]
%onnx::Conv_434[FLOAT, 64x128x1x1]
%onnx::Conv_437[FLOAT, 64x64x3x3]
%onnx::Conv_440[FLOAT, 128x128x1x1]
%onnx::Conv_443[FLOAT, 128x128x3x3]
%onnx::Conv_446[FLOAT, 128x256x1x1]
%onnx::Conv_449[FLOAT, 128x128x3x3]
%onnx::Conv_452[FLOAT, 128x256x1x1]
%onnx::Conv_455[FLOAT, 128x128x3x3]
%onnx::Conv_458[FLOAT, 256x256x1x1]
%onnx::Conv_459[FLOAT, 256]
%onnx::Conv_461[FLOAT, 256x256x3x3]
%onnx::Conv_464[FLOAT, 256x512x1x1]
%onnx::Conv_467[FLOAT, 256x256x3x3]
%onnx::Conv_470[FLOAT, 256x512x1x1]
%onnx::Conv_473[FLOAT, 256x256x3x3]
) {
%onnx::Conv_474 = Identity(%onnx::Conv_459)
%onnx::Conv_471 = Identity(%onnx::Conv_459)
%onnx::Conv_468 = Identity(%onnx::Conv_459)
%onnx::Conv_465 = Identity(%onnx::Conv_459)
%onnx::Conv_462 = Identity(%onnx::Conv_459)
%onnx::Conv_456 = Identity(%onnx::Conv_420)
%onnx::Conv_453 = Identity(%onnx::Conv_420)
%onnx::Conv_450 = Identity(%onnx::Conv_420)
%onnx::Conv_447 = Identity(%onnx::Conv_420)
%onnx::Conv_444 = Identity(%onnx::Conv_420)
%onnx::Conv_441 = Identity(%onnx::Conv_420)
%onnx::Conv_438 = Identity(%onnx::Conv_423)
%onnx::Conv_435 = Identity(%onnx::Conv_423)
%onnx::Conv_432 = Identity(%onnx::Conv_423)
%onnx::Conv_429 = Identity(%onnx::Conv_423)
%onnx::Conv_426 = Identity(%onnx::Conv_423)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_419, %onnx::Conv_420)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_422, %onnx::Conv_423)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_425, %onnx::Conv_426)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %/layers.1/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_428, %onnx::Conv_429)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_431, %onnx::Conv_432)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %/layers.2/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_434, %onnx::Conv_435)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_437, %onnx::Conv_438)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %/layers.3/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_440, %onnx::Conv_441)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_443, %onnx::Conv_444)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %/layers.5/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_446, %onnx::Conv_447)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_449, %onnx::Conv_450)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %/layers.6/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_452, %onnx::Conv_453)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_455, %onnx::Conv_456)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %/layers.7/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_458, %onnx::Conv_459)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_461, %onnx::Conv_462)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %/layers.9/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_464, %onnx::Conv_465)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_467, %onnx::Conv_468)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %/layers.10/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_470, %onnx::Conv_471)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_473, %onnx::Conv_474)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %/layers.11/vertex_op.4/maxpool/MaxPool_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%417 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %417
}
|
val_accuracy
| 89.022434
| 827,336,704
| 2,770,826
|
{'zcp_epe_nas': 103.33644356482556, 'zcp_fisher': 6.502097129821777, 'zcp_flops': 13237387264.0, 'zcp_grad_norm': 40.96675109863281, 'zcp_grasp': -1.501617431640625, 'zcp_jacov': -16.059234026434908, 'zcp_l2_norm': 348.86846923828125, 'zcp_nwot': 208.42914814204298, 'zcp_params': 2770826.0, 'zcp_plain': 0.018726067617535, 'zcp_snip': 245.83058166503906, 'zcp_synflow': 61.196179654816035, 'zcp_zen': 41.933719635009766, 'zcp_val_accuracy': 0.9308894276618951}
| |
NASBench101_422485
|
NASBench101
|
422485
|
ff4d09dc9fa1daa2f3655e4269995047
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_779[FLOAT, 128x3x3x3]
%onnx::Conv_780[FLOAT, 128]
%onnx::Conv_782[FLOAT, 64x128x1x1]
%onnx::Conv_783[FLOAT, 64]
%onnx::Conv_785[FLOAT, 64x64x3x3]
%onnx::Conv_788[FLOAT, 64x64x3x3]
%onnx::Conv_791[FLOAT, 64x128x1x1]
%onnx::Conv_794[FLOAT, 64x64x3x3]
%onnx::Conv_797[FLOAT, 64x128x1x1]
%onnx::Conv_800[FLOAT, 64x64x3x3]
%onnx::Conv_803[FLOAT, 64x64x3x3]
%onnx::Conv_806[FLOAT, 64x128x1x1]
%onnx::Conv_809[FLOAT, 64x64x3x3]
%onnx::Conv_812[FLOAT, 64x128x1x1]
%onnx::Conv_815[FLOAT, 64x64x3x3]
%onnx::Conv_818[FLOAT, 64x64x3x3]
%onnx::Conv_821[FLOAT, 64x128x1x1]
%onnx::Conv_824[FLOAT, 64x64x3x3]
%onnx::Conv_827[FLOAT, 128x128x1x1]
%onnx::Conv_830[FLOAT, 128x128x3x3]
%onnx::Conv_833[FLOAT, 128x128x3x3]
%onnx::Conv_836[FLOAT, 128x128x1x1]
%onnx::Conv_839[FLOAT, 128x128x3x3]
%onnx::Conv_842[FLOAT, 128x256x1x1]
%onnx::Conv_845[FLOAT, 128x128x3x3]
%onnx::Conv_848[FLOAT, 128x128x3x3]
%onnx::Conv_851[FLOAT, 128x256x1x1]
%onnx::Conv_854[FLOAT, 128x128x3x3]
%onnx::Conv_857[FLOAT, 128x256x1x1]
%onnx::Conv_860[FLOAT, 128x128x3x3]
%onnx::Conv_863[FLOAT, 128x128x3x3]
%onnx::Conv_866[FLOAT, 128x256x1x1]
%onnx::Conv_869[FLOAT, 128x128x3x3]
%onnx::Conv_872[FLOAT, 256x256x1x1]
%onnx::Conv_873[FLOAT, 256]
%onnx::Conv_875[FLOAT, 256x256x3x3]
%onnx::Conv_878[FLOAT, 256x256x3x3]
%onnx::Conv_881[FLOAT, 256x256x1x1]
%onnx::Conv_884[FLOAT, 256x256x3x3]
%onnx::Conv_887[FLOAT, 256x512x1x1]
%onnx::Conv_890[FLOAT, 256x256x3x3]
%onnx::Conv_893[FLOAT, 256x256x3x3]
%onnx::Conv_896[FLOAT, 256x512x1x1]
%onnx::Conv_899[FLOAT, 256x256x3x3]
%onnx::Conv_902[FLOAT, 256x512x1x1]
%onnx::Conv_905[FLOAT, 256x256x3x3]
%onnx::Conv_908[FLOAT, 256x256x3x3]
%onnx::Conv_911[FLOAT, 256x512x1x1]
%onnx::Conv_914[FLOAT, 256x256x3x3]
) {
%onnx::Conv_915 = Identity(%onnx::Conv_873)
%onnx::Conv_912 = Identity(%onnx::Conv_873)
%onnx::Conv_909 = Identity(%onnx::Conv_873)
%onnx::Conv_906 = Identity(%onnx::Conv_873)
%onnx::Conv_903 = Identity(%onnx::Conv_873)
%onnx::Conv_900 = Identity(%onnx::Conv_873)
%onnx::Conv_897 = Identity(%onnx::Conv_873)
%onnx::Conv_894 = Identity(%onnx::Conv_873)
%onnx::Conv_891 = Identity(%onnx::Conv_873)
%onnx::Conv_888 = Identity(%onnx::Conv_873)
%onnx::Conv_885 = Identity(%onnx::Conv_873)
%onnx::Conv_882 = Identity(%onnx::Conv_873)
%onnx::Conv_879 = Identity(%onnx::Conv_873)
%onnx::Conv_876 = Identity(%onnx::Conv_873)
%onnx::Conv_870 = Identity(%onnx::Conv_780)
%onnx::Conv_867 = Identity(%onnx::Conv_780)
%onnx::Conv_864 = Identity(%onnx::Conv_780)
%onnx::Conv_861 = Identity(%onnx::Conv_780)
%onnx::Conv_858 = Identity(%onnx::Conv_780)
%onnx::Conv_855 = Identity(%onnx::Conv_780)
%onnx::Conv_852 = Identity(%onnx::Conv_780)
%onnx::Conv_849 = Identity(%onnx::Conv_780)
%onnx::Conv_846 = Identity(%onnx::Conv_780)
%onnx::Conv_843 = Identity(%onnx::Conv_780)
%onnx::Conv_840 = Identity(%onnx::Conv_780)
%onnx::Conv_837 = Identity(%onnx::Conv_780)
%onnx::Conv_834 = Identity(%onnx::Conv_780)
%onnx::Conv_831 = Identity(%onnx::Conv_780)
%onnx::Conv_828 = Identity(%onnx::Conv_780)
%onnx::Conv_825 = Identity(%onnx::Conv_783)
%onnx::Conv_822 = Identity(%onnx::Conv_783)
%onnx::Conv_819 = Identity(%onnx::Conv_783)
%onnx::Conv_816 = Identity(%onnx::Conv_783)
%onnx::Conv_813 = Identity(%onnx::Conv_783)
%onnx::Conv_810 = Identity(%onnx::Conv_783)
%onnx::Conv_807 = Identity(%onnx::Conv_783)
%onnx::Conv_804 = Identity(%onnx::Conv_783)
%onnx::Conv_801 = Identity(%onnx::Conv_783)
%onnx::Conv_798 = Identity(%onnx::Conv_783)
%onnx::Conv_795 = Identity(%onnx::Conv_783)
%onnx::Conv_792 = Identity(%onnx::Conv_783)
%onnx::Conv_789 = Identity(%onnx::Conv_783)
%onnx::Conv_786 = Identity(%onnx::Conv_783)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_779, %onnx::Conv_780)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_782, %onnx::Conv_783)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_785, %onnx::Conv_786)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_788, %onnx::Conv_789)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.3/maxpool/MaxPool_output_0, %/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/Add_5_output_0 = Add(%/layers.1/Add_4_output_0, %/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_5_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_797, %onnx::Conv_798)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_800, %onnx::Conv_801)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_803, %onnx::Conv_804)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_806, %onnx::Conv_807)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.3/maxpool/MaxPool_output_0, %/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_809, %onnx::Conv_810)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/Add_5_output_0 = Add(%/layers.2/Add_4_output_0, %/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_5_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_812, %onnx::Conv_813)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_815, %onnx::Conv_816)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_818, %onnx::Conv_819)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_821, %onnx::Conv_822)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.3/maxpool/MaxPool_output_0, %/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_824, %onnx::Conv_825)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/Add_5_output_0 = Add(%/layers.3/Add_4_output_0, %/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_5_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_827, %onnx::Conv_828)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_830, %onnx::Conv_831)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_833, %onnx::Conv_834)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_836, %onnx::Conv_837)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.3/maxpool/MaxPool_output_0, %/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_839, %onnx::Conv_840)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/Add_5_output_0 = Add(%/layers.5/Add_4_output_0, %/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_5_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_842, %onnx::Conv_843)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_845, %onnx::Conv_846)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_848, %onnx::Conv_849)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_851, %onnx::Conv_852)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.3/maxpool/MaxPool_output_0, %/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_854, %onnx::Conv_855)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/Add_5_output_0 = Add(%/layers.6/Add_4_output_0, %/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_5_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_857, %onnx::Conv_858)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.3/maxpool/MaxPool_output_0, %/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/Add_5_output_0 = Add(%/layers.7/Add_4_output_0, %/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_5_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.3/maxpool/MaxPool_output_0, %/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/Add_5_output_0 = Add(%/layers.9/Add_4_output_0, %/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_5_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.3/maxpool/MaxPool_output_0, %/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/Add_5_output_0 = Add(%/layers.10/Add_4_output_0, %/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_5_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_908, %onnx::Conv_909)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_911, %onnx::Conv_912)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.3/maxpool/MaxPool_output_0, %/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_914, %onnx::Conv_915)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/Add_5_output_0 = Add(%/layers.11/Add_4_output_0, %/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_5_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.5/maxpool/MaxPool_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%777 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %777
}
|
val_accuracy
| 92.638218
| 2,328,766,464
| 7,857,930
|
{'zcp_epe_nas': 144.07321811446272, 'zcp_fisher': 124.78254699707031, 'zcp_flops': 37260263424.0, 'zcp_grad_norm': 199.72518920898438, 'zcp_grasp': -112.650390625, 'zcp_jacov': -16.04400845416389, 'zcp_l2_norm': 844.482666015625, 'zcp_nwot': 221.66640628871846, 'zcp_params': 7857930.0, 'zcp_plain': 0.056271102279424, 'zcp_snip': 1187.0421142578125, 'zcp_synflow': 127.17727834061321, 'zcp_zen': 99.07115173339844, 'zcp_val_accuracy': 0.9287860393524171}
| |
NASBench101_45483
|
NASBench101
|
45483
|
1b971177017dfc2b61f02c2a977b2344
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_608[FLOAT, 128x3x3x3]
%onnx::Conv_609[FLOAT, 128]
%onnx::Conv_611[FLOAT, 64x128x1x1]
%onnx::Conv_612[FLOAT, 64]
%onnx::Conv_614[FLOAT, 64x64x1x1]
%onnx::Conv_617[FLOAT, 64x64x3x3]
%onnx::Conv_620[FLOAT, 64x64x3x3]
%onnx::Conv_623[FLOAT, 64x128x1x1]
%onnx::Conv_626[FLOAT, 64x64x1x1]
%onnx::Conv_629[FLOAT, 64x64x3x3]
%onnx::Conv_632[FLOAT, 64x64x3x3]
%onnx::Conv_635[FLOAT, 64x128x1x1]
%onnx::Conv_638[FLOAT, 64x64x1x1]
%onnx::Conv_641[FLOAT, 64x64x3x3]
%onnx::Conv_644[FLOAT, 64x64x3x3]
%onnx::Conv_647[FLOAT, 128x128x1x1]
%onnx::Conv_650[FLOAT, 128x128x1x1]
%onnx::Conv_653[FLOAT, 128x128x3x3]
%onnx::Conv_656[FLOAT, 128x128x3x3]
%onnx::Conv_659[FLOAT, 128x256x1x1]
%onnx::Conv_662[FLOAT, 128x128x1x1]
%onnx::Conv_665[FLOAT, 128x128x3x3]
%onnx::Conv_668[FLOAT, 128x128x3x3]
%onnx::Conv_671[FLOAT, 128x256x1x1]
%onnx::Conv_674[FLOAT, 128x128x1x1]
%onnx::Conv_677[FLOAT, 128x128x3x3]
%onnx::Conv_680[FLOAT, 128x128x3x3]
%onnx::Conv_683[FLOAT, 256x256x1x1]
%onnx::Conv_684[FLOAT, 256]
%onnx::Conv_686[FLOAT, 256x256x1x1]
%onnx::Conv_689[FLOAT, 256x256x3x3]
%onnx::Conv_692[FLOAT, 256x256x3x3]
%onnx::Conv_695[FLOAT, 256x512x1x1]
%onnx::Conv_698[FLOAT, 256x256x1x1]
%onnx::Conv_701[FLOAT, 256x256x3x3]
%onnx::Conv_704[FLOAT, 256x256x3x3]
%onnx::Conv_707[FLOAT, 256x512x1x1]
%onnx::Conv_710[FLOAT, 256x256x1x1]
%onnx::Conv_713[FLOAT, 256x256x3x3]
%onnx::Conv_716[FLOAT, 256x256x3x3]
) {
%onnx::Conv_717 = Identity(%onnx::Conv_684)
%onnx::Conv_714 = Identity(%onnx::Conv_684)
%onnx::Conv_711 = Identity(%onnx::Conv_684)
%onnx::Conv_708 = Identity(%onnx::Conv_684)
%onnx::Conv_705 = Identity(%onnx::Conv_684)
%onnx::Conv_702 = Identity(%onnx::Conv_684)
%onnx::Conv_699 = Identity(%onnx::Conv_684)
%onnx::Conv_696 = Identity(%onnx::Conv_684)
%onnx::Conv_693 = Identity(%onnx::Conv_684)
%onnx::Conv_690 = Identity(%onnx::Conv_684)
%onnx::Conv_687 = Identity(%onnx::Conv_684)
%onnx::Conv_681 = Identity(%onnx::Conv_609)
%onnx::Conv_678 = Identity(%onnx::Conv_609)
%onnx::Conv_675 = Identity(%onnx::Conv_609)
%onnx::Conv_672 = Identity(%onnx::Conv_609)
%onnx::Conv_669 = Identity(%onnx::Conv_609)
%onnx::Conv_666 = Identity(%onnx::Conv_609)
%onnx::Conv_663 = Identity(%onnx::Conv_609)
%onnx::Conv_660 = Identity(%onnx::Conv_609)
%onnx::Conv_657 = Identity(%onnx::Conv_609)
%onnx::Conv_654 = Identity(%onnx::Conv_609)
%onnx::Conv_651 = Identity(%onnx::Conv_609)
%onnx::Conv_648 = Identity(%onnx::Conv_609)
%onnx::Conv_645 = Identity(%onnx::Conv_612)
%onnx::Conv_642 = Identity(%onnx::Conv_612)
%onnx::Conv_639 = Identity(%onnx::Conv_612)
%onnx::Conv_636 = Identity(%onnx::Conv_612)
%onnx::Conv_633 = Identity(%onnx::Conv_612)
%onnx::Conv_630 = Identity(%onnx::Conv_612)
%onnx::Conv_627 = Identity(%onnx::Conv_612)
%onnx::Conv_624 = Identity(%onnx::Conv_612)
%onnx::Conv_621 = Identity(%onnx::Conv_612)
%onnx::Conv_618 = Identity(%onnx::Conv_612)
%onnx::Conv_615 = Identity(%onnx::Conv_612)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_608, %onnx::Conv_609)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_611, %onnx::Conv_612)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_614, %onnx::Conv_615)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_617, %onnx::Conv_618)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_620, %onnx::Conv_621)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_623, %onnx::Conv_624)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_626, %onnx::Conv_627)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_629, %onnx::Conv_630)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_632, %onnx::Conv_633)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_635, %onnx::Conv_636)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_638, %onnx::Conv_639)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_641, %onnx::Conv_642)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_644, %onnx::Conv_645)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_647, %onnx::Conv_648)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_650, %onnx::Conv_651)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_653, %onnx::Conv_654)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_656, %onnx::Conv_657)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_659, %onnx::Conv_660)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_662, %onnx::Conv_663)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_665, %onnx::Conv_666)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_668, %onnx::Conv_669)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_671, %onnx::Conv_672)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_674, %onnx::Conv_675)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_677, %onnx::Conv_678)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_680, %onnx::Conv_681)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_683, %onnx::Conv_684)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_686, %onnx::Conv_687)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_689, %onnx::Conv_690)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_692, %onnx::Conv_693)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_695, %onnx::Conv_696)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_698, %onnx::Conv_699)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_701, %onnx::Conv_702)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_704, %onnx::Conv_705)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_707, %onnx::Conv_708)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_710, %onnx::Conv_711)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_713, %onnx::Conv_714)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_716, %onnx::Conv_717)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%606 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %606
}
|
val_accuracy
| 91.546476
| 1,587,816,448
| 5,356,682
|
{'zcp_epe_nas': 65.50901688237101, 'zcp_fisher': 5.808039665222168, 'zcp_flops': 25405063168.0, 'zcp_grad_norm': 42.69853591918945, 'zcp_grasp': -0.669036865234375, 'zcp_jacov': -16.05963756289249, 'zcp_l2_norm': 648.4828491210938, 'zcp_nwot': 218.09277147921603, 'zcp_params': 5356682.0, 'zcp_plain': 0.043052922934293004, 'zcp_snip': 275.9570007324219, 'zcp_synflow': 117.01313827347627, 'zcp_zen': 75.05058288574219, 'zcp_val_accuracy': 0.9102563858032221}
| |
NASBench101_35355
|
NASBench101
|
35355
|
156f2207a5ad81330ce946ec5b92fb46
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_869[FLOAT, 128x3x3x3]
%onnx::Conv_870[FLOAT, 128]
%onnx::Conv_872[FLOAT, 128x128x1x1]
%onnx::Conv_875[FLOAT, 128x128x1x1]
%onnx::Conv_878[FLOAT, 128x128x3x3]
%onnx::Conv_881[FLOAT, 128x128x3x3]
%onnx::Conv_884[FLOAT, 128x128x1x1]
%onnx::Conv_887[FLOAT, 128x128x1x1]
%onnx::Conv_890[FLOAT, 128x128x1x1]
%onnx::Conv_893[FLOAT, 128x128x1x1]
%onnx::Conv_896[FLOAT, 128x128x3x3]
%onnx::Conv_899[FLOAT, 128x128x3x3]
%onnx::Conv_902[FLOAT, 128x128x1x1]
%onnx::Conv_905[FLOAT, 128x128x1x1]
%onnx::Conv_908[FLOAT, 128x128x1x1]
%onnx::Conv_911[FLOAT, 128x128x1x1]
%onnx::Conv_914[FLOAT, 128x128x3x3]
%onnx::Conv_917[FLOAT, 128x128x3x3]
%onnx::Conv_920[FLOAT, 128x128x1x1]
%onnx::Conv_923[FLOAT, 128x128x1x1]
%onnx::Conv_926[FLOAT, 256x128x1x1]
%onnx::Conv_927[FLOAT, 256]
%onnx::Conv_929[FLOAT, 256x256x1x1]
%onnx::Conv_932[FLOAT, 256x256x3x3]
%onnx::Conv_935[FLOAT, 256x256x3x3]
%onnx::Conv_938[FLOAT, 256x128x1x1]
%onnx::Conv_941[FLOAT, 256x256x1x1]
%onnx::Conv_944[FLOAT, 256x256x1x1]
%onnx::Conv_947[FLOAT, 256x256x1x1]
%onnx::Conv_950[FLOAT, 256x256x3x3]
%onnx::Conv_953[FLOAT, 256x256x3x3]
%onnx::Conv_956[FLOAT, 256x256x1x1]
%onnx::Conv_959[FLOAT, 256x256x1x1]
%onnx::Conv_962[FLOAT, 256x256x1x1]
%onnx::Conv_965[FLOAT, 256x256x1x1]
%onnx::Conv_968[FLOAT, 256x256x3x3]
%onnx::Conv_971[FLOAT, 256x256x3x3]
%onnx::Conv_974[FLOAT, 256x256x1x1]
%onnx::Conv_977[FLOAT, 256x256x1x1]
%onnx::Conv_980[FLOAT, 512x256x1x1]
%onnx::Conv_981[FLOAT, 512]
%onnx::Conv_983[FLOAT, 512x512x1x1]
%onnx::Conv_986[FLOAT, 512x512x3x3]
%onnx::Conv_989[FLOAT, 512x512x3x3]
%onnx::Conv_992[FLOAT, 512x256x1x1]
%onnx::Conv_995[FLOAT, 512x512x1x1]
%onnx::Conv_998[FLOAT, 512x512x1x1]
%onnx::Conv_1001[FLOAT, 512x512x1x1]
%onnx::Conv_1004[FLOAT, 512x512x3x3]
%onnx::Conv_1007[FLOAT, 512x512x3x3]
%onnx::Conv_1010[FLOAT, 512x512x1x1]
%onnx::Conv_1013[FLOAT, 512x512x1x1]
%onnx::Conv_1016[FLOAT, 512x512x1x1]
%onnx::Conv_1019[FLOAT, 512x512x1x1]
%onnx::Conv_1022[FLOAT, 512x512x3x3]
%onnx::Conv_1025[FLOAT, 512x512x3x3]
%onnx::Conv_1028[FLOAT, 512x512x1x1]
%onnx::Conv_1031[FLOAT, 512x512x1x1]
) {
%onnx::Conv_1032 = Identity(%onnx::Conv_981)
%onnx::Conv_1029 = Identity(%onnx::Conv_981)
%onnx::Conv_1026 = Identity(%onnx::Conv_981)
%onnx::Conv_1023 = Identity(%onnx::Conv_981)
%onnx::Conv_1020 = Identity(%onnx::Conv_981)
%onnx::Conv_1017 = Identity(%onnx::Conv_981)
%onnx::Conv_1014 = Identity(%onnx::Conv_981)
%onnx::Conv_1011 = Identity(%onnx::Conv_981)
%onnx::Conv_1008 = Identity(%onnx::Conv_981)
%onnx::Conv_1005 = Identity(%onnx::Conv_981)
%onnx::Conv_1002 = Identity(%onnx::Conv_981)
%onnx::Conv_999 = Identity(%onnx::Conv_981)
%onnx::Conv_996 = Identity(%onnx::Conv_981)
%onnx::Conv_993 = Identity(%onnx::Conv_981)
%onnx::Conv_990 = Identity(%onnx::Conv_981)
%onnx::Conv_987 = Identity(%onnx::Conv_981)
%onnx::Conv_984 = Identity(%onnx::Conv_981)
%onnx::Conv_978 = Identity(%onnx::Conv_927)
%onnx::Conv_975 = Identity(%onnx::Conv_927)
%onnx::Conv_972 = Identity(%onnx::Conv_927)
%onnx::Conv_969 = Identity(%onnx::Conv_927)
%onnx::Conv_966 = Identity(%onnx::Conv_927)
%onnx::Conv_963 = Identity(%onnx::Conv_927)
%onnx::Conv_960 = Identity(%onnx::Conv_927)
%onnx::Conv_957 = Identity(%onnx::Conv_927)
%onnx::Conv_954 = Identity(%onnx::Conv_927)
%onnx::Conv_951 = Identity(%onnx::Conv_927)
%onnx::Conv_948 = Identity(%onnx::Conv_927)
%onnx::Conv_945 = Identity(%onnx::Conv_927)
%onnx::Conv_942 = Identity(%onnx::Conv_927)
%onnx::Conv_939 = Identity(%onnx::Conv_927)
%onnx::Conv_936 = Identity(%onnx::Conv_927)
%onnx::Conv_933 = Identity(%onnx::Conv_927)
%onnx::Conv_930 = Identity(%onnx::Conv_927)
%onnx::Conv_924 = Identity(%onnx::Conv_870)
%onnx::Conv_921 = Identity(%onnx::Conv_870)
%onnx::Conv_918 = Identity(%onnx::Conv_870)
%onnx::Conv_915 = Identity(%onnx::Conv_870)
%onnx::Conv_912 = Identity(%onnx::Conv_870)
%onnx::Conv_909 = Identity(%onnx::Conv_870)
%onnx::Conv_906 = Identity(%onnx::Conv_870)
%onnx::Conv_903 = Identity(%onnx::Conv_870)
%onnx::Conv_900 = Identity(%onnx::Conv_870)
%onnx::Conv_897 = Identity(%onnx::Conv_870)
%onnx::Conv_894 = Identity(%onnx::Conv_870)
%onnx::Conv_891 = Identity(%onnx::Conv_870)
%onnx::Conv_888 = Identity(%onnx::Conv_870)
%onnx::Conv_885 = Identity(%onnx::Conv_870)
%onnx::Conv_882 = Identity(%onnx::Conv_870)
%onnx::Conv_879 = Identity(%onnx::Conv_870)
%onnx::Conv_876 = Identity(%onnx::Conv_870)
%onnx::Conv_873 = Identity(%onnx::Conv_870)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_869, %onnx::Conv_870)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/Add_3_output_0, %/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_5_output_0 = Add(%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_4_output_0)
%/layers.1/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_5_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/Add_3_output_0, %/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_5_output_0 = Add(%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_4_output_0)
%/layers.2/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_5_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_908, %onnx::Conv_909)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_911, %onnx::Conv_912)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_914, %onnx::Conv_915)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_917, %onnx::Conv_918)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_920, %onnx::Conv_921)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/Add_3_output_0, %/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_4_output_0, %onnx::Conv_923, %onnx::Conv_924)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_5_output_0 = Add(%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_4_output_0)
%/layers.3/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_5_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_926, %onnx::Conv_927)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_929, %onnx::Conv_930)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_932, %onnx::Conv_933)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_935, %onnx::Conv_936)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_938, %onnx::Conv_939)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/Add_3_output_0, %/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_941, %onnx::Conv_942)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_5_output_0 = Add(%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_4_output_0)
%/layers.5/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_5_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_944, %onnx::Conv_945)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_947, %onnx::Conv_948)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_950, %onnx::Conv_951)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_953, %onnx::Conv_954)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_956, %onnx::Conv_957)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/Add_3_output_0, %/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_959, %onnx::Conv_960)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_5_output_0 = Add(%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_4_output_0)
%/layers.6/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_5_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_962, %onnx::Conv_963)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_965, %onnx::Conv_966)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_968, %onnx::Conv_969)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_974, %onnx::Conv_975)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/Add_3_output_0, %/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_4_output_0, %onnx::Conv_977, %onnx::Conv_978)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_5_output_0 = Add(%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_4_output_0)
%/layers.7/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_5_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/Add_3_output_0, %/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_5_output_0 = Add(%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_4_output_0)
%/layers.9/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_5_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/Add_3_output_0, %/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_5_output_0 = Add(%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_4_output_0)
%/layers.10/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_5_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_1025, %onnx::Conv_1026)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_1028, %onnx::Conv_1029)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/Add_3_output_0, %/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_4_output_0, %onnx::Conv_1031, %onnx::Conv_1032)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_5_output_0 = Add(%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_4_output_0)
%/layers.11/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_5_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/vertex_op.5/maxpool/MaxPool_output_0)
%867 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %867
}
|
val_accuracy
| 89.342946
| 6,617,835,520
| 22,421,642
|
{'zcp_epe_nas': 95.93024030104888, 'zcp_fisher': 8810.3798828125, 'zcp_flops': 105885368320.0, 'zcp_grad_norm': 1771.05078125, 'zcp_grasp': -67377.59375, 'zcp_jacov': -16.0545535227766, 'zcp_l2_norm': 1242.8134765625, 'zcp_nwot': 235.54554484986406, 'zcp_params': 22421642.0, 'zcp_plain': 0.071969725191593, 'zcp_snip': 13472.5517578125, 'zcp_synflow': 155.935515401995, 'zcp_zen': 109.66598510742188, 'zcp_val_accuracy': 0.9258813858032221}
| |
NASBench101_218342
|
NASBench101
|
218342
|
844c450615ff5efdf9e57a8f71fba8b0
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_923[FLOAT, 128x3x3x3]
%onnx::Conv_924[FLOAT, 128]
%onnx::Conv_926[FLOAT, 43x128x1x1]
%onnx::Conv_927[FLOAT, 43]
%onnx::Conv_929[FLOAT, 43x43x3x3]
%onnx::Conv_932[FLOAT, 43x128x1x1]
%onnx::Conv_935[FLOAT, 43x43x1x1]
%onnx::Conv_938[FLOAT, 42x42x3x3]
%onnx::Conv_939[FLOAT, 42]
%onnx::Conv_941[FLOAT, 42x42x3x3]
%onnx::Conv_944[FLOAT, 43x128x1x1]
%onnx::Conv_947[FLOAT, 43x43x3x3]
%onnx::Conv_950[FLOAT, 43x128x1x1]
%onnx::Conv_953[FLOAT, 43x43x1x1]
%onnx::Conv_956[FLOAT, 42x42x3x3]
%onnx::Conv_959[FLOAT, 42x42x3x3]
%onnx::Conv_962[FLOAT, 43x128x1x1]
%onnx::Conv_965[FLOAT, 43x43x3x3]
%onnx::Conv_968[FLOAT, 43x128x1x1]
%onnx::Conv_971[FLOAT, 43x43x1x1]
%onnx::Conv_974[FLOAT, 42x42x3x3]
%onnx::Conv_977[FLOAT, 42x42x3x3]
%onnx::Conv_980[FLOAT, 86x128x1x1]
%onnx::Conv_981[FLOAT, 86]
%onnx::Conv_983[FLOAT, 86x86x3x3]
%onnx::Conv_986[FLOAT, 85x128x1x1]
%onnx::Conv_987[FLOAT, 85]
%onnx::Conv_989[FLOAT, 85x85x1x1]
%onnx::Conv_992[FLOAT, 85x85x3x3]
%onnx::Conv_995[FLOAT, 85x85x3x3]
%onnx::Conv_998[FLOAT, 86x256x1x1]
%onnx::Conv_1001[FLOAT, 86x86x3x3]
%onnx::Conv_1004[FLOAT, 85x256x1x1]
%onnx::Conv_1007[FLOAT, 85x85x1x1]
%onnx::Conv_1010[FLOAT, 85x85x3x3]
%onnx::Conv_1013[FLOAT, 85x85x3x3]
%onnx::Conv_1016[FLOAT, 86x256x1x1]
%onnx::Conv_1019[FLOAT, 86x86x3x3]
%onnx::Conv_1022[FLOAT, 85x256x1x1]
%onnx::Conv_1025[FLOAT, 85x85x1x1]
%onnx::Conv_1028[FLOAT, 85x85x3x3]
%onnx::Conv_1031[FLOAT, 85x85x3x3]
%onnx::Conv_1034[FLOAT, 171x256x1x1]
%onnx::Conv_1035[FLOAT, 171]
%onnx::Conv_1037[FLOAT, 171x171x3x3]
%onnx::Conv_1040[FLOAT, 171x256x1x1]
%onnx::Conv_1043[FLOAT, 171x171x1x1]
%onnx::Conv_1046[FLOAT, 170x170x3x3]
%onnx::Conv_1047[FLOAT, 170]
%onnx::Conv_1049[FLOAT, 170x170x3x3]
%onnx::Conv_1052[FLOAT, 171x512x1x1]
%onnx::Conv_1055[FLOAT, 171x171x3x3]
%onnx::Conv_1058[FLOAT, 171x512x1x1]
%onnx::Conv_1061[FLOAT, 171x171x1x1]
%onnx::Conv_1064[FLOAT, 170x170x3x3]
%onnx::Conv_1067[FLOAT, 170x170x3x3]
%onnx::Conv_1070[FLOAT, 171x512x1x1]
%onnx::Conv_1073[FLOAT, 171x171x3x3]
%onnx::Conv_1076[FLOAT, 171x512x1x1]
%onnx::Conv_1079[FLOAT, 171x171x1x1]
%onnx::Conv_1082[FLOAT, 170x170x3x3]
%onnx::Conv_1085[FLOAT, 170x170x3x3]
) {
%onnx::Conv_1086 = Identity(%onnx::Conv_1047)
%onnx::Conv_1083 = Identity(%onnx::Conv_1047)
%onnx::Conv_1080 = Identity(%onnx::Conv_1035)
%onnx::Conv_1077 = Identity(%onnx::Conv_1035)
%onnx::Conv_1074 = Identity(%onnx::Conv_1035)
%onnx::Conv_1071 = Identity(%onnx::Conv_1035)
%onnx::Conv_1068 = Identity(%onnx::Conv_1047)
%onnx::Conv_1065 = Identity(%onnx::Conv_1047)
%onnx::Conv_1062 = Identity(%onnx::Conv_1035)
%onnx::Conv_1059 = Identity(%onnx::Conv_1035)
%onnx::Conv_1056 = Identity(%onnx::Conv_1035)
%onnx::Conv_1053 = Identity(%onnx::Conv_1035)
%onnx::Conv_1050 = Identity(%onnx::Conv_1047)
%onnx::Conv_1044 = Identity(%onnx::Conv_1035)
%onnx::Conv_1041 = Identity(%onnx::Conv_1035)
%onnx::Conv_1038 = Identity(%onnx::Conv_1035)
%onnx::Conv_1032 = Identity(%onnx::Conv_987)
%onnx::Conv_1029 = Identity(%onnx::Conv_987)
%onnx::Conv_1026 = Identity(%onnx::Conv_987)
%onnx::Conv_1023 = Identity(%onnx::Conv_987)
%onnx::Conv_1020 = Identity(%onnx::Conv_981)
%onnx::Conv_1017 = Identity(%onnx::Conv_981)
%onnx::Conv_1014 = Identity(%onnx::Conv_987)
%onnx::Conv_1011 = Identity(%onnx::Conv_987)
%onnx::Conv_1008 = Identity(%onnx::Conv_987)
%onnx::Conv_1005 = Identity(%onnx::Conv_987)
%onnx::Conv_1002 = Identity(%onnx::Conv_981)
%onnx::Conv_999 = Identity(%onnx::Conv_981)
%onnx::Conv_996 = Identity(%onnx::Conv_987)
%onnx::Conv_993 = Identity(%onnx::Conv_987)
%onnx::Conv_990 = Identity(%onnx::Conv_987)
%onnx::Conv_984 = Identity(%onnx::Conv_981)
%onnx::Conv_978 = Identity(%onnx::Conv_939)
%onnx::Conv_975 = Identity(%onnx::Conv_939)
%onnx::Conv_972 = Identity(%onnx::Conv_927)
%onnx::Conv_969 = Identity(%onnx::Conv_927)
%onnx::Conv_966 = Identity(%onnx::Conv_927)
%onnx::Conv_963 = Identity(%onnx::Conv_927)
%onnx::Conv_960 = Identity(%onnx::Conv_939)
%onnx::Conv_957 = Identity(%onnx::Conv_939)
%onnx::Conv_954 = Identity(%onnx::Conv_927)
%onnx::Conv_951 = Identity(%onnx::Conv_927)
%onnx::Conv_948 = Identity(%onnx::Conv_927)
%onnx::Conv_945 = Identity(%onnx::Conv_927)
%onnx::Conv_942 = Identity(%onnx::Conv_939)
%onnx::Conv_936 = Identity(%onnx::Conv_927)
%onnx::Conv_933 = Identity(%onnx::Conv_927)
%onnx::Conv_930 = Identity(%onnx::Conv_927)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_923, %onnx::Conv_924)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_926, %onnx::Conv_927)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_929, %onnx::Conv_930)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_932, %onnx::Conv_933)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_935, %onnx::Conv_936)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.1/Slice_output_0 = Slice(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0, %/layers.1/Constant_4_output_0, %/layers.1/Constant_2_output_0, %/layers.1/Constant_5_output_0)
%/layers.1/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/Slice_output_0, %/layers.1/Constant_6_output_0)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_938, %onnx::Conv_939)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_7_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0)
%/layers.1/Constant_8_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_8_output_0)
%/layers.1/Add_5_output_0 = Add(%/layers.1/Add_4_output_0, %/layers.1/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_5_output_0, %onnx::Conv_941, %onnx::Conv_942)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_944, %onnx::Conv_945)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_947, %onnx::Conv_948)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_950, %onnx::Conv_951)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_953, %onnx::Conv_954)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.2/Slice_output_0 = Slice(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0, %/layers.2/Constant_4_output_0, %/layers.2/Constant_2_output_0, %/layers.2/Constant_5_output_0)
%/layers.2/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/Slice_output_0, %/layers.2/Constant_6_output_0)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_956, %onnx::Conv_957)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_7_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0)
%/layers.2/Constant_8_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_8_output_0)
%/layers.2/Add_5_output_0 = Add(%/layers.2/Add_4_output_0, %/layers.2/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_5_output_0, %onnx::Conv_959, %onnx::Conv_960)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_962, %onnx::Conv_963)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_965, %onnx::Conv_966)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_968, %onnx::Conv_969)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.3/Slice_output_0 = Slice(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0, %/layers.3/Constant_4_output_0, %/layers.3/Constant_2_output_0, %/layers.3/Constant_5_output_0)
%/layers.3/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/Slice_output_0, %/layers.3/Constant_6_output_0)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_974, %onnx::Conv_975)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_7_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0)
%/layers.3/Constant_8_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_8_output_0)
%/layers.3/Add_5_output_0 = Add(%/layers.3/Add_4_output_0, %/layers.3/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_5_output_0, %onnx::Conv_977, %onnx::Conv_978)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.5/Slice_output_0 = Slice(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0, %/layers.5/Constant_4_output_0, %/layers.5/Constant_2_output_0, %/layers.5/Constant_5_output_0)
%/layers.5/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/Slice_output_0, %/layers.5/Constant_6_output_0)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_7_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0)
%/layers.5/Constant_8_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_8_output_0)
%/layers.5/Add_5_output_0 = Add(%/layers.5/Add_4_output_0, %/layers.5/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_5_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.6/Slice_output_0 = Slice(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0, %/layers.6/Constant_4_output_0, %/layers.6/Constant_2_output_0, %/layers.6/Constant_5_output_0)
%/layers.6/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/Slice_output_0, %/layers.6/Constant_6_output_0)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_7_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0)
%/layers.6/Constant_8_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_8_output_0)
%/layers.6/Add_5_output_0 = Add(%/layers.6/Add_4_output_0, %/layers.6/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_5_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_1025, %onnx::Conv_1026)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.7/Slice_output_0 = Slice(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0, %/layers.7/Constant_4_output_0, %/layers.7/Constant_2_output_0, %/layers.7/Constant_5_output_0)
%/layers.7/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/Slice_output_0, %/layers.7/Constant_6_output_0)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_1028, %onnx::Conv_1029)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_7_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0)
%/layers.7/Constant_8_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_8_output_0)
%/layers.7/Add_5_output_0 = Add(%/layers.7/Add_4_output_0, %/layers.7/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_5_output_0, %onnx::Conv_1031, %onnx::Conv_1032)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1034, %onnx::Conv_1035)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_1037, %onnx::Conv_1038)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1040, %onnx::Conv_1041)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_1043, %onnx::Conv_1044)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.9/Slice_output_0 = Slice(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0, %/layers.9/Constant_4_output_0, %/layers.9/Constant_2_output_0, %/layers.9/Constant_5_output_0)
%/layers.9/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/Slice_output_0, %/layers.9/Constant_6_output_0)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_1046, %onnx::Conv_1047)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_7_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0)
%/layers.9/Constant_8_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_8_output_0)
%/layers.9/Add_5_output_0 = Add(%/layers.9/Add_4_output_0, %/layers.9/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_5_output_0, %onnx::Conv_1049, %onnx::Conv_1050)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_1052, %onnx::Conv_1053)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_1055, %onnx::Conv_1056)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_1058, %onnx::Conv_1059)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_1061, %onnx::Conv_1062)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.10/Slice_output_0 = Slice(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0, %/layers.10/Constant_4_output_0, %/layers.10/Constant_2_output_0, %/layers.10/Constant_5_output_0)
%/layers.10/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/Slice_output_0, %/layers.10/Constant_6_output_0)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_1064, %onnx::Conv_1065)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_7_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0)
%/layers.10/Constant_8_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_8_output_0)
%/layers.10/Add_5_output_0 = Add(%/layers.10/Add_4_output_0, %/layers.10/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_5_output_0, %onnx::Conv_1067, %onnx::Conv_1068)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1070, %onnx::Conv_1071)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_1073, %onnx::Conv_1074)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1076, %onnx::Conv_1077)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_1079, %onnx::Conv_1080)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.11/Slice_output_0 = Slice(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0, %/layers.11/Constant_4_output_0, %/layers.11/Constant_2_output_0, %/layers.11/Constant_5_output_0)
%/layers.11/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/Slice_output_0, %/layers.11/Constant_6_output_0)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_1082, %onnx::Conv_1083)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_7_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0)
%/layers.11/Constant_8_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_8_output_0)
%/layers.11/Add_5_output_0 = Add(%/layers.11/Add_4_output_0, %/layers.11/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_5_output_0, %onnx::Conv_1085, %onnx::Conv_1086)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%921 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %921
}
|
val_accuracy
| 93.739986
| 1,132,672,256
| 3,799,891
|
{'zcp_epe_nas': 96.90126640985393, 'zcp_fisher': 5.518043041229248, 'zcp_flops': 18122756096.0, 'zcp_grad_norm': 51.375911712646484, 'zcp_grasp': -1.92718505859375, 'zcp_jacov': -16.060708119970357, 'zcp_l2_norm': 884.1173706054688, 'zcp_nwot': 218.32926013562644, 'zcp_params': 3799891.0, 'zcp_plain': 0.032842826098203, 'zcp_snip': 274.20611572265625, 'zcp_synflow': 125.34780814125418, 'zcp_zen': 98.20671844482422, 'zcp_val_accuracy': 0.9286859035491941}
| |
NASBench101_293356
|
NASBench101
|
293356
|
b19c8a2b85dc83590402cd9ad666fd0b
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_752[FLOAT, 128x3x3x3]
%onnx::Conv_753[FLOAT, 128]
%onnx::Conv_755[FLOAT, 128x128x1x1]
%onnx::Conv_758[FLOAT, 128x128x1x1]
%onnx::Conv_761[FLOAT, 128x128x1x1]
%onnx::Conv_764[FLOAT, 128x128x1x1]
%onnx::Conv_767[FLOAT, 128x128x1x1]
%onnx::Conv_770[FLOAT, 128x128x1x1]
%onnx::Conv_773[FLOAT, 128x128x1x1]
%onnx::Conv_776[FLOAT, 128x128x1x1]
%onnx::Conv_779[FLOAT, 128x128x1x1]
%onnx::Conv_782[FLOAT, 128x128x1x1]
%onnx::Conv_785[FLOAT, 128x128x1x1]
%onnx::Conv_788[FLOAT, 128x128x1x1]
%onnx::Conv_791[FLOAT, 128x128x1x1]
%onnx::Conv_794[FLOAT, 128x128x1x1]
%onnx::Conv_797[FLOAT, 128x128x1x1]
%onnx::Conv_800[FLOAT, 256x128x1x1]
%onnx::Conv_801[FLOAT, 256]
%onnx::Conv_803[FLOAT, 256x128x1x1]
%onnx::Conv_806[FLOAT, 256x256x1x1]
%onnx::Conv_809[FLOAT, 256x256x1x1]
%onnx::Conv_812[FLOAT, 256x128x1x1]
%onnx::Conv_815[FLOAT, 256x256x1x1]
%onnx::Conv_818[FLOAT, 256x256x1x1]
%onnx::Conv_821[FLOAT, 256x256x1x1]
%onnx::Conv_824[FLOAT, 256x256x1x1]
%onnx::Conv_827[FLOAT, 256x256x1x1]
%onnx::Conv_830[FLOAT, 256x256x1x1]
%onnx::Conv_833[FLOAT, 256x256x1x1]
%onnx::Conv_836[FLOAT, 256x256x1x1]
%onnx::Conv_839[FLOAT, 256x256x1x1]
%onnx::Conv_842[FLOAT, 256x256x1x1]
%onnx::Conv_845[FLOAT, 512x256x1x1]
%onnx::Conv_846[FLOAT, 512]
%onnx::Conv_848[FLOAT, 512x256x1x1]
%onnx::Conv_851[FLOAT, 512x512x1x1]
%onnx::Conv_854[FLOAT, 512x512x1x1]
%onnx::Conv_857[FLOAT, 512x256x1x1]
%onnx::Conv_860[FLOAT, 512x512x1x1]
%onnx::Conv_863[FLOAT, 512x512x1x1]
%onnx::Conv_866[FLOAT, 512x512x1x1]
%onnx::Conv_869[FLOAT, 512x512x1x1]
%onnx::Conv_872[FLOAT, 512x512x1x1]
%onnx::Conv_875[FLOAT, 512x512x1x1]
%onnx::Conv_878[FLOAT, 512x512x1x1]
%onnx::Conv_881[FLOAT, 512x512x1x1]
%onnx::Conv_884[FLOAT, 512x512x1x1]
%onnx::Conv_887[FLOAT, 512x512x1x1]
) {
%onnx::Conv_888 = Identity(%onnx::Conv_846)
%onnx::Conv_885 = Identity(%onnx::Conv_846)
%onnx::Conv_882 = Identity(%onnx::Conv_846)
%onnx::Conv_879 = Identity(%onnx::Conv_846)
%onnx::Conv_876 = Identity(%onnx::Conv_846)
%onnx::Conv_873 = Identity(%onnx::Conv_846)
%onnx::Conv_870 = Identity(%onnx::Conv_846)
%onnx::Conv_867 = Identity(%onnx::Conv_846)
%onnx::Conv_864 = Identity(%onnx::Conv_846)
%onnx::Conv_861 = Identity(%onnx::Conv_846)
%onnx::Conv_858 = Identity(%onnx::Conv_846)
%onnx::Conv_855 = Identity(%onnx::Conv_846)
%onnx::Conv_852 = Identity(%onnx::Conv_846)
%onnx::Conv_849 = Identity(%onnx::Conv_846)
%onnx::Conv_843 = Identity(%onnx::Conv_801)
%onnx::Conv_840 = Identity(%onnx::Conv_801)
%onnx::Conv_837 = Identity(%onnx::Conv_801)
%onnx::Conv_834 = Identity(%onnx::Conv_801)
%onnx::Conv_831 = Identity(%onnx::Conv_801)
%onnx::Conv_828 = Identity(%onnx::Conv_801)
%onnx::Conv_825 = Identity(%onnx::Conv_801)
%onnx::Conv_822 = Identity(%onnx::Conv_801)
%onnx::Conv_819 = Identity(%onnx::Conv_801)
%onnx::Conv_816 = Identity(%onnx::Conv_801)
%onnx::Conv_813 = Identity(%onnx::Conv_801)
%onnx::Conv_810 = Identity(%onnx::Conv_801)
%onnx::Conv_807 = Identity(%onnx::Conv_801)
%onnx::Conv_804 = Identity(%onnx::Conv_801)
%onnx::Conv_798 = Identity(%onnx::Conv_753)
%onnx::Conv_795 = Identity(%onnx::Conv_753)
%onnx::Conv_792 = Identity(%onnx::Conv_753)
%onnx::Conv_789 = Identity(%onnx::Conv_753)
%onnx::Conv_786 = Identity(%onnx::Conv_753)
%onnx::Conv_783 = Identity(%onnx::Conv_753)
%onnx::Conv_780 = Identity(%onnx::Conv_753)
%onnx::Conv_777 = Identity(%onnx::Conv_753)
%onnx::Conv_774 = Identity(%onnx::Conv_753)
%onnx::Conv_771 = Identity(%onnx::Conv_753)
%onnx::Conv_768 = Identity(%onnx::Conv_753)
%onnx::Conv_765 = Identity(%onnx::Conv_753)
%onnx::Conv_762 = Identity(%onnx::Conv_753)
%onnx::Conv_759 = Identity(%onnx::Conv_753)
%onnx::Conv_756 = Identity(%onnx::Conv_753)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_752, %onnx::Conv_753)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_755, %onnx::Conv_756)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_758, %onnx::Conv_759)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_761, %onnx::Conv_762)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_764, %onnx::Conv_765)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0)
%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_767, %onnx::Conv_768)
%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.5/maxpool/MaxPool_output_0, %/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_770, %onnx::Conv_771)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_773, %onnx::Conv_774)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_776, %onnx::Conv_777)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_779, %onnx::Conv_780)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0)
%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_782, %onnx::Conv_783)
%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.5/maxpool/MaxPool_output_0, %/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_785, %onnx::Conv_786)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_788, %onnx::Conv_789)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0)
%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_797, %onnx::Conv_798)
%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.5/maxpool/MaxPool_output_0, %/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Add_4_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_800, %onnx::Conv_801)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_803, %onnx::Conv_804)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_806, %onnx::Conv_807)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_809, %onnx::Conv_810)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0)
%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_812, %onnx::Conv_813)
%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.5/maxpool/MaxPool_output_0, %/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_815, %onnx::Conv_816)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_818, %onnx::Conv_819)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_821, %onnx::Conv_822)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_824, %onnx::Conv_825)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0)
%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_827, %onnx::Conv_828)
%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.5/maxpool/MaxPool_output_0, %/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_830, %onnx::Conv_831)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_833, %onnx::Conv_834)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_836, %onnx::Conv_837)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_839, %onnx::Conv_840)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0)
%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_842, %onnx::Conv_843)
%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.5/maxpool/MaxPool_output_0, %/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Add_4_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_845, %onnx::Conv_846)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_848, %onnx::Conv_849)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_851, %onnx::Conv_852)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_854, %onnx::Conv_855)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0)
%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_857, %onnx::Conv_858)
%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.5/maxpool/MaxPool_output_0, %/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0)
%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.5/maxpool/MaxPool_output_0, %/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0)
%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.5/maxpool/MaxPool_output_0, %/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Add_4_output_0)
%750 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %750
}
|
val_accuracy
| 89.933896
| 1,444,947,968
| 4,705,162
|
{'zcp_epe_nas': 111.47684384755206, 'zcp_fisher': 34.66154098510742, 'zcp_flops': 23119167488.0, 'zcp_grad_norm': 129.70455932617188, 'zcp_grasp': -35.74951171875, 'zcp_jacov': -16.06463291773293, 'zcp_l2_norm': 1014.8059692382812, 'zcp_nwot': 232.40421779594672, 'zcp_params': 4705162.0, 'zcp_plain': 0.018470745533704, 'zcp_snip': 937.3181762695312, 'zcp_synflow': 89.9674543499744, 'zcp_zen': 90.89616394042969, 'zcp_val_accuracy': 0.914863765239715}
| |
NASBench101_254144
|
NASBench101
|
254144
|
99db138bab3e95728f04ae878fc7ff31
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_968[FLOAT, 128x3x3x3]
%onnx::Conv_969[FLOAT, 128]
%onnx::Conv_971[FLOAT, 64x128x1x1]
%onnx::Conv_972[FLOAT, 64]
%onnx::Conv_974[FLOAT, 64x64x3x3]
%onnx::Conv_977[FLOAT, 64x128x1x1]
%onnx::Conv_980[FLOAT, 64x64x3x3]
%onnx::Conv_983[FLOAT, 64x64x3x3]
%onnx::Conv_986[FLOAT, 64x64x1x1]
%onnx::Conv_989[FLOAT, 128x128x1x1]
%onnx::Conv_992[FLOAT, 64x128x1x1]
%onnx::Conv_995[FLOAT, 64x64x3x3]
%onnx::Conv_998[FLOAT, 64x128x1x1]
%onnx::Conv_1001[FLOAT, 64x64x3x3]
%onnx::Conv_1004[FLOAT, 64x64x3x3]
%onnx::Conv_1007[FLOAT, 64x64x1x1]
%onnx::Conv_1010[FLOAT, 128x128x1x1]
%onnx::Conv_1013[FLOAT, 64x128x1x1]
%onnx::Conv_1016[FLOAT, 64x64x3x3]
%onnx::Conv_1019[FLOAT, 64x128x1x1]
%onnx::Conv_1022[FLOAT, 64x64x3x3]
%onnx::Conv_1025[FLOAT, 64x64x3x3]
%onnx::Conv_1028[FLOAT, 64x64x1x1]
%onnx::Conv_1031[FLOAT, 128x128x1x1]
%onnx::Conv_1034[FLOAT, 128x128x1x1]
%onnx::Conv_1037[FLOAT, 128x128x3x3]
%onnx::Conv_1040[FLOAT, 128x128x1x1]
%onnx::Conv_1043[FLOAT, 128x128x3x3]
%onnx::Conv_1046[FLOAT, 128x128x3x3]
%onnx::Conv_1049[FLOAT, 128x128x1x1]
%onnx::Conv_1052[FLOAT, 256x128x1x1]
%onnx::Conv_1053[FLOAT, 256]
%onnx::Conv_1055[FLOAT, 128x256x1x1]
%onnx::Conv_1058[FLOAT, 128x128x3x3]
%onnx::Conv_1061[FLOAT, 128x256x1x1]
%onnx::Conv_1064[FLOAT, 128x128x3x3]
%onnx::Conv_1067[FLOAT, 128x128x3x3]
%onnx::Conv_1070[FLOAT, 128x128x1x1]
%onnx::Conv_1073[FLOAT, 256x256x1x1]
%onnx::Conv_1076[FLOAT, 128x256x1x1]
%onnx::Conv_1079[FLOAT, 128x128x3x3]
%onnx::Conv_1082[FLOAT, 128x256x1x1]
%onnx::Conv_1085[FLOAT, 128x128x3x3]
%onnx::Conv_1088[FLOAT, 128x128x3x3]
%onnx::Conv_1091[FLOAT, 128x128x1x1]
%onnx::Conv_1094[FLOAT, 256x256x1x1]
%onnx::Conv_1097[FLOAT, 256x256x1x1]
%onnx::Conv_1100[FLOAT, 256x256x3x3]
%onnx::Conv_1103[FLOAT, 256x256x1x1]
%onnx::Conv_1106[FLOAT, 256x256x3x3]
%onnx::Conv_1109[FLOAT, 256x256x3x3]
%onnx::Conv_1112[FLOAT, 256x256x1x1]
%onnx::Conv_1115[FLOAT, 512x256x1x1]
%onnx::Conv_1116[FLOAT, 512]
%onnx::Conv_1118[FLOAT, 256x512x1x1]
%onnx::Conv_1121[FLOAT, 256x256x3x3]
%onnx::Conv_1124[FLOAT, 256x512x1x1]
%onnx::Conv_1127[FLOAT, 256x256x3x3]
%onnx::Conv_1130[FLOAT, 256x256x3x3]
%onnx::Conv_1133[FLOAT, 256x256x1x1]
%onnx::Conv_1136[FLOAT, 512x512x1x1]
%onnx::Conv_1139[FLOAT, 256x512x1x1]
%onnx::Conv_1142[FLOAT, 256x256x3x3]
%onnx::Conv_1145[FLOAT, 256x512x1x1]
%onnx::Conv_1148[FLOAT, 256x256x3x3]
%onnx::Conv_1151[FLOAT, 256x256x3x3]
%onnx::Conv_1154[FLOAT, 256x256x1x1]
%onnx::Conv_1157[FLOAT, 512x512x1x1]
) {
%onnx::Conv_1158 = Identity(%onnx::Conv_1116)
%onnx::Conv_1155 = Identity(%onnx::Conv_1053)
%onnx::Conv_1152 = Identity(%onnx::Conv_1053)
%onnx::Conv_1149 = Identity(%onnx::Conv_1053)
%onnx::Conv_1146 = Identity(%onnx::Conv_1053)
%onnx::Conv_1143 = Identity(%onnx::Conv_1053)
%onnx::Conv_1140 = Identity(%onnx::Conv_1053)
%onnx::Conv_1137 = Identity(%onnx::Conv_1116)
%onnx::Conv_1134 = Identity(%onnx::Conv_1053)
%onnx::Conv_1131 = Identity(%onnx::Conv_1053)
%onnx::Conv_1128 = Identity(%onnx::Conv_1053)
%onnx::Conv_1125 = Identity(%onnx::Conv_1053)
%onnx::Conv_1122 = Identity(%onnx::Conv_1053)
%onnx::Conv_1119 = Identity(%onnx::Conv_1053)
%onnx::Conv_1113 = Identity(%onnx::Conv_1053)
%onnx::Conv_1110 = Identity(%onnx::Conv_1053)
%onnx::Conv_1107 = Identity(%onnx::Conv_1053)
%onnx::Conv_1104 = Identity(%onnx::Conv_1053)
%onnx::Conv_1101 = Identity(%onnx::Conv_1053)
%onnx::Conv_1098 = Identity(%onnx::Conv_1053)
%onnx::Conv_1095 = Identity(%onnx::Conv_1053)
%onnx::Conv_1092 = Identity(%onnx::Conv_969)
%onnx::Conv_1089 = Identity(%onnx::Conv_969)
%onnx::Conv_1086 = Identity(%onnx::Conv_969)
%onnx::Conv_1083 = Identity(%onnx::Conv_969)
%onnx::Conv_1080 = Identity(%onnx::Conv_969)
%onnx::Conv_1077 = Identity(%onnx::Conv_969)
%onnx::Conv_1074 = Identity(%onnx::Conv_1053)
%onnx::Conv_1071 = Identity(%onnx::Conv_969)
%onnx::Conv_1068 = Identity(%onnx::Conv_969)
%onnx::Conv_1065 = Identity(%onnx::Conv_969)
%onnx::Conv_1062 = Identity(%onnx::Conv_969)
%onnx::Conv_1059 = Identity(%onnx::Conv_969)
%onnx::Conv_1056 = Identity(%onnx::Conv_969)
%onnx::Conv_1050 = Identity(%onnx::Conv_969)
%onnx::Conv_1047 = Identity(%onnx::Conv_969)
%onnx::Conv_1044 = Identity(%onnx::Conv_969)
%onnx::Conv_1041 = Identity(%onnx::Conv_969)
%onnx::Conv_1038 = Identity(%onnx::Conv_969)
%onnx::Conv_1035 = Identity(%onnx::Conv_969)
%onnx::Conv_1032 = Identity(%onnx::Conv_969)
%onnx::Conv_1029 = Identity(%onnx::Conv_972)
%onnx::Conv_1026 = Identity(%onnx::Conv_972)
%onnx::Conv_1023 = Identity(%onnx::Conv_972)
%onnx::Conv_1020 = Identity(%onnx::Conv_972)
%onnx::Conv_1017 = Identity(%onnx::Conv_972)
%onnx::Conv_1014 = Identity(%onnx::Conv_972)
%onnx::Conv_1011 = Identity(%onnx::Conv_969)
%onnx::Conv_1008 = Identity(%onnx::Conv_972)
%onnx::Conv_1005 = Identity(%onnx::Conv_972)
%onnx::Conv_1002 = Identity(%onnx::Conv_972)
%onnx::Conv_999 = Identity(%onnx::Conv_972)
%onnx::Conv_996 = Identity(%onnx::Conv_972)
%onnx::Conv_993 = Identity(%onnx::Conv_972)
%onnx::Conv_990 = Identity(%onnx::Conv_969)
%onnx::Conv_987 = Identity(%onnx::Conv_972)
%onnx::Conv_984 = Identity(%onnx::Conv_972)
%onnx::Conv_981 = Identity(%onnx::Conv_972)
%onnx::Conv_978 = Identity(%onnx::Conv_972)
%onnx::Conv_975 = Identity(%onnx::Conv_972)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_968, %onnx::Conv_969)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_974, %onnx::Conv_975)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_977, %onnx::Conv_978)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/Add_3_output_0, %/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_4_output_0)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.4/maxpool/MaxPool_output_0, %/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_5_output_0 = Add(%/layers.1/Concat_output_0, %/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_5_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_5_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/Add_3_output_0, %/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_4_output_0)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.4/maxpool/MaxPool_output_0, %/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_5_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_5_output_0 = Add(%/layers.2/Concat_output_0, %/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_5_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_5_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_1025, %onnx::Conv_1026)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/Add_3_output_0, %/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_4_output_0)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_1028, %onnx::Conv_1029)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.4/maxpool/MaxPool_output_0, %/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_5_output_0, %onnx::Conv_1031, %onnx::Conv_1032)
%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_5_output_0 = Add(%/layers.3/Concat_output_0, %/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Add_5_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1034, %onnx::Conv_1035)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_1037, %onnx::Conv_1038)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1040, %onnx::Conv_1041)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_1043, %onnx::Conv_1044)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_1046, %onnx::Conv_1047)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/Add_3_output_0, %/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_4_output_0)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_1049, %onnx::Conv_1050)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.4/maxpool/MaxPool_output_0, %/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1052, %onnx::Conv_1053)
%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_5_output_0 = Add(%/layers.5/Concat_output_0, %/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_5_output_0, %onnx::Conv_1055, %onnx::Conv_1056)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_1058, %onnx::Conv_1059)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_5_output_0, %onnx::Conv_1061, %onnx::Conv_1062)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_1064, %onnx::Conv_1065)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_1067, %onnx::Conv_1068)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/Add_3_output_0, %/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_4_output_0)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_1070, %onnx::Conv_1071)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.4/maxpool/MaxPool_output_0, %/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_5_output_0, %onnx::Conv_1073, %onnx::Conv_1074)
%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_5_output_0 = Add(%/layers.6/Concat_output_0, %/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_5_output_0, %onnx::Conv_1076, %onnx::Conv_1077)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_1079, %onnx::Conv_1080)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_5_output_0, %onnx::Conv_1082, %onnx::Conv_1083)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_1085, %onnx::Conv_1086)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_1088, %onnx::Conv_1089)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/Add_3_output_0, %/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_4_output_0)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_1091, %onnx::Conv_1092)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.4/maxpool/MaxPool_output_0, %/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_5_output_0, %onnx::Conv_1094, %onnx::Conv_1095)
%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_5_output_0 = Add(%/layers.7/Concat_output_0, %/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Add_5_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1097, %onnx::Conv_1098)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_1100, %onnx::Conv_1101)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1103, %onnx::Conv_1104)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_1106, %onnx::Conv_1107)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_1109, %onnx::Conv_1110)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/Add_3_output_0, %/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_4_output_0)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_1112, %onnx::Conv_1113)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.4/maxpool/MaxPool_output_0, %/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1115, %onnx::Conv_1116)
%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_5_output_0 = Add(%/layers.9/Concat_output_0, %/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_5_output_0, %onnx::Conv_1118, %onnx::Conv_1119)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_1121, %onnx::Conv_1122)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_5_output_0, %onnx::Conv_1124, %onnx::Conv_1125)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_1127, %onnx::Conv_1128)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_1130, %onnx::Conv_1131)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/Add_3_output_0, %/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_4_output_0)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_1133, %onnx::Conv_1134)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.4/maxpool/MaxPool_output_0, %/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_5_output_0, %onnx::Conv_1136, %onnx::Conv_1137)
%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_5_output_0 = Add(%/layers.10/Concat_output_0, %/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_5_output_0, %onnx::Conv_1139, %onnx::Conv_1140)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_1142, %onnx::Conv_1143)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_5_output_0, %onnx::Conv_1145, %onnx::Conv_1146)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_1148, %onnx::Conv_1149)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_1151, %onnx::Conv_1152)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/Add_3_output_0, %/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_4_output_0)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_1154, %onnx::Conv_1155)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.4/maxpool/MaxPool_output_0, %/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_5_output_0, %onnx::Conv_1157, %onnx::Conv_1158)
%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_5_output_0 = Add(%/layers.11/Concat_output_0, %/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Add_5_output_0)
%966 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %966
}
|
val_accuracy
| 92.778444
| 2,680,956,928
| 8,992,394
|
{'zcp_epe_nas': 91.35670127097544, 'zcp_fisher': 61.65487289428711, 'zcp_flops': 42895310848.0, 'zcp_grad_norm': 168.5249481201172, 'zcp_grasp': -39.526611328125, 'zcp_jacov': -16.05896726314154, 'zcp_l2_norm': 1190.2396240234375, 'zcp_nwot': 229.0709505880269, 'zcp_params': 8992394.0, 'zcp_plain': 0.12053034454584101, 'zcp_snip': 1114.98779296875, 'zcp_synflow': 118.15265839068255, 'zcp_zen': 127.8485107421875, 'zcp_val_accuracy': 0.8958333134651181}
| |
NASBench101_307937
|
NASBench101
|
307937
|
ba50a794fe80d49cde129a8d2b102e66
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_920[FLOAT, 128x3x3x3]
%onnx::Conv_921[FLOAT, 128]
%onnx::Conv_923[FLOAT, 43x128x1x1]
%onnx::Conv_924[FLOAT, 43]
%onnx::Conv_926[FLOAT, 43x43x1x1]
%onnx::Conv_929[FLOAT, 43x128x1x1]
%onnx::Conv_932[FLOAT, 43x43x1x1]
%onnx::Conv_935[FLOAT, 42x42x3x3]
%onnx::Conv_936[FLOAT, 42]
%onnx::Conv_938[FLOAT, 42x42x3x3]
%onnx::Conv_941[FLOAT, 43x128x1x1]
%onnx::Conv_944[FLOAT, 43x43x1x1]
%onnx::Conv_947[FLOAT, 43x128x1x1]
%onnx::Conv_950[FLOAT, 43x43x1x1]
%onnx::Conv_953[FLOAT, 42x42x3x3]
%onnx::Conv_956[FLOAT, 42x42x3x3]
%onnx::Conv_959[FLOAT, 43x128x1x1]
%onnx::Conv_962[FLOAT, 43x43x1x1]
%onnx::Conv_965[FLOAT, 43x128x1x1]
%onnx::Conv_968[FLOAT, 43x43x1x1]
%onnx::Conv_971[FLOAT, 42x42x3x3]
%onnx::Conv_974[FLOAT, 42x42x3x3]
%onnx::Conv_977[FLOAT, 86x128x1x1]
%onnx::Conv_978[FLOAT, 86]
%onnx::Conv_980[FLOAT, 86x86x1x1]
%onnx::Conv_983[FLOAT, 85x128x1x1]
%onnx::Conv_984[FLOAT, 85]
%onnx::Conv_986[FLOAT, 85x85x1x1]
%onnx::Conv_989[FLOAT, 85x85x3x3]
%onnx::Conv_992[FLOAT, 85x85x3x3]
%onnx::Conv_995[FLOAT, 86x256x1x1]
%onnx::Conv_998[FLOAT, 86x86x1x1]
%onnx::Conv_1001[FLOAT, 85x256x1x1]
%onnx::Conv_1004[FLOAT, 85x85x1x1]
%onnx::Conv_1007[FLOAT, 85x85x3x3]
%onnx::Conv_1010[FLOAT, 85x85x3x3]
%onnx::Conv_1013[FLOAT, 86x256x1x1]
%onnx::Conv_1016[FLOAT, 86x86x1x1]
%onnx::Conv_1019[FLOAT, 85x256x1x1]
%onnx::Conv_1022[FLOAT, 85x85x1x1]
%onnx::Conv_1025[FLOAT, 85x85x3x3]
%onnx::Conv_1028[FLOAT, 85x85x3x3]
%onnx::Conv_1031[FLOAT, 171x256x1x1]
%onnx::Conv_1032[FLOAT, 171]
%onnx::Conv_1034[FLOAT, 171x171x1x1]
%onnx::Conv_1037[FLOAT, 171x256x1x1]
%onnx::Conv_1040[FLOAT, 171x171x1x1]
%onnx::Conv_1043[FLOAT, 170x170x3x3]
%onnx::Conv_1044[FLOAT, 170]
%onnx::Conv_1046[FLOAT, 170x170x3x3]
%onnx::Conv_1049[FLOAT, 171x512x1x1]
%onnx::Conv_1052[FLOAT, 171x171x1x1]
%onnx::Conv_1055[FLOAT, 171x512x1x1]
%onnx::Conv_1058[FLOAT, 171x171x1x1]
%onnx::Conv_1061[FLOAT, 170x170x3x3]
%onnx::Conv_1064[FLOAT, 170x170x3x3]
%onnx::Conv_1067[FLOAT, 171x512x1x1]
%onnx::Conv_1070[FLOAT, 171x171x1x1]
%onnx::Conv_1073[FLOAT, 171x512x1x1]
%onnx::Conv_1076[FLOAT, 171x171x1x1]
%onnx::Conv_1079[FLOAT, 170x170x3x3]
%onnx::Conv_1082[FLOAT, 170x170x3x3]
) {
%onnx::Conv_1083 = Identity(%onnx::Conv_1044)
%onnx::Conv_1080 = Identity(%onnx::Conv_1044)
%onnx::Conv_1077 = Identity(%onnx::Conv_1032)
%onnx::Conv_1074 = Identity(%onnx::Conv_1032)
%onnx::Conv_1071 = Identity(%onnx::Conv_1032)
%onnx::Conv_1068 = Identity(%onnx::Conv_1032)
%onnx::Conv_1065 = Identity(%onnx::Conv_1044)
%onnx::Conv_1062 = Identity(%onnx::Conv_1044)
%onnx::Conv_1059 = Identity(%onnx::Conv_1032)
%onnx::Conv_1056 = Identity(%onnx::Conv_1032)
%onnx::Conv_1053 = Identity(%onnx::Conv_1032)
%onnx::Conv_1050 = Identity(%onnx::Conv_1032)
%onnx::Conv_1047 = Identity(%onnx::Conv_1044)
%onnx::Conv_1041 = Identity(%onnx::Conv_1032)
%onnx::Conv_1038 = Identity(%onnx::Conv_1032)
%onnx::Conv_1035 = Identity(%onnx::Conv_1032)
%onnx::Conv_1029 = Identity(%onnx::Conv_984)
%onnx::Conv_1026 = Identity(%onnx::Conv_984)
%onnx::Conv_1023 = Identity(%onnx::Conv_984)
%onnx::Conv_1020 = Identity(%onnx::Conv_984)
%onnx::Conv_1017 = Identity(%onnx::Conv_978)
%onnx::Conv_1014 = Identity(%onnx::Conv_978)
%onnx::Conv_1011 = Identity(%onnx::Conv_984)
%onnx::Conv_1008 = Identity(%onnx::Conv_984)
%onnx::Conv_1005 = Identity(%onnx::Conv_984)
%onnx::Conv_1002 = Identity(%onnx::Conv_984)
%onnx::Conv_999 = Identity(%onnx::Conv_978)
%onnx::Conv_996 = Identity(%onnx::Conv_978)
%onnx::Conv_993 = Identity(%onnx::Conv_984)
%onnx::Conv_990 = Identity(%onnx::Conv_984)
%onnx::Conv_987 = Identity(%onnx::Conv_984)
%onnx::Conv_981 = Identity(%onnx::Conv_978)
%onnx::Conv_975 = Identity(%onnx::Conv_936)
%onnx::Conv_972 = Identity(%onnx::Conv_936)
%onnx::Conv_969 = Identity(%onnx::Conv_924)
%onnx::Conv_966 = Identity(%onnx::Conv_924)
%onnx::Conv_963 = Identity(%onnx::Conv_924)
%onnx::Conv_960 = Identity(%onnx::Conv_924)
%onnx::Conv_957 = Identity(%onnx::Conv_936)
%onnx::Conv_954 = Identity(%onnx::Conv_936)
%onnx::Conv_951 = Identity(%onnx::Conv_924)
%onnx::Conv_948 = Identity(%onnx::Conv_924)
%onnx::Conv_945 = Identity(%onnx::Conv_924)
%onnx::Conv_942 = Identity(%onnx::Conv_924)
%onnx::Conv_939 = Identity(%onnx::Conv_936)
%onnx::Conv_933 = Identity(%onnx::Conv_924)
%onnx::Conv_930 = Identity(%onnx::Conv_924)
%onnx::Conv_927 = Identity(%onnx::Conv_924)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_920, %onnx::Conv_921)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_923, %onnx::Conv_924)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_926, %onnx::Conv_927)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_929, %onnx::Conv_930)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/Add_1_output_0, %/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_932, %onnx::Conv_933)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.1/Slice_output_0 = Slice(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0, %/layers.1/Constant_4_output_0, %/layers.1/Constant_2_output_0, %/layers.1/Constant_5_output_0)
%/layers.1/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/Slice_output_0, %/layers.1/Constant_6_output_0)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_935, %onnx::Conv_936)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_7_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_4_output_0)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_938, %onnx::Conv_939)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_941, %onnx::Conv_942)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_944, %onnx::Conv_945)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_947, %onnx::Conv_948)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/Add_1_output_0, %/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_950, %onnx::Conv_951)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.2/Slice_output_0 = Slice(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0, %/layers.2/Constant_4_output_0, %/layers.2/Constant_2_output_0, %/layers.2/Constant_5_output_0)
%/layers.2/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/Slice_output_0, %/layers.2/Constant_6_output_0)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_953, %onnx::Conv_954)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_7_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_4_output_0)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_956, %onnx::Conv_957)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_959, %onnx::Conv_960)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_962, %onnx::Conv_963)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_965, %onnx::Conv_966)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/Add_1_output_0, %/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_968, %onnx::Conv_969)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.3/Slice_output_0 = Slice(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0, %/layers.3/Constant_4_output_0, %/layers.3/Constant_2_output_0, %/layers.3/Constant_5_output_0)
%/layers.3/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/Slice_output_0, %/layers.3/Constant_6_output_0)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_7_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_4_output_0)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_974, %onnx::Conv_975)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_977, %onnx::Conv_978)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.5/Slice_output_0 = Slice(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0, %/layers.5/Constant_3_output_0, %/layers.5/Constant_1_output_0, %/layers.5/Constant_4_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_5_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/Slice_output_0, %/layers.5/Constant_5_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/Add_1_output_0, %/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.5/Slice_1_output_0 = Slice(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_7_output_0, %/layers.5/Constant_8_output_0, %/layers.5/Constant_6_output_0, %/layers.5/Constant_9_output_0)
%/layers.5/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/Slice_1_output_0, %/layers.5/Constant_10_output_0)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_11_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_11_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_4_output_0)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.6/Slice_output_0 = Slice(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0, %/layers.6/Constant_3_output_0, %/layers.6/Constant_1_output_0, %/layers.6/Constant_4_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_5_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/Slice_output_0, %/layers.6/Constant_5_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/Add_1_output_0, %/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.6/Slice_1_output_0 = Slice(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_7_output_0, %/layers.6/Constant_8_output_0, %/layers.6/Constant_6_output_0, %/layers.6/Constant_9_output_0)
%/layers.6/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/Slice_1_output_0, %/layers.6/Constant_10_output_0)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_11_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_11_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_4_output_0)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.7/Slice_output_0 = Slice(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0, %/layers.7/Constant_3_output_0, %/layers.7/Constant_1_output_0, %/layers.7/Constant_4_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_5_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/Slice_output_0, %/layers.7/Constant_5_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/Add_1_output_0, %/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.7/Slice_1_output_0 = Slice(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_7_output_0, %/layers.7/Constant_8_output_0, %/layers.7/Constant_6_output_0, %/layers.7/Constant_9_output_0)
%/layers.7/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/Slice_1_output_0, %/layers.7/Constant_10_output_0)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_1025, %onnx::Conv_1026)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_11_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_11_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_4_output_0)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_1028, %onnx::Conv_1029)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1031, %onnx::Conv_1032)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_1034, %onnx::Conv_1035)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1037, %onnx::Conv_1038)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/Add_1_output_0, %/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_1040, %onnx::Conv_1041)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.9/Slice_output_0 = Slice(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0, %/layers.9/Constant_4_output_0, %/layers.9/Constant_2_output_0, %/layers.9/Constant_5_output_0)
%/layers.9/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/Slice_output_0, %/layers.9/Constant_6_output_0)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_1043, %onnx::Conv_1044)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_7_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_4_output_0)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_1046, %onnx::Conv_1047)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_1049, %onnx::Conv_1050)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_1052, %onnx::Conv_1053)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_1055, %onnx::Conv_1056)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/Add_1_output_0, %/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_1058, %onnx::Conv_1059)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.10/Slice_output_0 = Slice(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0, %/layers.10/Constant_4_output_0, %/layers.10/Constant_2_output_0, %/layers.10/Constant_5_output_0)
%/layers.10/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/Slice_output_0, %/layers.10/Constant_6_output_0)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_1061, %onnx::Conv_1062)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_7_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_4_output_0)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_1064, %onnx::Conv_1065)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1067, %onnx::Conv_1068)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_1070, %onnx::Conv_1071)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1073, %onnx::Conv_1074)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/Add_1_output_0, %/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_1076, %onnx::Conv_1077)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.11/Slice_output_0 = Slice(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0, %/layers.11/Constant_4_output_0, %/layers.11/Constant_2_output_0, %/layers.11/Constant_5_output_0)
%/layers.11/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/Slice_output_0, %/layers.11/Constant_6_output_0)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_1079, %onnx::Conv_1080)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_7_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_4_output_0)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_1082, %onnx::Conv_1083)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%918 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %918
}
|
val_accuracy
| 93.098956
| 861,079,808
| 2,876,227
|
{'zcp_epe_nas': 49.33748031540234, 'zcp_fisher': 26.874330520629883, 'zcp_flops': 13777276928.0, 'zcp_grad_norm': 97.1728744506836, 'zcp_grasp': 27.6383056640625, 'zcp_jacov': -16.05623949883555, 'zcp_l2_norm': 883.5715942382812, 'zcp_nwot': 218.41976292776698, 'zcp_params': 2876227.0, 'zcp_plain': 0.013134658336639002, 'zcp_snip': 479.81939697265625, 'zcp_synflow': 109.7123152152813, 'zcp_zen': 89.15776824951172, 'zcp_val_accuracy': 0.917768418788909}
| |
NASBench101_256109
|
NASBench101
|
256109
|
9b14cfe4f7f3856db709bcf9a40a8936
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_887[FLOAT, 128x3x3x3]
%onnx::Conv_888[FLOAT, 128]
%onnx::Conv_890[FLOAT, 64x128x1x1]
%onnx::Conv_891[FLOAT, 64]
%onnx::Conv_893[FLOAT, 64x64x1x1]
%onnx::Conv_896[FLOAT, 64x64x1x1]
%onnx::Conv_899[FLOAT, 64x64x1x1]
%onnx::Conv_902[FLOAT, 64x128x1x1]
%onnx::Conv_905[FLOAT, 64x64x3x3]
%onnx::Conv_908[FLOAT, 64x128x1x1]
%onnx::Conv_911[FLOAT, 64x64x1x1]
%onnx::Conv_914[FLOAT, 64x64x1x1]
%onnx::Conv_917[FLOAT, 64x64x1x1]
%onnx::Conv_920[FLOAT, 64x128x1x1]
%onnx::Conv_923[FLOAT, 64x64x3x3]
%onnx::Conv_926[FLOAT, 64x128x1x1]
%onnx::Conv_929[FLOAT, 64x64x1x1]
%onnx::Conv_932[FLOAT, 64x64x1x1]
%onnx::Conv_935[FLOAT, 64x64x1x1]
%onnx::Conv_938[FLOAT, 64x128x1x1]
%onnx::Conv_941[FLOAT, 64x64x3x3]
%onnx::Conv_944[FLOAT, 128x128x1x1]
%onnx::Conv_947[FLOAT, 128x128x1x1]
%onnx::Conv_950[FLOAT, 128x128x1x1]
%onnx::Conv_953[FLOAT, 128x128x1x1]
%onnx::Conv_956[FLOAT, 128x128x1x1]
%onnx::Conv_959[FLOAT, 128x128x3x3]
%onnx::Conv_962[FLOAT, 128x256x1x1]
%onnx::Conv_965[FLOAT, 128x128x1x1]
%onnx::Conv_968[FLOAT, 128x128x1x1]
%onnx::Conv_971[FLOAT, 128x128x1x1]
%onnx::Conv_974[FLOAT, 128x256x1x1]
%onnx::Conv_977[FLOAT, 128x128x3x3]
%onnx::Conv_980[FLOAT, 128x256x1x1]
%onnx::Conv_983[FLOAT, 128x128x1x1]
%onnx::Conv_986[FLOAT, 128x128x1x1]
%onnx::Conv_989[FLOAT, 128x128x1x1]
%onnx::Conv_992[FLOAT, 128x256x1x1]
%onnx::Conv_995[FLOAT, 128x128x3x3]
%onnx::Conv_998[FLOAT, 256x256x1x1]
%onnx::Conv_999[FLOAT, 256]
%onnx::Conv_1001[FLOAT, 256x256x1x1]
%onnx::Conv_1004[FLOAT, 256x256x1x1]
%onnx::Conv_1007[FLOAT, 256x256x1x1]
%onnx::Conv_1010[FLOAT, 256x256x1x1]
%onnx::Conv_1013[FLOAT, 256x256x3x3]
%onnx::Conv_1016[FLOAT, 256x512x1x1]
%onnx::Conv_1019[FLOAT, 256x256x1x1]
%onnx::Conv_1022[FLOAT, 256x256x1x1]
%onnx::Conv_1025[FLOAT, 256x256x1x1]
%onnx::Conv_1028[FLOAT, 256x512x1x1]
%onnx::Conv_1031[FLOAT, 256x256x3x3]
%onnx::Conv_1034[FLOAT, 256x512x1x1]
%onnx::Conv_1037[FLOAT, 256x256x1x1]
%onnx::Conv_1040[FLOAT, 256x256x1x1]
%onnx::Conv_1043[FLOAT, 256x256x1x1]
%onnx::Conv_1046[FLOAT, 256x512x1x1]
%onnx::Conv_1049[FLOAT, 256x256x3x3]
) {
%onnx::Conv_1050 = Identity(%onnx::Conv_999)
%onnx::Conv_1047 = Identity(%onnx::Conv_999)
%onnx::Conv_1044 = Identity(%onnx::Conv_999)
%onnx::Conv_1041 = Identity(%onnx::Conv_999)
%onnx::Conv_1038 = Identity(%onnx::Conv_999)
%onnx::Conv_1035 = Identity(%onnx::Conv_999)
%onnx::Conv_1032 = Identity(%onnx::Conv_999)
%onnx::Conv_1029 = Identity(%onnx::Conv_999)
%onnx::Conv_1026 = Identity(%onnx::Conv_999)
%onnx::Conv_1023 = Identity(%onnx::Conv_999)
%onnx::Conv_1020 = Identity(%onnx::Conv_999)
%onnx::Conv_1017 = Identity(%onnx::Conv_999)
%onnx::Conv_1014 = Identity(%onnx::Conv_999)
%onnx::Conv_1011 = Identity(%onnx::Conv_999)
%onnx::Conv_1008 = Identity(%onnx::Conv_999)
%onnx::Conv_1005 = Identity(%onnx::Conv_999)
%onnx::Conv_1002 = Identity(%onnx::Conv_999)
%onnx::Conv_996 = Identity(%onnx::Conv_888)
%onnx::Conv_993 = Identity(%onnx::Conv_888)
%onnx::Conv_990 = Identity(%onnx::Conv_888)
%onnx::Conv_987 = Identity(%onnx::Conv_888)
%onnx::Conv_984 = Identity(%onnx::Conv_888)
%onnx::Conv_981 = Identity(%onnx::Conv_888)
%onnx::Conv_978 = Identity(%onnx::Conv_888)
%onnx::Conv_975 = Identity(%onnx::Conv_888)
%onnx::Conv_972 = Identity(%onnx::Conv_888)
%onnx::Conv_969 = Identity(%onnx::Conv_888)
%onnx::Conv_966 = Identity(%onnx::Conv_888)
%onnx::Conv_963 = Identity(%onnx::Conv_888)
%onnx::Conv_960 = Identity(%onnx::Conv_888)
%onnx::Conv_957 = Identity(%onnx::Conv_888)
%onnx::Conv_954 = Identity(%onnx::Conv_888)
%onnx::Conv_951 = Identity(%onnx::Conv_888)
%onnx::Conv_948 = Identity(%onnx::Conv_888)
%onnx::Conv_945 = Identity(%onnx::Conv_888)
%onnx::Conv_942 = Identity(%onnx::Conv_891)
%onnx::Conv_939 = Identity(%onnx::Conv_891)
%onnx::Conv_936 = Identity(%onnx::Conv_891)
%onnx::Conv_933 = Identity(%onnx::Conv_891)
%onnx::Conv_930 = Identity(%onnx::Conv_891)
%onnx::Conv_927 = Identity(%onnx::Conv_891)
%onnx::Conv_924 = Identity(%onnx::Conv_891)
%onnx::Conv_921 = Identity(%onnx::Conv_891)
%onnx::Conv_918 = Identity(%onnx::Conv_891)
%onnx::Conv_915 = Identity(%onnx::Conv_891)
%onnx::Conv_912 = Identity(%onnx::Conv_891)
%onnx::Conv_909 = Identity(%onnx::Conv_891)
%onnx::Conv_906 = Identity(%onnx::Conv_891)
%onnx::Conv_903 = Identity(%onnx::Conv_891)
%onnx::Conv_900 = Identity(%onnx::Conv_891)
%onnx::Conv_897 = Identity(%onnx::Conv_891)
%onnx::Conv_894 = Identity(%onnx::Conv_891)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_887, %onnx::Conv_888)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/Add_3_output_0, %/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_5_output_0 = Add(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_4_output_0)
%/layers.1/Add_6_output_0 = Add(%/layers.1/Add_5_output_0, %/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_6_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_908, %onnx::Conv_909)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_911, %onnx::Conv_912)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_914, %onnx::Conv_915)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_917, %onnx::Conv_918)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_920, %onnx::Conv_921)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/Add_3_output_0, %/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_923, %onnx::Conv_924)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_5_output_0 = Add(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_4_output_0)
%/layers.2/Add_6_output_0 = Add(%/layers.2/Add_5_output_0, %/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_6_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_926, %onnx::Conv_927)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_929, %onnx::Conv_930)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_932, %onnx::Conv_933)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_935, %onnx::Conv_936)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_938, %onnx::Conv_939)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/Add_3_output_0, %/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_4_output_0, %onnx::Conv_941, %onnx::Conv_942)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_5_output_0 = Add(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_4_output_0)
%/layers.3/Add_6_output_0 = Add(%/layers.3/Add_5_output_0, %/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_6_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_944, %onnx::Conv_945)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_947, %onnx::Conv_948)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_950, %onnx::Conv_951)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_953, %onnx::Conv_954)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_956, %onnx::Conv_957)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/Add_3_output_0, %/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_959, %onnx::Conv_960)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_5_output_0 = Add(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_4_output_0)
%/layers.5/Add_6_output_0 = Add(%/layers.5/Add_5_output_0, %/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_6_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_962, %onnx::Conv_963)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_965, %onnx::Conv_966)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_968, %onnx::Conv_969)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_974, %onnx::Conv_975)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/Add_3_output_0, %/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_977, %onnx::Conv_978)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_5_output_0 = Add(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_4_output_0)
%/layers.6/Add_6_output_0 = Add(%/layers.6/Add_5_output_0, %/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_6_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/Add_3_output_0, %/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_4_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_5_output_0 = Add(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_4_output_0)
%/layers.7/Add_6_output_0 = Add(%/layers.7/Add_5_output_0, %/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_6_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/Add_3_output_0, %/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_5_output_0 = Add(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_4_output_0)
%/layers.9/Add_6_output_0 = Add(%/layers.9/Add_5_output_0, %/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_6_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_1025, %onnx::Conv_1026)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_1028, %onnx::Conv_1029)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/Add_3_output_0, %/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_1031, %onnx::Conv_1032)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_5_output_0 = Add(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_4_output_0)
%/layers.10/Add_6_output_0 = Add(%/layers.10/Add_5_output_0, %/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_6_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1034, %onnx::Conv_1035)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_1037, %onnx::Conv_1038)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_1040, %onnx::Conv_1041)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_1043, %onnx::Conv_1044)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1046, %onnx::Conv_1047)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/Add_3_output_0, %/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_4_output_0, %onnx::Conv_1049, %onnx::Conv_1050)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_5_output_0 = Add(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_4_output_0)
%/layers.11/Add_6_output_0 = Add(%/layers.11/Add_5_output_0, %/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_6_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.5/maxpool/MaxPool_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%885 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %885
}
|
val_accuracy
| 90.144229
| 1,199,056,896
| 3,989,898
|
{'zcp_epe_nas': 107.85168567862073, 'zcp_fisher': 1030.2381591796875, 'zcp_flops': 19184910336.0, 'zcp_grad_norm': 611.1709594726562, 'zcp_grasp': 1522.5625, 'zcp_jacov': -16.0405327254986, 'zcp_l2_norm': 993.6826171875, 'zcp_nwot': 224.8982088800027, 'zcp_params': 3989898.0, 'zcp_plain': 0.053053632378578006, 'zcp_snip': 3052.359375, 'zcp_synflow': 104.86798746421155, 'zcp_zen': 90.14356994628906, 'zcp_val_accuracy': 0.9201722741127011}
| |
NASBench101_65546
|
NASBench101
|
65546
|
27c5e0958fb3521051cfcdf8190cb44b
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_842[FLOAT, 128x3x3x3]
%onnx::Conv_843[FLOAT, 128]
%onnx::Conv_845[FLOAT, 64x128x1x1]
%onnx::Conv_846[FLOAT, 64]
%onnx::Conv_848[FLOAT, 64x64x3x3]
%onnx::Conv_851[FLOAT, 64x128x1x1]
%onnx::Conv_854[FLOAT, 64x128x1x1]
%onnx::Conv_857[FLOAT, 64x64x3x3]
%onnx::Conv_860[FLOAT, 64x64x1x1]
%onnx::Conv_863[FLOAT, 64x128x1x1]
%onnx::Conv_866[FLOAT, 64x64x3x3]
%onnx::Conv_869[FLOAT, 64x128x1x1]
%onnx::Conv_872[FLOAT, 64x128x1x1]
%onnx::Conv_875[FLOAT, 64x64x3x3]
%onnx::Conv_878[FLOAT, 64x64x1x1]
%onnx::Conv_881[FLOAT, 64x128x1x1]
%onnx::Conv_884[FLOAT, 64x64x3x3]
%onnx::Conv_887[FLOAT, 64x128x1x1]
%onnx::Conv_890[FLOAT, 64x128x1x1]
%onnx::Conv_893[FLOAT, 64x64x3x3]
%onnx::Conv_896[FLOAT, 64x64x1x1]
%onnx::Conv_899[FLOAT, 128x128x1x1]
%onnx::Conv_902[FLOAT, 128x128x3x3]
%onnx::Conv_905[FLOAT, 128x128x1x1]
%onnx::Conv_908[FLOAT, 128x128x1x1]
%onnx::Conv_911[FLOAT, 128x128x3x3]
%onnx::Conv_914[FLOAT, 128x128x1x1]
%onnx::Conv_917[FLOAT, 128x256x1x1]
%onnx::Conv_920[FLOAT, 128x128x3x3]
%onnx::Conv_923[FLOAT, 128x256x1x1]
%onnx::Conv_926[FLOAT, 128x256x1x1]
%onnx::Conv_929[FLOAT, 128x128x3x3]
%onnx::Conv_932[FLOAT, 128x128x1x1]
%onnx::Conv_935[FLOAT, 128x256x1x1]
%onnx::Conv_938[FLOAT, 128x128x3x3]
%onnx::Conv_941[FLOAT, 128x256x1x1]
%onnx::Conv_944[FLOAT, 128x256x1x1]
%onnx::Conv_947[FLOAT, 128x128x3x3]
%onnx::Conv_950[FLOAT, 128x128x1x1]
%onnx::Conv_953[FLOAT, 256x256x1x1]
%onnx::Conv_954[FLOAT, 256]
%onnx::Conv_956[FLOAT, 256x256x3x3]
%onnx::Conv_959[FLOAT, 256x256x1x1]
%onnx::Conv_962[FLOAT, 256x256x1x1]
%onnx::Conv_965[FLOAT, 256x256x3x3]
%onnx::Conv_968[FLOAT, 256x256x1x1]
%onnx::Conv_971[FLOAT, 256x512x1x1]
%onnx::Conv_974[FLOAT, 256x256x3x3]
%onnx::Conv_977[FLOAT, 256x512x1x1]
%onnx::Conv_980[FLOAT, 256x512x1x1]
%onnx::Conv_983[FLOAT, 256x256x3x3]
%onnx::Conv_986[FLOAT, 256x256x1x1]
%onnx::Conv_989[FLOAT, 256x512x1x1]
%onnx::Conv_992[FLOAT, 256x256x3x3]
%onnx::Conv_995[FLOAT, 256x512x1x1]
%onnx::Conv_998[FLOAT, 256x512x1x1]
%onnx::Conv_1001[FLOAT, 256x256x3x3]
%onnx::Conv_1004[FLOAT, 256x256x1x1]
) {
%onnx::Conv_1005 = Identity(%onnx::Conv_954)
%onnx::Conv_1002 = Identity(%onnx::Conv_954)
%onnx::Conv_999 = Identity(%onnx::Conv_954)
%onnx::Conv_996 = Identity(%onnx::Conv_954)
%onnx::Conv_993 = Identity(%onnx::Conv_954)
%onnx::Conv_990 = Identity(%onnx::Conv_954)
%onnx::Conv_987 = Identity(%onnx::Conv_954)
%onnx::Conv_984 = Identity(%onnx::Conv_954)
%onnx::Conv_981 = Identity(%onnx::Conv_954)
%onnx::Conv_978 = Identity(%onnx::Conv_954)
%onnx::Conv_975 = Identity(%onnx::Conv_954)
%onnx::Conv_972 = Identity(%onnx::Conv_954)
%onnx::Conv_969 = Identity(%onnx::Conv_954)
%onnx::Conv_966 = Identity(%onnx::Conv_954)
%onnx::Conv_963 = Identity(%onnx::Conv_954)
%onnx::Conv_960 = Identity(%onnx::Conv_954)
%onnx::Conv_957 = Identity(%onnx::Conv_954)
%onnx::Conv_951 = Identity(%onnx::Conv_843)
%onnx::Conv_948 = Identity(%onnx::Conv_843)
%onnx::Conv_945 = Identity(%onnx::Conv_843)
%onnx::Conv_942 = Identity(%onnx::Conv_843)
%onnx::Conv_939 = Identity(%onnx::Conv_843)
%onnx::Conv_936 = Identity(%onnx::Conv_843)
%onnx::Conv_933 = Identity(%onnx::Conv_843)
%onnx::Conv_930 = Identity(%onnx::Conv_843)
%onnx::Conv_927 = Identity(%onnx::Conv_843)
%onnx::Conv_924 = Identity(%onnx::Conv_843)
%onnx::Conv_921 = Identity(%onnx::Conv_843)
%onnx::Conv_918 = Identity(%onnx::Conv_843)
%onnx::Conv_915 = Identity(%onnx::Conv_843)
%onnx::Conv_912 = Identity(%onnx::Conv_843)
%onnx::Conv_909 = Identity(%onnx::Conv_843)
%onnx::Conv_906 = Identity(%onnx::Conv_843)
%onnx::Conv_903 = Identity(%onnx::Conv_843)
%onnx::Conv_900 = Identity(%onnx::Conv_843)
%onnx::Conv_897 = Identity(%onnx::Conv_846)
%onnx::Conv_894 = Identity(%onnx::Conv_846)
%onnx::Conv_891 = Identity(%onnx::Conv_846)
%onnx::Conv_888 = Identity(%onnx::Conv_846)
%onnx::Conv_885 = Identity(%onnx::Conv_846)
%onnx::Conv_882 = Identity(%onnx::Conv_846)
%onnx::Conv_879 = Identity(%onnx::Conv_846)
%onnx::Conv_876 = Identity(%onnx::Conv_846)
%onnx::Conv_873 = Identity(%onnx::Conv_846)
%onnx::Conv_870 = Identity(%onnx::Conv_846)
%onnx::Conv_867 = Identity(%onnx::Conv_846)
%onnx::Conv_864 = Identity(%onnx::Conv_846)
%onnx::Conv_861 = Identity(%onnx::Conv_846)
%onnx::Conv_858 = Identity(%onnx::Conv_846)
%onnx::Conv_855 = Identity(%onnx::Conv_846)
%onnx::Conv_852 = Identity(%onnx::Conv_846)
%onnx::Conv_849 = Identity(%onnx::Conv_846)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_842, %onnx::Conv_843)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_845, %onnx::Conv_846)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_848, %onnx::Conv_849)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_851, %onnx::Conv_852)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_854, %onnx::Conv_855)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.3/maxpool/MaxPool_output_0, %/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_857, %onnx::Conv_858)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.3/maxpool/MaxPool_output_0, %/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.3/maxpool/MaxPool_output_0, %/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_908, %onnx::Conv_909)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.3/maxpool/MaxPool_output_0, %/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_911, %onnx::Conv_912)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_914, %onnx::Conv_915)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_917, %onnx::Conv_918)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_920, %onnx::Conv_921)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_923, %onnx::Conv_924)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_926, %onnx::Conv_927)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.3/maxpool/MaxPool_output_0, %/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_929, %onnx::Conv_930)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_932, %onnx::Conv_933)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_935, %onnx::Conv_936)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_938, %onnx::Conv_939)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_941, %onnx::Conv_942)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_944, %onnx::Conv_945)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.3/maxpool/MaxPool_output_0, %/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_947, %onnx::Conv_948)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_950, %onnx::Conv_951)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_953, %onnx::Conv_954)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_956, %onnx::Conv_957)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_959, %onnx::Conv_960)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_962, %onnx::Conv_963)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.3/maxpool/MaxPool_output_0, %/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_965, %onnx::Conv_966)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_968, %onnx::Conv_969)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_974, %onnx::Conv_975)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_977, %onnx::Conv_978)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.3/maxpool/MaxPool_output_0, %/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.3/maxpool/MaxPool_output_0, %/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%840 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %840
}
|
val_accuracy
| 91.877002
| 1,861,756,928
| 6,230,410
|
{'zcp_epe_nas': 92.89028235001567, 'zcp_fisher': 28.401762008666992, 'zcp_flops': 29788110848.0, 'zcp_grad_norm': 93.4677505493164, 'zcp_grasp': -12.065185546875, 'zcp_jacov': -16.05583268988411, 'zcp_l2_norm': 1040.2454833984375, 'zcp_nwot': 224.0651101470309, 'zcp_params': 6230410.0, 'zcp_plain': 0.05187225714325901, 'zcp_snip': 600.6484985351562, 'zcp_synflow': 96.98177925686291, 'zcp_zen': 105.2791748046875, 'zcp_val_accuracy': 0.9120593070983881}
| |
NASBench101_84279
|
NASBench101
|
84279
|
331396a6181b05a2335b9fe9300f3e82
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_545[FLOAT, 128x3x3x3]
%onnx::Conv_546[FLOAT, 128]
%onnx::Conv_548[FLOAT, 32x128x1x1]
%onnx::Conv_549[FLOAT, 32]
%onnx::Conv_551[FLOAT, 32x128x1x1]
%onnx::Conv_554[FLOAT, 32x32x1x1]
%onnx::Conv_557[FLOAT, 32x128x1x1]
%onnx::Conv_560[FLOAT, 32x128x1x1]
%onnx::Conv_563[FLOAT, 32x32x1x1]
%onnx::Conv_566[FLOAT, 32x128x1x1]
%onnx::Conv_569[FLOAT, 32x128x1x1]
%onnx::Conv_572[FLOAT, 32x32x1x1]
%onnx::Conv_575[FLOAT, 64x128x1x1]
%onnx::Conv_576[FLOAT, 64]
%onnx::Conv_578[FLOAT, 64x128x1x1]
%onnx::Conv_581[FLOAT, 64x64x1x1]
%onnx::Conv_584[FLOAT, 64x256x1x1]
%onnx::Conv_587[FLOAT, 64x256x1x1]
%onnx::Conv_590[FLOAT, 64x64x1x1]
%onnx::Conv_593[FLOAT, 64x256x1x1]
%onnx::Conv_596[FLOAT, 64x256x1x1]
%onnx::Conv_599[FLOAT, 64x64x1x1]
%onnx::Conv_602[FLOAT, 128x256x1x1]
%onnx::Conv_605[FLOAT, 128x256x1x1]
%onnx::Conv_608[FLOAT, 128x128x1x1]
%onnx::Conv_611[FLOAT, 128x512x1x1]
%onnx::Conv_614[FLOAT, 128x512x1x1]
%onnx::Conv_617[FLOAT, 128x128x1x1]
%onnx::Conv_620[FLOAT, 128x512x1x1]
%onnx::Conv_623[FLOAT, 128x512x1x1]
%onnx::Conv_626[FLOAT, 128x128x1x1]
) {
%onnx::Conv_627 = Identity(%onnx::Conv_546)
%onnx::Conv_624 = Identity(%onnx::Conv_546)
%onnx::Conv_621 = Identity(%onnx::Conv_546)
%onnx::Conv_618 = Identity(%onnx::Conv_546)
%onnx::Conv_615 = Identity(%onnx::Conv_546)
%onnx::Conv_612 = Identity(%onnx::Conv_546)
%onnx::Conv_609 = Identity(%onnx::Conv_546)
%onnx::Conv_606 = Identity(%onnx::Conv_546)
%onnx::Conv_603 = Identity(%onnx::Conv_546)
%onnx::Conv_600 = Identity(%onnx::Conv_576)
%onnx::Conv_597 = Identity(%onnx::Conv_576)
%onnx::Conv_594 = Identity(%onnx::Conv_576)
%onnx::Conv_591 = Identity(%onnx::Conv_576)
%onnx::Conv_588 = Identity(%onnx::Conv_576)
%onnx::Conv_585 = Identity(%onnx::Conv_576)
%onnx::Conv_582 = Identity(%onnx::Conv_576)
%onnx::Conv_579 = Identity(%onnx::Conv_576)
%onnx::Conv_573 = Identity(%onnx::Conv_549)
%onnx::Conv_570 = Identity(%onnx::Conv_549)
%onnx::Conv_567 = Identity(%onnx::Conv_549)
%onnx::Conv_564 = Identity(%onnx::Conv_549)
%onnx::Conv_561 = Identity(%onnx::Conv_549)
%onnx::Conv_558 = Identity(%onnx::Conv_549)
%onnx::Conv_555 = Identity(%onnx::Conv_549)
%onnx::Conv_552 = Identity(%onnx::Conv_549)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_545, %onnx::Conv_546)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_548, %onnx::Conv_549)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_551, %onnx::Conv_552)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_554, %onnx::Conv_555)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_557, %onnx::Conv_558)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_560, %onnx::Conv_561)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_563, %onnx::Conv_564)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_566, %onnx::Conv_567)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_569, %onnx::Conv_570)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_572, %onnx::Conv_573)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_575, %onnx::Conv_576)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_578, %onnx::Conv_579)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_581, %onnx::Conv_582)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_584, %onnx::Conv_585)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_587, %onnx::Conv_588)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_590, %onnx::Conv_591)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_593, %onnx::Conv_594)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_596, %onnx::Conv_597)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_599, %onnx::Conv_600)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_602, %onnx::Conv_603)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_605, %onnx::Conv_606)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_608, %onnx::Conv_609)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_611, %onnx::Conv_612)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_614, %onnx::Conv_615)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_617, %onnx::Conv_618)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_620, %onnx::Conv_621)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_623, %onnx::Conv_624)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_626, %onnx::Conv_627)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.4/maxpool/MaxPool_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%543 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %543
}
|
val_accuracy
| 87.610179
| 165,357,568
| 511,562
|
{'zcp_epe_nas': 85.33124900386271, 'zcp_fisher': 1.342757105827331, 'zcp_flops': 2645721088.0, 'zcp_grad_norm': 19.66354751586914, 'zcp_grasp': -0.44181060791015603, 'zcp_jacov': -16.066962475160977, 'zcp_l2_norm': 500.768310546875, 'zcp_nwot': 204.40239524501433, 'zcp_params': 511562.0, 'zcp_plain': 0.008188031613826, 'zcp_snip': 95.57157897949219, 'zcp_synflow': 57.79674072604773, 'zcp_zen': 45.66751480102539, 'zcp_val_accuracy': 0.9201722741127011}
| |
NASBench101_106425
|
NASBench101
|
106425
|
4050f0291bd06ac68937b4d7e04c59b6
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_986[FLOAT, 128x3x3x3]
%onnx::Conv_987[FLOAT, 128]
%onnx::Conv_989[FLOAT, 128x128x1x1]
%onnx::Conv_992[FLOAT, 128x128x1x1]
%onnx::Conv_995[FLOAT, 128x128x1x1]
%onnx::Conv_998[FLOAT, 128x128x3x3]
%onnx::Conv_1001[FLOAT, 128x128x1x1]
%onnx::Conv_1004[FLOAT, 128x128x1x1]
%onnx::Conv_1007[FLOAT, 128x128x1x1]
%onnx::Conv_1010[FLOAT, 128x128x1x1]
%onnx::Conv_1013[FLOAT, 128x128x1x1]
%onnx::Conv_1016[FLOAT, 128x128x1x1]
%onnx::Conv_1019[FLOAT, 128x128x3x3]
%onnx::Conv_1022[FLOAT, 128x128x1x1]
%onnx::Conv_1025[FLOAT, 128x128x1x1]
%onnx::Conv_1028[FLOAT, 128x128x1x1]
%onnx::Conv_1031[FLOAT, 128x128x1x1]
%onnx::Conv_1034[FLOAT, 128x128x1x1]
%onnx::Conv_1037[FLOAT, 128x128x1x1]
%onnx::Conv_1040[FLOAT, 128x128x3x3]
%onnx::Conv_1043[FLOAT, 128x128x1x1]
%onnx::Conv_1046[FLOAT, 128x128x1x1]
%onnx::Conv_1049[FLOAT, 128x128x1x1]
%onnx::Conv_1052[FLOAT, 256x128x1x1]
%onnx::Conv_1053[FLOAT, 256]
%onnx::Conv_1055[FLOAT, 256x256x1x1]
%onnx::Conv_1058[FLOAT, 256x128x1x1]
%onnx::Conv_1061[FLOAT, 256x256x3x3]
%onnx::Conv_1064[FLOAT, 256x128x1x1]
%onnx::Conv_1067[FLOAT, 256x256x1x1]
%onnx::Conv_1070[FLOAT, 256x256x1x1]
%onnx::Conv_1073[FLOAT, 256x256x1x1]
%onnx::Conv_1076[FLOAT, 256x256x1x1]
%onnx::Conv_1079[FLOAT, 256x256x1x1]
%onnx::Conv_1082[FLOAT, 256x256x3x3]
%onnx::Conv_1085[FLOAT, 256x256x1x1]
%onnx::Conv_1088[FLOAT, 256x256x1x1]
%onnx::Conv_1091[FLOAT, 256x256x1x1]
%onnx::Conv_1094[FLOAT, 256x256x1x1]
%onnx::Conv_1097[FLOAT, 256x256x1x1]
%onnx::Conv_1100[FLOAT, 256x256x1x1]
%onnx::Conv_1103[FLOAT, 256x256x3x3]
%onnx::Conv_1106[FLOAT, 256x256x1x1]
%onnx::Conv_1109[FLOAT, 256x256x1x1]
%onnx::Conv_1112[FLOAT, 256x256x1x1]
%onnx::Conv_1115[FLOAT, 512x256x1x1]
%onnx::Conv_1116[FLOAT, 512]
%onnx::Conv_1118[FLOAT, 512x512x1x1]
%onnx::Conv_1121[FLOAT, 512x256x1x1]
%onnx::Conv_1124[FLOAT, 512x512x3x3]
%onnx::Conv_1127[FLOAT, 512x256x1x1]
%onnx::Conv_1130[FLOAT, 512x512x1x1]
%onnx::Conv_1133[FLOAT, 512x512x1x1]
%onnx::Conv_1136[FLOAT, 512x512x1x1]
%onnx::Conv_1139[FLOAT, 512x512x1x1]
%onnx::Conv_1142[FLOAT, 512x512x1x1]
%onnx::Conv_1145[FLOAT, 512x512x3x3]
%onnx::Conv_1148[FLOAT, 512x512x1x1]
%onnx::Conv_1151[FLOAT, 512x512x1x1]
%onnx::Conv_1154[FLOAT, 512x512x1x1]
%onnx::Conv_1157[FLOAT, 512x512x1x1]
%onnx::Conv_1160[FLOAT, 512x512x1x1]
%onnx::Conv_1163[FLOAT, 512x512x1x1]
%onnx::Conv_1166[FLOAT, 512x512x3x3]
%onnx::Conv_1169[FLOAT, 512x512x1x1]
%onnx::Conv_1172[FLOAT, 512x512x1x1]
%onnx::Conv_1175[FLOAT, 512x512x1x1]
) {
%onnx::Conv_1176 = Identity(%onnx::Conv_1116)
%onnx::Conv_1173 = Identity(%onnx::Conv_1116)
%onnx::Conv_1170 = Identity(%onnx::Conv_1116)
%onnx::Conv_1167 = Identity(%onnx::Conv_1116)
%onnx::Conv_1164 = Identity(%onnx::Conv_1116)
%onnx::Conv_1161 = Identity(%onnx::Conv_1116)
%onnx::Conv_1158 = Identity(%onnx::Conv_1116)
%onnx::Conv_1155 = Identity(%onnx::Conv_1116)
%onnx::Conv_1152 = Identity(%onnx::Conv_1116)
%onnx::Conv_1149 = Identity(%onnx::Conv_1116)
%onnx::Conv_1146 = Identity(%onnx::Conv_1116)
%onnx::Conv_1143 = Identity(%onnx::Conv_1116)
%onnx::Conv_1140 = Identity(%onnx::Conv_1116)
%onnx::Conv_1137 = Identity(%onnx::Conv_1116)
%onnx::Conv_1134 = Identity(%onnx::Conv_1116)
%onnx::Conv_1131 = Identity(%onnx::Conv_1116)
%onnx::Conv_1128 = Identity(%onnx::Conv_1116)
%onnx::Conv_1125 = Identity(%onnx::Conv_1116)
%onnx::Conv_1122 = Identity(%onnx::Conv_1116)
%onnx::Conv_1119 = Identity(%onnx::Conv_1116)
%onnx::Conv_1113 = Identity(%onnx::Conv_1053)
%onnx::Conv_1110 = Identity(%onnx::Conv_1053)
%onnx::Conv_1107 = Identity(%onnx::Conv_1053)
%onnx::Conv_1104 = Identity(%onnx::Conv_1053)
%onnx::Conv_1101 = Identity(%onnx::Conv_1053)
%onnx::Conv_1098 = Identity(%onnx::Conv_1053)
%onnx::Conv_1095 = Identity(%onnx::Conv_1053)
%onnx::Conv_1092 = Identity(%onnx::Conv_1053)
%onnx::Conv_1089 = Identity(%onnx::Conv_1053)
%onnx::Conv_1086 = Identity(%onnx::Conv_1053)
%onnx::Conv_1083 = Identity(%onnx::Conv_1053)
%onnx::Conv_1080 = Identity(%onnx::Conv_1053)
%onnx::Conv_1077 = Identity(%onnx::Conv_1053)
%onnx::Conv_1074 = Identity(%onnx::Conv_1053)
%onnx::Conv_1071 = Identity(%onnx::Conv_1053)
%onnx::Conv_1068 = Identity(%onnx::Conv_1053)
%onnx::Conv_1065 = Identity(%onnx::Conv_1053)
%onnx::Conv_1062 = Identity(%onnx::Conv_1053)
%onnx::Conv_1059 = Identity(%onnx::Conv_1053)
%onnx::Conv_1056 = Identity(%onnx::Conv_1053)
%onnx::Conv_1050 = Identity(%onnx::Conv_987)
%onnx::Conv_1047 = Identity(%onnx::Conv_987)
%onnx::Conv_1044 = Identity(%onnx::Conv_987)
%onnx::Conv_1041 = Identity(%onnx::Conv_987)
%onnx::Conv_1038 = Identity(%onnx::Conv_987)
%onnx::Conv_1035 = Identity(%onnx::Conv_987)
%onnx::Conv_1032 = Identity(%onnx::Conv_987)
%onnx::Conv_1029 = Identity(%onnx::Conv_987)
%onnx::Conv_1026 = Identity(%onnx::Conv_987)
%onnx::Conv_1023 = Identity(%onnx::Conv_987)
%onnx::Conv_1020 = Identity(%onnx::Conv_987)
%onnx::Conv_1017 = Identity(%onnx::Conv_987)
%onnx::Conv_1014 = Identity(%onnx::Conv_987)
%onnx::Conv_1011 = Identity(%onnx::Conv_987)
%onnx::Conv_1008 = Identity(%onnx::Conv_987)
%onnx::Conv_1005 = Identity(%onnx::Conv_987)
%onnx::Conv_1002 = Identity(%onnx::Conv_987)
%onnx::Conv_999 = Identity(%onnx::Conv_987)
%onnx::Conv_996 = Identity(%onnx::Conv_987)
%onnx::Conv_993 = Identity(%onnx::Conv_987)
%onnx::Conv_990 = Identity(%onnx::Conv_987)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_986, %onnx::Conv_987)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/Add_1_output_0, %/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/Add_5_output_0 = Add(%/layers.1/Add_4_output_0, %/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_5_output_0)
%/layers.1/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_6_output_0 = Add(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_4_output_0)
%/layers.1/Add_7_output_0 = Add(%/layers.1/Add_6_output_0, %/layers.1/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_7_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/Add_1_output_0, %/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_1025, %onnx::Conv_1026)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/Add_5_output_0 = Add(%/layers.2/Add_4_output_0, %/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_5_output_0)
%/layers.2/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_6_output_0 = Add(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_4_output_0)
%/layers.2/Add_7_output_0 = Add(%/layers.2/Add_6_output_0, %/layers.2/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_7_output_0, %onnx::Conv_1028, %onnx::Conv_1029)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1031, %onnx::Conv_1032)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_1034, %onnx::Conv_1035)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1037, %onnx::Conv_1038)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/Add_1_output_0, %/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_1040, %onnx::Conv_1041)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1043, %onnx::Conv_1044)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_1046, %onnx::Conv_1047)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/Add_5_output_0 = Add(%/layers.3/Add_4_output_0, %/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_5_output_0)
%/layers.3/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_6_output_0 = Add(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_4_output_0)
%/layers.3/Add_7_output_0 = Add(%/layers.3/Add_6_output_0, %/layers.3/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_7_output_0, %onnx::Conv_1049, %onnx::Conv_1050)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1052, %onnx::Conv_1053)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_1055, %onnx::Conv_1056)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1058, %onnx::Conv_1059)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/Add_1_output_0, %/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_1061, %onnx::Conv_1062)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1064, %onnx::Conv_1065)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_1067, %onnx::Conv_1068)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/Add_5_output_0 = Add(%/layers.5/Add_4_output_0, %/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_5_output_0)
%/layers.5/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_6_output_0 = Add(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_4_output_0)
%/layers.5/Add_7_output_0 = Add(%/layers.5/Add_6_output_0, %/layers.5/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_7_output_0, %onnx::Conv_1070, %onnx::Conv_1071)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1073, %onnx::Conv_1074)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_1076, %onnx::Conv_1077)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1079, %onnx::Conv_1080)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/Add_1_output_0, %/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_1082, %onnx::Conv_1083)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1085, %onnx::Conv_1086)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_1088, %onnx::Conv_1089)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/Add_5_output_0 = Add(%/layers.6/Add_4_output_0, %/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_5_output_0)
%/layers.6/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_6_output_0 = Add(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_4_output_0)
%/layers.6/Add_7_output_0 = Add(%/layers.6/Add_6_output_0, %/layers.6/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_7_output_0, %onnx::Conv_1091, %onnx::Conv_1092)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1094, %onnx::Conv_1095)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_1097, %onnx::Conv_1098)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1100, %onnx::Conv_1101)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/Add_1_output_0, %/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_1103, %onnx::Conv_1104)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1106, %onnx::Conv_1107)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_1109, %onnx::Conv_1110)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/Add_5_output_0 = Add(%/layers.7/Add_4_output_0, %/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_5_output_0)
%/layers.7/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_6_output_0 = Add(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_4_output_0)
%/layers.7/Add_7_output_0 = Add(%/layers.7/Add_6_output_0, %/layers.7/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_7_output_0, %onnx::Conv_1112, %onnx::Conv_1113)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1115, %onnx::Conv_1116)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_1118, %onnx::Conv_1119)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1121, %onnx::Conv_1122)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/Add_1_output_0, %/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_1124, %onnx::Conv_1125)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1127, %onnx::Conv_1128)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_1130, %onnx::Conv_1131)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/Add_5_output_0 = Add(%/layers.9/Add_4_output_0, %/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_5_output_0)
%/layers.9/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_6_output_0 = Add(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_4_output_0)
%/layers.9/Add_7_output_0 = Add(%/layers.9/Add_6_output_0, %/layers.9/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_7_output_0, %onnx::Conv_1133, %onnx::Conv_1134)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1136, %onnx::Conv_1137)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_1139, %onnx::Conv_1140)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1142, %onnx::Conv_1143)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/Add_1_output_0, %/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_1145, %onnx::Conv_1146)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1148, %onnx::Conv_1149)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_1151, %onnx::Conv_1152)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/Add_5_output_0 = Add(%/layers.10/Add_4_output_0, %/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_5_output_0)
%/layers.10/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_6_output_0 = Add(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_4_output_0)
%/layers.10/Add_7_output_0 = Add(%/layers.10/Add_6_output_0, %/layers.10/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_7_output_0, %onnx::Conv_1154, %onnx::Conv_1155)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1157, %onnx::Conv_1158)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_1160, %onnx::Conv_1161)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1163, %onnx::Conv_1164)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/Add_1_output_0, %/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_1166, %onnx::Conv_1167)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1169, %onnx::Conv_1170)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_1172, %onnx::Conv_1173)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/Add_5_output_0 = Add(%/layers.11/Add_4_output_0, %/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_5_output_0)
%/layers.11/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_6_output_0 = Add(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_4_output_0)
%/layers.11/Add_7_output_0 = Add(%/layers.11/Add_6_output_0, %/layers.11/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_7_output_0, %onnx::Conv_1175, %onnx::Conv_1176)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%984 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %984
}
|
val_accuracy
| 90.925479
| 4,475,856,896
| 15,037,834
|
{'zcp_epe_nas': 123.69138339737191, 'zcp_fisher': 150.51922607421875, 'zcp_flops': 71613710336.0, 'zcp_grad_norm': 254.52655029296875, 'zcp_grasp': -7.7353515625, 'zcp_jacov': -16.06437590653865, 'zcp_l2_norm': 1439.2486572265625, 'zcp_nwot': 237.88553378687146, 'zcp_params': 15037834.0, 'zcp_plain': -0.0005952141364100001, 'zcp_snip': 1901.4063720703125, 'zcp_synflow': 120.26646383656575, 'zcp_zen': 117.07764434814453, 'zcp_val_accuracy': 0.904947936534881}
| |
NASBench101_336583
|
NASBench101
|
336583
|
cb8815d2de1129d7076c8f13ad622336
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_992[FLOAT, 128x3x3x3]
%onnx::Conv_993[FLOAT, 128]
%onnx::Conv_995[FLOAT, 43x128x1x1]
%onnx::Conv_996[FLOAT, 43]
%onnx::Conv_998[FLOAT, 43x43x3x3]
%onnx::Conv_1001[FLOAT, 43x43x1x1]
%onnx::Conv_1004[FLOAT, 42x128x1x1]
%onnx::Conv_1005[FLOAT, 42]
%onnx::Conv_1007[FLOAT, 42x42x3x3]
%onnx::Conv_1010[FLOAT, 42x42x3x3]
%onnx::Conv_1013[FLOAT, 42x128x1x1]
%onnx::Conv_1016[FLOAT, 43x128x1x1]
%onnx::Conv_1019[FLOAT, 43x43x3x3]
%onnx::Conv_1022[FLOAT, 43x43x1x1]
%onnx::Conv_1025[FLOAT, 42x128x1x1]
%onnx::Conv_1028[FLOAT, 42x42x3x3]
%onnx::Conv_1031[FLOAT, 42x42x3x3]
%onnx::Conv_1034[FLOAT, 42x128x1x1]
%onnx::Conv_1037[FLOAT, 43x128x1x1]
%onnx::Conv_1040[FLOAT, 43x43x3x3]
%onnx::Conv_1043[FLOAT, 43x43x1x1]
%onnx::Conv_1046[FLOAT, 42x128x1x1]
%onnx::Conv_1049[FLOAT, 42x42x3x3]
%onnx::Conv_1052[FLOAT, 42x42x3x3]
%onnx::Conv_1055[FLOAT, 42x128x1x1]
%onnx::Conv_1058[FLOAT, 86x128x1x1]
%onnx::Conv_1059[FLOAT, 86]
%onnx::Conv_1061[FLOAT, 86x86x3x3]
%onnx::Conv_1064[FLOAT, 85x85x1x1]
%onnx::Conv_1065[FLOAT, 85]
%onnx::Conv_1067[FLOAT, 85x128x1x1]
%onnx::Conv_1070[FLOAT, 85x85x3x3]
%onnx::Conv_1073[FLOAT, 85x85x3x3]
%onnx::Conv_1076[FLOAT, 85x128x1x1]
%onnx::Conv_1079[FLOAT, 86x256x1x1]
%onnx::Conv_1082[FLOAT, 86x86x3x3]
%onnx::Conv_1085[FLOAT, 85x85x1x1]
%onnx::Conv_1088[FLOAT, 85x256x1x1]
%onnx::Conv_1091[FLOAT, 85x85x3x3]
%onnx::Conv_1094[FLOAT, 85x85x3x3]
%onnx::Conv_1097[FLOAT, 85x256x1x1]
%onnx::Conv_1100[FLOAT, 86x256x1x1]
%onnx::Conv_1103[FLOAT, 86x86x3x3]
%onnx::Conv_1106[FLOAT, 85x85x1x1]
%onnx::Conv_1109[FLOAT, 85x256x1x1]
%onnx::Conv_1112[FLOAT, 85x85x3x3]
%onnx::Conv_1115[FLOAT, 85x85x3x3]
%onnx::Conv_1118[FLOAT, 85x256x1x1]
%onnx::Conv_1121[FLOAT, 171x256x1x1]
%onnx::Conv_1122[FLOAT, 171]
%onnx::Conv_1124[FLOAT, 171x171x3x3]
%onnx::Conv_1127[FLOAT, 171x171x1x1]
%onnx::Conv_1130[FLOAT, 170x256x1x1]
%onnx::Conv_1131[FLOAT, 170]
%onnx::Conv_1133[FLOAT, 170x170x3x3]
%onnx::Conv_1136[FLOAT, 170x170x3x3]
%onnx::Conv_1139[FLOAT, 170x256x1x1]
%onnx::Conv_1142[FLOAT, 171x512x1x1]
%onnx::Conv_1145[FLOAT, 171x171x3x3]
%onnx::Conv_1148[FLOAT, 171x171x1x1]
%onnx::Conv_1151[FLOAT, 170x512x1x1]
%onnx::Conv_1154[FLOAT, 170x170x3x3]
%onnx::Conv_1157[FLOAT, 170x170x3x3]
%onnx::Conv_1160[FLOAT, 170x512x1x1]
%onnx::Conv_1163[FLOAT, 171x512x1x1]
%onnx::Conv_1166[FLOAT, 171x171x3x3]
%onnx::Conv_1169[FLOAT, 171x171x1x1]
%onnx::Conv_1172[FLOAT, 170x512x1x1]
%onnx::Conv_1175[FLOAT, 170x170x3x3]
%onnx::Conv_1178[FLOAT, 170x170x3x3]
%onnx::Conv_1181[FLOAT, 170x512x1x1]
) {
%onnx::Conv_1182 = Identity(%onnx::Conv_1131)
%onnx::Conv_1179 = Identity(%onnx::Conv_1131)
%onnx::Conv_1176 = Identity(%onnx::Conv_1131)
%onnx::Conv_1173 = Identity(%onnx::Conv_1131)
%onnx::Conv_1170 = Identity(%onnx::Conv_1122)
%onnx::Conv_1167 = Identity(%onnx::Conv_1122)
%onnx::Conv_1164 = Identity(%onnx::Conv_1122)
%onnx::Conv_1161 = Identity(%onnx::Conv_1131)
%onnx::Conv_1158 = Identity(%onnx::Conv_1131)
%onnx::Conv_1155 = Identity(%onnx::Conv_1131)
%onnx::Conv_1152 = Identity(%onnx::Conv_1131)
%onnx::Conv_1149 = Identity(%onnx::Conv_1122)
%onnx::Conv_1146 = Identity(%onnx::Conv_1122)
%onnx::Conv_1143 = Identity(%onnx::Conv_1122)
%onnx::Conv_1140 = Identity(%onnx::Conv_1131)
%onnx::Conv_1137 = Identity(%onnx::Conv_1131)
%onnx::Conv_1134 = Identity(%onnx::Conv_1131)
%onnx::Conv_1128 = Identity(%onnx::Conv_1122)
%onnx::Conv_1125 = Identity(%onnx::Conv_1122)
%onnx::Conv_1119 = Identity(%onnx::Conv_1065)
%onnx::Conv_1116 = Identity(%onnx::Conv_1065)
%onnx::Conv_1113 = Identity(%onnx::Conv_1065)
%onnx::Conv_1110 = Identity(%onnx::Conv_1065)
%onnx::Conv_1107 = Identity(%onnx::Conv_1065)
%onnx::Conv_1104 = Identity(%onnx::Conv_1059)
%onnx::Conv_1101 = Identity(%onnx::Conv_1059)
%onnx::Conv_1098 = Identity(%onnx::Conv_1065)
%onnx::Conv_1095 = Identity(%onnx::Conv_1065)
%onnx::Conv_1092 = Identity(%onnx::Conv_1065)
%onnx::Conv_1089 = Identity(%onnx::Conv_1065)
%onnx::Conv_1086 = Identity(%onnx::Conv_1065)
%onnx::Conv_1083 = Identity(%onnx::Conv_1059)
%onnx::Conv_1080 = Identity(%onnx::Conv_1059)
%onnx::Conv_1077 = Identity(%onnx::Conv_1065)
%onnx::Conv_1074 = Identity(%onnx::Conv_1065)
%onnx::Conv_1071 = Identity(%onnx::Conv_1065)
%onnx::Conv_1068 = Identity(%onnx::Conv_1065)
%onnx::Conv_1062 = Identity(%onnx::Conv_1059)
%onnx::Conv_1056 = Identity(%onnx::Conv_1005)
%onnx::Conv_1053 = Identity(%onnx::Conv_1005)
%onnx::Conv_1050 = Identity(%onnx::Conv_1005)
%onnx::Conv_1047 = Identity(%onnx::Conv_1005)
%onnx::Conv_1044 = Identity(%onnx::Conv_996)
%onnx::Conv_1041 = Identity(%onnx::Conv_996)
%onnx::Conv_1038 = Identity(%onnx::Conv_996)
%onnx::Conv_1035 = Identity(%onnx::Conv_1005)
%onnx::Conv_1032 = Identity(%onnx::Conv_1005)
%onnx::Conv_1029 = Identity(%onnx::Conv_1005)
%onnx::Conv_1026 = Identity(%onnx::Conv_1005)
%onnx::Conv_1023 = Identity(%onnx::Conv_996)
%onnx::Conv_1020 = Identity(%onnx::Conv_996)
%onnx::Conv_1017 = Identity(%onnx::Conv_996)
%onnx::Conv_1014 = Identity(%onnx::Conv_1005)
%onnx::Conv_1011 = Identity(%onnx::Conv_1005)
%onnx::Conv_1008 = Identity(%onnx::Conv_1005)
%onnx::Conv_1002 = Identity(%onnx::Conv_996)
%onnx::Conv_999 = Identity(%onnx::Conv_996)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_992, %onnx::Conv_993)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_4_output_0)
%/layers.1/Add_5_output_0 = Add(%/layers.1/Add_4_output_0, %/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_5_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_1025, %onnx::Conv_1026)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_1028, %onnx::Conv_1029)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_1031, %onnx::Conv_1032)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_1034, %onnx::Conv_1035)
%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_4_output_0)
%/layers.2/Add_5_output_0 = Add(%/layers.2/Add_4_output_0, %/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_5_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_1037, %onnx::Conv_1038)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_1040, %onnx::Conv_1041)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_1043, %onnx::Conv_1044)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_1046, %onnx::Conv_1047)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_1049, %onnx::Conv_1050)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_1052, %onnx::Conv_1053)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_1055, %onnx::Conv_1056)
%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_4_output_0)
%/layers.3/Add_5_output_0 = Add(%/layers.3/Add_4_output_0, %/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_5_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1058, %onnx::Conv_1059)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_1061, %onnx::Conv_1062)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.5/Slice_output_0 = Slice(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0, %/layers.5/Constant_3_output_0, %/layers.5/Constant_1_output_0, %/layers.5/Constant_4_output_0)
%/layers.5/Constant_5_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/Slice_output_0, %/layers.5/Constant_5_output_0)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_1064, %onnx::Conv_1065)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1067, %onnx::Conv_1068)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_6_output_0)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_1070, %onnx::Conv_1071)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_7_output_0)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_1073, %onnx::Conv_1074)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1076, %onnx::Conv_1077)
%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_8_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_8_output_0)
%/layers.5/Add_5_output_0 = Add(%/layers.5/Add_4_output_0, %/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_5_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_1079, %onnx::Conv_1080)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_1082, %onnx::Conv_1083)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.6/Slice_output_0 = Slice(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0, %/layers.6/Constant_3_output_0, %/layers.6/Constant_1_output_0, %/layers.6/Constant_4_output_0)
%/layers.6/Constant_5_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/Slice_output_0, %/layers.6/Constant_5_output_0)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_1085, %onnx::Conv_1086)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_1088, %onnx::Conv_1089)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_6_output_0)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_1091, %onnx::Conv_1092)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_7_output_0)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_1094, %onnx::Conv_1095)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_1097, %onnx::Conv_1098)
%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_8_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_8_output_0)
%/layers.6/Add_5_output_0 = Add(%/layers.6/Add_4_output_0, %/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_5_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_1100, %onnx::Conv_1101)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_1103, %onnx::Conv_1104)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.7/Slice_output_0 = Slice(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0, %/layers.7/Constant_3_output_0, %/layers.7/Constant_1_output_0, %/layers.7/Constant_4_output_0)
%/layers.7/Constant_5_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/Slice_output_0, %/layers.7/Constant_5_output_0)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_1106, %onnx::Conv_1107)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_1109, %onnx::Conv_1110)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_6_output_0)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_1112, %onnx::Conv_1113)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_7_output_0)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_1115, %onnx::Conv_1116)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_1118, %onnx::Conv_1119)
%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_8_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_8_output_0)
%/layers.7/Add_5_output_0 = Add(%/layers.7/Add_4_output_0, %/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_5_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1121, %onnx::Conv_1122)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_1124, %onnx::Conv_1125)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_1127, %onnx::Conv_1128)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1130, %onnx::Conv_1131)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_1133, %onnx::Conv_1134)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_1136, %onnx::Conv_1137)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1139, %onnx::Conv_1140)
%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_4_output_0)
%/layers.9/Add_5_output_0 = Add(%/layers.9/Add_4_output_0, %/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_5_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_1142, %onnx::Conv_1143)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_1145, %onnx::Conv_1146)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_1148, %onnx::Conv_1149)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_1151, %onnx::Conv_1152)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_1154, %onnx::Conv_1155)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_1157, %onnx::Conv_1158)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_1160, %onnx::Conv_1161)
%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_4_output_0)
%/layers.10/Add_5_output_0 = Add(%/layers.10/Add_4_output_0, %/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_5_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1163, %onnx::Conv_1164)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_1166, %onnx::Conv_1167)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_1169, %onnx::Conv_1170)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1172, %onnx::Conv_1173)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_1175, %onnx::Conv_1176)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_1178, %onnx::Conv_1179)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1181, %onnx::Conv_1182)
%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_4_output_0)
%/layers.11/Add_5_output_0 = Add(%/layers.11/Add_4_output_0, %/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_5_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.5/maxpool/MaxPool_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%990 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %990
}
|
val_accuracy
| 92.207533
| 1,222,247,168
| 4,088,125
|
{'zcp_epe_nas': 105.72980622442053, 'zcp_fisher': 25.697580337524414, 'zcp_flops': 19555954688.0, 'zcp_grad_norm': 111.59355926513672, 'zcp_grasp': -13.3509521484375, 'zcp_jacov': -16.04988236891156, 'zcp_l2_norm': 1080.63427734375, 'zcp_nwot': 220.82925319110575, 'zcp_params': 4088125.0, 'zcp_plain': 0.065088488161563, 'zcp_snip': 605.5975952148438, 'zcp_synflow': 91.27179508055167, 'zcp_zen': 109.3454360961914, 'zcp_val_accuracy': 0.931390225887298}
| |
NASBench101_176398
|
NASBench101
|
176398
|
6acf6388b79c183b025638056db1fcbd
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_860[FLOAT, 128x3x3x3]
%onnx::Conv_861[FLOAT, 128]
%onnx::Conv_863[FLOAT, 128x128x1x1]
%onnx::Conv_866[FLOAT, 128x128x1x1]
%onnx::Conv_869[FLOAT, 128x128x3x3]
%onnx::Conv_872[FLOAT, 128x128x1x1]
%onnx::Conv_875[FLOAT, 128x128x1x1]
%onnx::Conv_878[FLOAT, 128x128x1x1]
%onnx::Conv_881[FLOAT, 128x128x1x1]
%onnx::Conv_884[FLOAT, 128x128x1x1]
%onnx::Conv_887[FLOAT, 128x128x3x3]
%onnx::Conv_890[FLOAT, 128x128x1x1]
%onnx::Conv_893[FLOAT, 128x128x1x1]
%onnx::Conv_896[FLOAT, 128x128x1x1]
%onnx::Conv_899[FLOAT, 128x128x1x1]
%onnx::Conv_902[FLOAT, 128x128x1x1]
%onnx::Conv_905[FLOAT, 128x128x3x3]
%onnx::Conv_908[FLOAT, 128x128x1x1]
%onnx::Conv_911[FLOAT, 128x128x1x1]
%onnx::Conv_914[FLOAT, 128x128x1x1]
%onnx::Conv_917[FLOAT, 256x128x1x1]
%onnx::Conv_918[FLOAT, 256]
%onnx::Conv_920[FLOAT, 256x128x1x1]
%onnx::Conv_923[FLOAT, 256x256x3x3]
%onnx::Conv_926[FLOAT, 256x128x1x1]
%onnx::Conv_929[FLOAT, 256x256x1x1]
%onnx::Conv_932[FLOAT, 256x256x1x1]
%onnx::Conv_935[FLOAT, 256x256x1x1]
%onnx::Conv_938[FLOAT, 256x256x1x1]
%onnx::Conv_941[FLOAT, 256x256x3x3]
%onnx::Conv_944[FLOAT, 256x256x1x1]
%onnx::Conv_947[FLOAT, 256x256x1x1]
%onnx::Conv_950[FLOAT, 256x256x1x1]
%onnx::Conv_953[FLOAT, 256x256x1x1]
%onnx::Conv_956[FLOAT, 256x256x1x1]
%onnx::Conv_959[FLOAT, 256x256x3x3]
%onnx::Conv_962[FLOAT, 256x256x1x1]
%onnx::Conv_965[FLOAT, 256x256x1x1]
%onnx::Conv_968[FLOAT, 256x256x1x1]
%onnx::Conv_971[FLOAT, 512x256x1x1]
%onnx::Conv_972[FLOAT, 512]
%onnx::Conv_974[FLOAT, 512x256x1x1]
%onnx::Conv_977[FLOAT, 512x512x3x3]
%onnx::Conv_980[FLOAT, 512x256x1x1]
%onnx::Conv_983[FLOAT, 512x512x1x1]
%onnx::Conv_986[FLOAT, 512x512x1x1]
%onnx::Conv_989[FLOAT, 512x512x1x1]
%onnx::Conv_992[FLOAT, 512x512x1x1]
%onnx::Conv_995[FLOAT, 512x512x3x3]
%onnx::Conv_998[FLOAT, 512x512x1x1]
%onnx::Conv_1001[FLOAT, 512x512x1x1]
%onnx::Conv_1004[FLOAT, 512x512x1x1]
%onnx::Conv_1007[FLOAT, 512x512x1x1]
%onnx::Conv_1010[FLOAT, 512x512x1x1]
%onnx::Conv_1013[FLOAT, 512x512x3x3]
%onnx::Conv_1016[FLOAT, 512x512x1x1]
%onnx::Conv_1019[FLOAT, 512x512x1x1]
%onnx::Conv_1022[FLOAT, 512x512x1x1]
) {
%onnx::Conv_1023 = Identity(%onnx::Conv_972)
%onnx::Conv_1020 = Identity(%onnx::Conv_972)
%onnx::Conv_1017 = Identity(%onnx::Conv_972)
%onnx::Conv_1014 = Identity(%onnx::Conv_972)
%onnx::Conv_1011 = Identity(%onnx::Conv_972)
%onnx::Conv_1008 = Identity(%onnx::Conv_972)
%onnx::Conv_1005 = Identity(%onnx::Conv_972)
%onnx::Conv_1002 = Identity(%onnx::Conv_972)
%onnx::Conv_999 = Identity(%onnx::Conv_972)
%onnx::Conv_996 = Identity(%onnx::Conv_972)
%onnx::Conv_993 = Identity(%onnx::Conv_972)
%onnx::Conv_990 = Identity(%onnx::Conv_972)
%onnx::Conv_987 = Identity(%onnx::Conv_972)
%onnx::Conv_984 = Identity(%onnx::Conv_972)
%onnx::Conv_981 = Identity(%onnx::Conv_972)
%onnx::Conv_978 = Identity(%onnx::Conv_972)
%onnx::Conv_975 = Identity(%onnx::Conv_972)
%onnx::Conv_969 = Identity(%onnx::Conv_918)
%onnx::Conv_966 = Identity(%onnx::Conv_918)
%onnx::Conv_963 = Identity(%onnx::Conv_918)
%onnx::Conv_960 = Identity(%onnx::Conv_918)
%onnx::Conv_957 = Identity(%onnx::Conv_918)
%onnx::Conv_954 = Identity(%onnx::Conv_918)
%onnx::Conv_951 = Identity(%onnx::Conv_918)
%onnx::Conv_948 = Identity(%onnx::Conv_918)
%onnx::Conv_945 = Identity(%onnx::Conv_918)
%onnx::Conv_942 = Identity(%onnx::Conv_918)
%onnx::Conv_939 = Identity(%onnx::Conv_918)
%onnx::Conv_936 = Identity(%onnx::Conv_918)
%onnx::Conv_933 = Identity(%onnx::Conv_918)
%onnx::Conv_930 = Identity(%onnx::Conv_918)
%onnx::Conv_927 = Identity(%onnx::Conv_918)
%onnx::Conv_924 = Identity(%onnx::Conv_918)
%onnx::Conv_921 = Identity(%onnx::Conv_918)
%onnx::Conv_915 = Identity(%onnx::Conv_861)
%onnx::Conv_912 = Identity(%onnx::Conv_861)
%onnx::Conv_909 = Identity(%onnx::Conv_861)
%onnx::Conv_906 = Identity(%onnx::Conv_861)
%onnx::Conv_903 = Identity(%onnx::Conv_861)
%onnx::Conv_900 = Identity(%onnx::Conv_861)
%onnx::Conv_897 = Identity(%onnx::Conv_861)
%onnx::Conv_894 = Identity(%onnx::Conv_861)
%onnx::Conv_891 = Identity(%onnx::Conv_861)
%onnx::Conv_888 = Identity(%onnx::Conv_861)
%onnx::Conv_885 = Identity(%onnx::Conv_861)
%onnx::Conv_882 = Identity(%onnx::Conv_861)
%onnx::Conv_879 = Identity(%onnx::Conv_861)
%onnx::Conv_876 = Identity(%onnx::Conv_861)
%onnx::Conv_873 = Identity(%onnx::Conv_861)
%onnx::Conv_870 = Identity(%onnx::Conv_861)
%onnx::Conv_867 = Identity(%onnx::Conv_861)
%onnx::Conv_864 = Identity(%onnx::Conv_861)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_860, %onnx::Conv_861)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/Add_2_output_0, %/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_4_output_0)
%/layers.1/Add_5_output_0 = Add(%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_5_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/Add_2_output_0, %/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_4_output_0)
%/layers.2/Add_5_output_0 = Add(%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_5_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_908, %onnx::Conv_909)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/Add_2_output_0, %/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_911, %onnx::Conv_912)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_4_output_0)
%/layers.3/Add_5_output_0 = Add(%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_5_output_0, %onnx::Conv_914, %onnx::Conv_915)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_917, %onnx::Conv_918)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_920, %onnx::Conv_921)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_923, %onnx::Conv_924)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_926, %onnx::Conv_927)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/Add_2_output_0, %/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_929, %onnx::Conv_930)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_4_output_0)
%/layers.5/Add_5_output_0 = Add(%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_5_output_0, %onnx::Conv_932, %onnx::Conv_933)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_935, %onnx::Conv_936)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_938, %onnx::Conv_939)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_941, %onnx::Conv_942)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_944, %onnx::Conv_945)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/Add_2_output_0, %/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_947, %onnx::Conv_948)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_4_output_0)
%/layers.6/Add_5_output_0 = Add(%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_5_output_0, %onnx::Conv_950, %onnx::Conv_951)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_953, %onnx::Conv_954)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_956, %onnx::Conv_957)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_959, %onnx::Conv_960)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_962, %onnx::Conv_963)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/Add_2_output_0, %/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_965, %onnx::Conv_966)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_4_output_0)
%/layers.7/Add_5_output_0 = Add(%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_5_output_0, %onnx::Conv_968, %onnx::Conv_969)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_974, %onnx::Conv_975)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_977, %onnx::Conv_978)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/Add_2_output_0, %/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_4_output_0)
%/layers.9/Add_5_output_0 = Add(%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_5_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/Add_2_output_0, %/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_4_output_0)
%/layers.10/Add_5_output_0 = Add(%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_5_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/Add_2_output_0, %/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_4_output_0)
%/layers.11/Add_5_output_0 = Add(%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_5_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%858 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %858
}
|
val_accuracy
| 91.636616
| 4,168,361,984
| 14,000,266
|
{'zcp_epe_nas': 167.56945857959676, 'zcp_fisher': 70.9063491821289, 'zcp_flops': 66693791744.0, 'zcp_grad_norm': 155.15895080566406, 'zcp_grasp': 14.897705078125, 'zcp_jacov': -16.06120421147319, 'zcp_l2_norm': 1225.9698486328125, 'zcp_nwot': 235.06020383569407, 'zcp_params': 14000266.0, 'zcp_plain': -0.001722998684272, 'zcp_snip': 1216.041748046875, 'zcp_synflow': 126.22126129593558, 'zcp_zen': 106.36054229736328, 'zcp_val_accuracy': 0.925380587577819}
| |
NASBench101_285767
|
NASBench101
|
285767
|
acfa4bb639bb975d5f26362c97e2e992
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_761[FLOAT, 128x3x3x3]
%onnx::Conv_762[FLOAT, 128]
%onnx::Conv_764[FLOAT, 128x128x1x1]
%onnx::Conv_767[FLOAT, 128x128x3x3]
%onnx::Conv_770[FLOAT, 128x128x3x3]
%onnx::Conv_773[FLOAT, 128x128x1x1]
%onnx::Conv_776[FLOAT, 128x128x1x1]
%onnx::Conv_779[FLOAT, 128x128x1x1]
%onnx::Conv_782[FLOAT, 128x128x3x3]
%onnx::Conv_785[FLOAT, 128x128x3x3]
%onnx::Conv_788[FLOAT, 128x128x1x1]
%onnx::Conv_791[FLOAT, 128x128x1x1]
%onnx::Conv_794[FLOAT, 128x128x1x1]
%onnx::Conv_797[FLOAT, 128x128x3x3]
%onnx::Conv_800[FLOAT, 128x128x3x3]
%onnx::Conv_803[FLOAT, 128x128x1x1]
%onnx::Conv_806[FLOAT, 128x128x1x1]
%onnx::Conv_809[FLOAT, 256x128x1x1]
%onnx::Conv_810[FLOAT, 256]
%onnx::Conv_812[FLOAT, 256x256x3x3]
%onnx::Conv_815[FLOAT, 256x256x3x3]
%onnx::Conv_818[FLOAT, 256x128x1x1]
%onnx::Conv_821[FLOAT, 256x256x1x1]
%onnx::Conv_824[FLOAT, 256x256x1x1]
%onnx::Conv_827[FLOAT, 256x256x3x3]
%onnx::Conv_830[FLOAT, 256x256x3x3]
%onnx::Conv_833[FLOAT, 256x256x1x1]
%onnx::Conv_836[FLOAT, 256x256x1x1]
%onnx::Conv_839[FLOAT, 256x256x1x1]
%onnx::Conv_842[FLOAT, 256x256x3x3]
%onnx::Conv_845[FLOAT, 256x256x3x3]
%onnx::Conv_848[FLOAT, 256x256x1x1]
%onnx::Conv_851[FLOAT, 256x256x1x1]
%onnx::Conv_854[FLOAT, 512x256x1x1]
%onnx::Conv_855[FLOAT, 512]
%onnx::Conv_857[FLOAT, 512x512x3x3]
%onnx::Conv_860[FLOAT, 512x512x3x3]
%onnx::Conv_863[FLOAT, 512x256x1x1]
%onnx::Conv_866[FLOAT, 512x512x1x1]
%onnx::Conv_869[FLOAT, 512x512x1x1]
%onnx::Conv_872[FLOAT, 512x512x3x3]
%onnx::Conv_875[FLOAT, 512x512x3x3]
%onnx::Conv_878[FLOAT, 512x512x1x1]
%onnx::Conv_881[FLOAT, 512x512x1x1]
%onnx::Conv_884[FLOAT, 512x512x1x1]
%onnx::Conv_887[FLOAT, 512x512x3x3]
%onnx::Conv_890[FLOAT, 512x512x3x3]
%onnx::Conv_893[FLOAT, 512x512x1x1]
%onnx::Conv_896[FLOAT, 512x512x1x1]
) {
%onnx::Conv_897 = Identity(%onnx::Conv_855)
%onnx::Conv_894 = Identity(%onnx::Conv_855)
%onnx::Conv_891 = Identity(%onnx::Conv_855)
%onnx::Conv_888 = Identity(%onnx::Conv_855)
%onnx::Conv_885 = Identity(%onnx::Conv_855)
%onnx::Conv_882 = Identity(%onnx::Conv_855)
%onnx::Conv_879 = Identity(%onnx::Conv_855)
%onnx::Conv_876 = Identity(%onnx::Conv_855)
%onnx::Conv_873 = Identity(%onnx::Conv_855)
%onnx::Conv_870 = Identity(%onnx::Conv_855)
%onnx::Conv_867 = Identity(%onnx::Conv_855)
%onnx::Conv_864 = Identity(%onnx::Conv_855)
%onnx::Conv_861 = Identity(%onnx::Conv_855)
%onnx::Conv_858 = Identity(%onnx::Conv_855)
%onnx::Conv_852 = Identity(%onnx::Conv_810)
%onnx::Conv_849 = Identity(%onnx::Conv_810)
%onnx::Conv_846 = Identity(%onnx::Conv_810)
%onnx::Conv_843 = Identity(%onnx::Conv_810)
%onnx::Conv_840 = Identity(%onnx::Conv_810)
%onnx::Conv_837 = Identity(%onnx::Conv_810)
%onnx::Conv_834 = Identity(%onnx::Conv_810)
%onnx::Conv_831 = Identity(%onnx::Conv_810)
%onnx::Conv_828 = Identity(%onnx::Conv_810)
%onnx::Conv_825 = Identity(%onnx::Conv_810)
%onnx::Conv_822 = Identity(%onnx::Conv_810)
%onnx::Conv_819 = Identity(%onnx::Conv_810)
%onnx::Conv_816 = Identity(%onnx::Conv_810)
%onnx::Conv_813 = Identity(%onnx::Conv_810)
%onnx::Conv_807 = Identity(%onnx::Conv_762)
%onnx::Conv_804 = Identity(%onnx::Conv_762)
%onnx::Conv_801 = Identity(%onnx::Conv_762)
%onnx::Conv_798 = Identity(%onnx::Conv_762)
%onnx::Conv_795 = Identity(%onnx::Conv_762)
%onnx::Conv_792 = Identity(%onnx::Conv_762)
%onnx::Conv_789 = Identity(%onnx::Conv_762)
%onnx::Conv_786 = Identity(%onnx::Conv_762)
%onnx::Conv_783 = Identity(%onnx::Conv_762)
%onnx::Conv_780 = Identity(%onnx::Conv_762)
%onnx::Conv_777 = Identity(%onnx::Conv_762)
%onnx::Conv_774 = Identity(%onnx::Conv_762)
%onnx::Conv_771 = Identity(%onnx::Conv_762)
%onnx::Conv_768 = Identity(%onnx::Conv_762)
%onnx::Conv_765 = Identity(%onnx::Conv_762)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_761, %onnx::Conv_762)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_764, %onnx::Conv_765)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_767, %onnx::Conv_768)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/Add_2_output_0, %/layers.1/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_770, %onnx::Conv_771)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_773, %onnx::Conv_774)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/Add_5_output_0 = Add(%/layers.1/Add_4_output_0, %/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_5_output_0)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_776, %onnx::Conv_777)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_779, %onnx::Conv_780)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_782, %onnx::Conv_783)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/Add_2_output_0, %/layers.2/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_785, %onnx::Conv_786)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_788, %onnx::Conv_789)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/Add_5_output_0 = Add(%/layers.2/Add_4_output_0, %/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_5_output_0)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_797, %onnx::Conv_798)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/Add_2_output_0, %/layers.3/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_800, %onnx::Conv_801)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_803, %onnx::Conv_804)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/Add_5_output_0 = Add(%/layers.3/Add_4_output_0, %/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_5_output_0)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_806, %onnx::Conv_807)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_809, %onnx::Conv_810)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_812, %onnx::Conv_813)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/Add_2_output_0, %/layers.5/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_815, %onnx::Conv_816)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_818, %onnx::Conv_819)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/Add_5_output_0 = Add(%/layers.5/Add_4_output_0, %/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_5_output_0)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_821, %onnx::Conv_822)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_824, %onnx::Conv_825)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_827, %onnx::Conv_828)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/Add_2_output_0, %/layers.6/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_830, %onnx::Conv_831)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_833, %onnx::Conv_834)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/Add_5_output_0 = Add(%/layers.6/Add_4_output_0, %/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_5_output_0)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_836, %onnx::Conv_837)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_839, %onnx::Conv_840)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_842, %onnx::Conv_843)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/Add_2_output_0, %/layers.7/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_845, %onnx::Conv_846)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_848, %onnx::Conv_849)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/Add_5_output_0 = Add(%/layers.7/Add_4_output_0, %/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_5_output_0)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_851, %onnx::Conv_852)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_854, %onnx::Conv_855)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_857, %onnx::Conv_858)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/Add_2_output_0, %/layers.9/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/Add_5_output_0 = Add(%/layers.9/Add_4_output_0, %/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_5_output_0)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/Add_2_output_0, %/layers.10/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/Add_5_output_0 = Add(%/layers.10/Add_4_output_0, %/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_5_output_0)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/Add_2_output_0, %/layers.11/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/Add_5_output_0 = Add(%/layers.11/Add_4_output_0, %/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_5_output_0)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%759 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %759
}
|
val_accuracy
| 91.095752
| 6,310,340,608
| 21,384,074
|
{'zcp_epe_nas': 97.3582546269004, 'zcp_fisher': 254.49839782714844, 'zcp_flops': 100965449728.0, 'zcp_grad_norm': 232.07672119140625, 'zcp_grasp': -58.0830078125, 'zcp_jacov': -16.046349686019305, 'zcp_l2_norm': 1030.20556640625, 'zcp_nwot': 232.07086146803778, 'zcp_params': 21384074.0, 'zcp_plain': 0.041091822087764004, 'zcp_snip': 1971.3603515625, 'zcp_synflow': 136.05953502180304, 'zcp_zen': 103.12381744384766, 'zcp_val_accuracy': 0.9405047893524171}
| |
NASBench101_331675
|
NASBench101
|
331675
|
c8a33bf8667a5733e6ddf66e46eb621b
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_437[FLOAT, 128x3x3x3]
%onnx::Conv_438[FLOAT, 128]
%onnx::Conv_440[FLOAT, 64x128x1x1]
%onnx::Conv_441[FLOAT, 64]
%onnx::Conv_443[FLOAT, 128x128x1x1]
%onnx::Conv_446[FLOAT, 64x128x1x1]
%onnx::Conv_449[FLOAT, 128x128x1x1]
%onnx::Conv_452[FLOAT, 64x128x1x1]
%onnx::Conv_455[FLOAT, 128x128x1x1]
%onnx::Conv_458[FLOAT, 128x128x1x1]
%onnx::Conv_461[FLOAT, 256x128x1x1]
%onnx::Conv_462[FLOAT, 256]
%onnx::Conv_464[FLOAT, 128x256x1x1]
%onnx::Conv_467[FLOAT, 256x256x1x1]
%onnx::Conv_470[FLOAT, 128x256x1x1]
%onnx::Conv_473[FLOAT, 256x256x1x1]
%onnx::Conv_476[FLOAT, 256x256x1x1]
%onnx::Conv_479[FLOAT, 512x256x1x1]
%onnx::Conv_480[FLOAT, 512]
%onnx::Conv_482[FLOAT, 256x512x1x1]
%onnx::Conv_485[FLOAT, 512x512x1x1]
%onnx::Conv_488[FLOAT, 256x512x1x1]
%onnx::Conv_491[FLOAT, 512x512x1x1]
) {
%onnx::Conv_492 = Identity(%onnx::Conv_480)
%onnx::Conv_489 = Identity(%onnx::Conv_462)
%onnx::Conv_486 = Identity(%onnx::Conv_480)
%onnx::Conv_483 = Identity(%onnx::Conv_462)
%onnx::Conv_477 = Identity(%onnx::Conv_462)
%onnx::Conv_474 = Identity(%onnx::Conv_462)
%onnx::Conv_471 = Identity(%onnx::Conv_438)
%onnx::Conv_468 = Identity(%onnx::Conv_462)
%onnx::Conv_465 = Identity(%onnx::Conv_438)
%onnx::Conv_459 = Identity(%onnx::Conv_438)
%onnx::Conv_456 = Identity(%onnx::Conv_438)
%onnx::Conv_453 = Identity(%onnx::Conv_441)
%onnx::Conv_450 = Identity(%onnx::Conv_438)
%onnx::Conv_447 = Identity(%onnx::Conv_441)
%onnx::Conv_444 = Identity(%onnx::Conv_438)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_437, %onnx::Conv_438)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_440, %onnx::Conv_441)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.1/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.4/maxpool/MaxPool_output_0, %/layers.1/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_443, %onnx::Conv_444)
%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_1_output_0 = Add(%/layers.1/Concat_output_0, %/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_446, %onnx::Conv_447)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.2/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.4/maxpool/MaxPool_output_0, %/layers.2/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_449, %onnx::Conv_450)
%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_1_output_0 = Add(%/layers.2/Concat_output_0, %/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_452, %onnx::Conv_453)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.3/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.4/maxpool/MaxPool_output_0, %/layers.3/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_455, %onnx::Conv_456)
%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_1_output_0 = Add(%/layers.3/Concat_output_0, %/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Add_1_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_458, %onnx::Conv_459)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.5/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.4/maxpool/MaxPool_output_0, %/layers.5/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_461, %onnx::Conv_462)
%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_1_output_0 = Add(%/layers.5/Concat_output_0, %/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_464, %onnx::Conv_465)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.6/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.4/maxpool/MaxPool_output_0, %/layers.6/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_467, %onnx::Conv_468)
%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_1_output_0 = Add(%/layers.6/Concat_output_0, %/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_470, %onnx::Conv_471)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.7/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.4/maxpool/MaxPool_output_0, %/layers.7/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_473, %onnx::Conv_474)
%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_1_output_0 = Add(%/layers.7/Concat_output_0, %/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Add_1_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_476, %onnx::Conv_477)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.9/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.4/maxpool/MaxPool_output_0, %/layers.9/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_479, %onnx::Conv_480)
%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_1_output_0 = Add(%/layers.9/Concat_output_0, %/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_482, %onnx::Conv_483)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.10/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.4/maxpool/MaxPool_output_0, %/layers.10/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_485, %onnx::Conv_486)
%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_1_output_0 = Add(%/layers.10/Concat_output_0, %/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_488, %onnx::Conv_489)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.11/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.4/maxpool/MaxPool_output_0, %/layers.11/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_491, %onnx::Conv_492)
%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_1_output_0 = Add(%/layers.11/Concat_output_0, %/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Add_1_output_0)
%435 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %435
}
|
val_accuracy
| 86.848956
| 419,047,424
| 1,319,434
|
{'zcp_epe_nas': 68.39553753491911, 'zcp_fisher': 4.524311065673828, 'zcp_flops': 6704758784.0, 'zcp_grad_norm': 36.17601776123047, 'zcp_grasp': -11.202255249023438, 'zcp_jacov': -16.05637519592221, 'zcp_l2_norm': 395.0531921386719, 'zcp_nwot': 213.96839756191235, 'zcp_params': 1319434.0, 'zcp_plain': 0.170478835701942, 'zcp_snip': 206.3662109375, 'zcp_synflow': 38.59353660493142, 'zcp_zen': 38.38414001464844, 'zcp_val_accuracy': 0.8663862347602841}
| |
NASBench101_299241
|
NASBench101
|
299241
|
b5140457757b012fbcf9ebe5719a8f12
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_698[FLOAT, 128x3x3x3]
%onnx::Conv_699[FLOAT, 128]
%onnx::Conv_701[FLOAT, 43x128x1x1]
%onnx::Conv_702[FLOAT, 43]
%onnx::Conv_704[FLOAT, 43x43x1x1]
%onnx::Conv_707[FLOAT, 43x128x1x1]
%onnx::Conv_710[FLOAT, 128x128x1x1]
%onnx::Conv_713[FLOAT, 43x128x1x1]
%onnx::Conv_716[FLOAT, 43x43x1x1]
%onnx::Conv_719[FLOAT, 43x128x1x1]
%onnx::Conv_722[FLOAT, 128x128x1x1]
%onnx::Conv_725[FLOAT, 43x128x1x1]
%onnx::Conv_728[FLOAT, 43x43x1x1]
%onnx::Conv_731[FLOAT, 43x128x1x1]
%onnx::Conv_734[FLOAT, 128x128x1x1]
%onnx::Conv_737[FLOAT, 86x128x1x1]
%onnx::Conv_738[FLOAT, 86]
%onnx::Conv_740[FLOAT, 86x86x1x1]
%onnx::Conv_743[FLOAT, 85x128x1x1]
%onnx::Conv_744[FLOAT, 85]
%onnx::Conv_746[FLOAT, 256x128x1x1]
%onnx::Conv_747[FLOAT, 256]
%onnx::Conv_749[FLOAT, 86x256x1x1]
%onnx::Conv_752[FLOAT, 86x86x1x1]
%onnx::Conv_755[FLOAT, 85x256x1x1]
%onnx::Conv_758[FLOAT, 256x256x1x1]
%onnx::Conv_761[FLOAT, 86x256x1x1]
%onnx::Conv_764[FLOAT, 86x86x1x1]
%onnx::Conv_767[FLOAT, 85x256x1x1]
%onnx::Conv_770[FLOAT, 256x256x1x1]
%onnx::Conv_773[FLOAT, 171x256x1x1]
%onnx::Conv_774[FLOAT, 171]
%onnx::Conv_776[FLOAT, 171x171x1x1]
%onnx::Conv_779[FLOAT, 171x256x1x1]
%onnx::Conv_782[FLOAT, 512x256x1x1]
%onnx::Conv_783[FLOAT, 512]
%onnx::Conv_785[FLOAT, 171x512x1x1]
%onnx::Conv_788[FLOAT, 171x171x1x1]
%onnx::Conv_791[FLOAT, 171x512x1x1]
%onnx::Conv_794[FLOAT, 512x512x1x1]
%onnx::Conv_797[FLOAT, 171x512x1x1]
%onnx::Conv_800[FLOAT, 171x171x1x1]
%onnx::Conv_803[FLOAT, 171x512x1x1]
%onnx::Conv_806[FLOAT, 512x512x1x1]
) {
%onnx::Conv_807 = Identity(%onnx::Conv_783)
%onnx::Conv_804 = Identity(%onnx::Conv_774)
%onnx::Conv_801 = Identity(%onnx::Conv_774)
%onnx::Conv_798 = Identity(%onnx::Conv_774)
%onnx::Conv_795 = Identity(%onnx::Conv_783)
%onnx::Conv_792 = Identity(%onnx::Conv_774)
%onnx::Conv_789 = Identity(%onnx::Conv_774)
%onnx::Conv_786 = Identity(%onnx::Conv_774)
%onnx::Conv_780 = Identity(%onnx::Conv_774)
%onnx::Conv_777 = Identity(%onnx::Conv_774)
%onnx::Conv_771 = Identity(%onnx::Conv_747)
%onnx::Conv_768 = Identity(%onnx::Conv_744)
%onnx::Conv_765 = Identity(%onnx::Conv_738)
%onnx::Conv_762 = Identity(%onnx::Conv_738)
%onnx::Conv_759 = Identity(%onnx::Conv_747)
%onnx::Conv_756 = Identity(%onnx::Conv_744)
%onnx::Conv_753 = Identity(%onnx::Conv_738)
%onnx::Conv_750 = Identity(%onnx::Conv_738)
%onnx::Conv_741 = Identity(%onnx::Conv_738)
%onnx::Conv_735 = Identity(%onnx::Conv_699)
%onnx::Conv_732 = Identity(%onnx::Conv_702)
%onnx::Conv_729 = Identity(%onnx::Conv_702)
%onnx::Conv_726 = Identity(%onnx::Conv_702)
%onnx::Conv_723 = Identity(%onnx::Conv_699)
%onnx::Conv_720 = Identity(%onnx::Conv_702)
%onnx::Conv_717 = Identity(%onnx::Conv_702)
%onnx::Conv_714 = Identity(%onnx::Conv_702)
%onnx::Conv_711 = Identity(%onnx::Conv_699)
%onnx::Conv_708 = Identity(%onnx::Conv_702)
%onnx::Conv_705 = Identity(%onnx::Conv_702)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_698, %onnx::Conv_699)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_701, %onnx::Conv_702)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_704, %onnx::Conv_705)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_707, %onnx::Conv_708)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/Add_2_output_0, %/layers.1/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.1/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.1/Slice_output_0 = Slice(%/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %/layers.1/Constant_4_output_0, %/layers.1/Constant_5_output_0, %/layers.1/Constant_3_output_0, %/layers.1/Constant_6_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Slice_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.3/maxpool/MaxPool_output_0, %/layers.1/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_710, %onnx::Conv_711)
%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/Concat_output_0, %/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_713, %onnx::Conv_714)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_716, %onnx::Conv_717)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_719, %onnx::Conv_720)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/Add_2_output_0, %/layers.2/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.2/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.2/Slice_output_0 = Slice(%/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %/layers.2/Constant_4_output_0, %/layers.2/Constant_5_output_0, %/layers.2/Constant_3_output_0, %/layers.2/Constant_6_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Slice_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.3/maxpool/MaxPool_output_0, %/layers.2/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_722, %onnx::Conv_723)
%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/Concat_output_0, %/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_725, %onnx::Conv_726)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_728, %onnx::Conv_729)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_731, %onnx::Conv_732)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/Add_2_output_0, %/layers.3/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.3/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.3/Slice_output_0 = Slice(%/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %/layers.3/Constant_4_output_0, %/layers.3/Constant_5_output_0, %/layers.3/Constant_3_output_0, %/layers.3/Constant_6_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Slice_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.3/maxpool/MaxPool_output_0, %/layers.3/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_734, %onnx::Conv_735)
%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/Concat_output_0, %/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Add_4_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_737, %onnx::Conv_738)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_740, %onnx::Conv_741)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_743, %onnx::Conv_744)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.5/Slice_output_0 = Slice(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0, %/layers.5/Constant_4_output_0, %/layers.5/Constant_2_output_0, %/layers.5/Constant_5_output_0)
%/layers.5/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/Slice_output_0, %/layers.5/Constant_6_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/Add_2_output_0, %/layers.5/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.5/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.3/maxpool/MaxPool_output_0, %/layers.5/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_746, %onnx::Conv_747)
%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/Concat_output_0, %/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_749, %onnx::Conv_750)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_752, %onnx::Conv_753)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_755, %onnx::Conv_756)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.6/Slice_output_0 = Slice(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0, %/layers.6/Constant_4_output_0, %/layers.6/Constant_2_output_0, %/layers.6/Constant_5_output_0)
%/layers.6/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/Slice_output_0, %/layers.6/Constant_6_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/Add_2_output_0, %/layers.6/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.6/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.3/maxpool/MaxPool_output_0, %/layers.6/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_758, %onnx::Conv_759)
%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/Concat_output_0, %/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_761, %onnx::Conv_762)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_764, %onnx::Conv_765)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_767, %onnx::Conv_768)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.7/Slice_output_0 = Slice(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0, %/layers.7/Constant_4_output_0, %/layers.7/Constant_2_output_0, %/layers.7/Constant_5_output_0)
%/layers.7/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/Slice_output_0, %/layers.7/Constant_6_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/Add_2_output_0, %/layers.7/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.7/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.3/maxpool/MaxPool_output_0, %/layers.7/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_770, %onnx::Conv_771)
%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/Concat_output_0, %/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Add_4_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_773, %onnx::Conv_774)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_776, %onnx::Conv_777)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_779, %onnx::Conv_780)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/Add_2_output_0, %/layers.9/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.9/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.9/Slice_output_0 = Slice(%/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %/layers.9/Constant_4_output_0, %/layers.9/Constant_5_output_0, %/layers.9/Constant_3_output_0, %/layers.9/Constant_6_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Slice_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.3/maxpool/MaxPool_output_0, %/layers.9/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_782, %onnx::Conv_783)
%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/Concat_output_0, %/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_785, %onnx::Conv_786)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_788, %onnx::Conv_789)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/Add_2_output_0, %/layers.10/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.10/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.10/Slice_output_0 = Slice(%/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %/layers.10/Constant_4_output_0, %/layers.10/Constant_5_output_0, %/layers.10/Constant_3_output_0, %/layers.10/Constant_6_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Slice_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.3/maxpool/MaxPool_output_0, %/layers.10/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/Concat_output_0, %/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_797, %onnx::Conv_798)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_800, %onnx::Conv_801)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_803, %onnx::Conv_804)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/Add_2_output_0, %/layers.11/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.11/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.11/Slice_output_0 = Slice(%/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %/layers.11/Constant_4_output_0, %/layers.11/Constant_5_output_0, %/layers.11/Constant_3_output_0, %/layers.11/Constant_6_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Slice_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.3/maxpool/MaxPool_output_0, %/layers.11/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_806, %onnx::Conv_807)
%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/Concat_output_0, %/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Add_4_output_0)
%696 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %696
}
|
val_accuracy
| 87.660259
| 501,263,232
| 1,583,646
|
{'zcp_epe_nas': 133.5767926646605, 'zcp_fisher': 2.343942403793335, 'zcp_flops': 8020211712.0, 'zcp_grad_norm': 33.76853942871094, 'zcp_grasp': 6.9599609375, 'zcp_jacov': -16.05688641678529, 'zcp_l2_norm': 713.5333251953125, 'zcp_nwot': 218.29212035233485, 'zcp_params': 1583646.0, 'zcp_plain': -0.027797207236289003, 'zcp_snip': 179.36627197265625, 'zcp_synflow': 55.09874808119618, 'zcp_zen': 64.16126251220703, 'zcp_val_accuracy': 0.9131610393524171}
| |
NASBench101_168231
|
NASBench101
|
168231
|
65d1db4ceae71373fb5c19cdbba26073
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_779[FLOAT, 128x3x3x3]
%onnx::Conv_780[FLOAT, 128]
%onnx::Conv_782[FLOAT, 128x128x1x1]
%onnx::Conv_785[FLOAT, 128x128x3x3]
%onnx::Conv_788[FLOAT, 128x128x1x1]
%onnx::Conv_791[FLOAT, 128x128x1x1]
%onnx::Conv_794[FLOAT, 128x128x1x1]
%onnx::Conv_797[FLOAT, 128x128x1x1]
%onnx::Conv_800[FLOAT, 128x128x3x3]
%onnx::Conv_803[FLOAT, 128x128x1x1]
%onnx::Conv_806[FLOAT, 128x128x1x1]
%onnx::Conv_809[FLOAT, 128x128x1x1]
%onnx::Conv_812[FLOAT, 128x128x1x1]
%onnx::Conv_815[FLOAT, 128x128x3x3]
%onnx::Conv_818[FLOAT, 128x128x1x1]
%onnx::Conv_821[FLOAT, 128x128x1x1]
%onnx::Conv_824[FLOAT, 128x128x1x1]
%onnx::Conv_827[FLOAT, 256x128x1x1]
%onnx::Conv_828[FLOAT, 256]
%onnx::Conv_830[FLOAT, 256x256x3x3]
%onnx::Conv_833[FLOAT, 256x256x1x1]
%onnx::Conv_836[FLOAT, 256x128x1x1]
%onnx::Conv_839[FLOAT, 256x256x1x1]
%onnx::Conv_842[FLOAT, 256x256x1x1]
%onnx::Conv_845[FLOAT, 256x256x3x3]
%onnx::Conv_848[FLOAT, 256x256x1x1]
%onnx::Conv_851[FLOAT, 256x256x1x1]
%onnx::Conv_854[FLOAT, 256x256x1x1]
%onnx::Conv_857[FLOAT, 256x256x1x1]
%onnx::Conv_860[FLOAT, 256x256x3x3]
%onnx::Conv_863[FLOAT, 256x256x1x1]
%onnx::Conv_866[FLOAT, 256x256x1x1]
%onnx::Conv_869[FLOAT, 256x256x1x1]
%onnx::Conv_872[FLOAT, 512x256x1x1]
%onnx::Conv_873[FLOAT, 512]
%onnx::Conv_875[FLOAT, 512x512x3x3]
%onnx::Conv_878[FLOAT, 512x512x1x1]
%onnx::Conv_881[FLOAT, 512x256x1x1]
%onnx::Conv_884[FLOAT, 512x512x1x1]
%onnx::Conv_887[FLOAT, 512x512x1x1]
%onnx::Conv_890[FLOAT, 512x512x3x3]
%onnx::Conv_893[FLOAT, 512x512x1x1]
%onnx::Conv_896[FLOAT, 512x512x1x1]
%onnx::Conv_899[FLOAT, 512x512x1x1]
%onnx::Conv_902[FLOAT, 512x512x1x1]
%onnx::Conv_905[FLOAT, 512x512x3x3]
%onnx::Conv_908[FLOAT, 512x512x1x1]
%onnx::Conv_911[FLOAT, 512x512x1x1]
%onnx::Conv_914[FLOAT, 512x512x1x1]
) {
%onnx::Conv_915 = Identity(%onnx::Conv_873)
%onnx::Conv_912 = Identity(%onnx::Conv_873)
%onnx::Conv_909 = Identity(%onnx::Conv_873)
%onnx::Conv_906 = Identity(%onnx::Conv_873)
%onnx::Conv_903 = Identity(%onnx::Conv_873)
%onnx::Conv_900 = Identity(%onnx::Conv_873)
%onnx::Conv_897 = Identity(%onnx::Conv_873)
%onnx::Conv_894 = Identity(%onnx::Conv_873)
%onnx::Conv_891 = Identity(%onnx::Conv_873)
%onnx::Conv_888 = Identity(%onnx::Conv_873)
%onnx::Conv_885 = Identity(%onnx::Conv_873)
%onnx::Conv_882 = Identity(%onnx::Conv_873)
%onnx::Conv_879 = Identity(%onnx::Conv_873)
%onnx::Conv_876 = Identity(%onnx::Conv_873)
%onnx::Conv_870 = Identity(%onnx::Conv_828)
%onnx::Conv_867 = Identity(%onnx::Conv_828)
%onnx::Conv_864 = Identity(%onnx::Conv_828)
%onnx::Conv_861 = Identity(%onnx::Conv_828)
%onnx::Conv_858 = Identity(%onnx::Conv_828)
%onnx::Conv_855 = Identity(%onnx::Conv_828)
%onnx::Conv_852 = Identity(%onnx::Conv_828)
%onnx::Conv_849 = Identity(%onnx::Conv_828)
%onnx::Conv_846 = Identity(%onnx::Conv_828)
%onnx::Conv_843 = Identity(%onnx::Conv_828)
%onnx::Conv_840 = Identity(%onnx::Conv_828)
%onnx::Conv_837 = Identity(%onnx::Conv_828)
%onnx::Conv_834 = Identity(%onnx::Conv_828)
%onnx::Conv_831 = Identity(%onnx::Conv_828)
%onnx::Conv_825 = Identity(%onnx::Conv_780)
%onnx::Conv_822 = Identity(%onnx::Conv_780)
%onnx::Conv_819 = Identity(%onnx::Conv_780)
%onnx::Conv_816 = Identity(%onnx::Conv_780)
%onnx::Conv_813 = Identity(%onnx::Conv_780)
%onnx::Conv_810 = Identity(%onnx::Conv_780)
%onnx::Conv_807 = Identity(%onnx::Conv_780)
%onnx::Conv_804 = Identity(%onnx::Conv_780)
%onnx::Conv_801 = Identity(%onnx::Conv_780)
%onnx::Conv_798 = Identity(%onnx::Conv_780)
%onnx::Conv_795 = Identity(%onnx::Conv_780)
%onnx::Conv_792 = Identity(%onnx::Conv_780)
%onnx::Conv_789 = Identity(%onnx::Conv_780)
%onnx::Conv_786 = Identity(%onnx::Conv_780)
%onnx::Conv_783 = Identity(%onnx::Conv_780)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_779, %onnx::Conv_780)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_782, %onnx::Conv_783)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_785, %onnx::Conv_786)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_788, %onnx::Conv_789)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/Add_2_output_0, %/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/Add_5_output_0 = Add(%/layers.1/Add_4_output_0, %/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/Add_6_output_0 = Add(%/layers.1/Add_5_output_0, %/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_6_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_797, %onnx::Conv_798)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_800, %onnx::Conv_801)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_803, %onnx::Conv_804)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_806, %onnx::Conv_807)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/Add_2_output_0, %/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_809, %onnx::Conv_810)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/Add_5_output_0 = Add(%/layers.2/Add_4_output_0, %/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/Add_6_output_0 = Add(%/layers.2/Add_5_output_0, %/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_6_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_812, %onnx::Conv_813)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_815, %onnx::Conv_816)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_818, %onnx::Conv_819)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_821, %onnx::Conv_822)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/Add_2_output_0, %/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_824, %onnx::Conv_825)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/Add_5_output_0 = Add(%/layers.3/Add_4_output_0, %/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/Add_6_output_0 = Add(%/layers.3/Add_5_output_0, %/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_6_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_827, %onnx::Conv_828)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_830, %onnx::Conv_831)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_833, %onnx::Conv_834)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_836, %onnx::Conv_837)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/Add_2_output_0, %/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_839, %onnx::Conv_840)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/Add_5_output_0 = Add(%/layers.5/Add_4_output_0, %/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/Add_6_output_0 = Add(%/layers.5/Add_5_output_0, %/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_6_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_842, %onnx::Conv_843)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_845, %onnx::Conv_846)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_848, %onnx::Conv_849)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_851, %onnx::Conv_852)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/Add_2_output_0, %/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_854, %onnx::Conv_855)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/Add_5_output_0 = Add(%/layers.6/Add_4_output_0, %/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/Add_6_output_0 = Add(%/layers.6/Add_5_output_0, %/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_6_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_857, %onnx::Conv_858)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/Add_2_output_0, %/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/Add_5_output_0 = Add(%/layers.7/Add_4_output_0, %/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/Add_6_output_0 = Add(%/layers.7/Add_5_output_0, %/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_6_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/Add_2_output_0, %/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/Add_5_output_0 = Add(%/layers.9/Add_4_output_0, %/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/Add_6_output_0 = Add(%/layers.9/Add_5_output_0, %/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_6_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/Add_2_output_0, %/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/Add_5_output_0 = Add(%/layers.10/Add_4_output_0, %/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/Add_6_output_0 = Add(%/layers.10/Add_5_output_0, %/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_6_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_908, %onnx::Conv_909)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_911, %onnx::Conv_912)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/Add_2_output_0, %/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_914, %onnx::Conv_915)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/Add_5_output_0 = Add(%/layers.11/Add_4_output_0, %/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/Add_6_output_0 = Add(%/layers.11/Add_5_output_0, %/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_6_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/vertex_op.5/maxpool/MaxPool_output_0)
%777 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %777
}
|
val_accuracy
| 88.411456
| 3,894,421,504
| 13,126,538
|
{'zcp_epe_nas': 86.57781359268662, 'zcp_fisher': 765.5813598632812, 'zcp_flops': 62310744064.0, 'zcp_grad_norm': 432.3982238769531, 'zcp_grasp': -396.07421875, 'zcp_jacov': -16.064655359382648, 'zcp_l2_norm': 1030.7265625, 'zcp_nwot': 231.95618909962798, 'zcp_params': 13126538.0, 'zcp_plain': 0.19517119228839802, 'zcp_snip': 3423.604736328125, 'zcp_synflow': 120.47097372178004, 'zcp_zen': 95.87364196777344, 'zcp_val_accuracy': 0.922475934028625}
| |
NASBench101_112222
|
NASBench101
|
112222
|
43c249a2643ceb1287d0fe9665340fc1
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_743[FLOAT, 128x3x3x3]
%onnx::Conv_744[FLOAT, 128]
%onnx::Conv_746[FLOAT, 128x128x1x1]
%onnx::Conv_749[FLOAT, 128x128x3x3]
%onnx::Conv_752[FLOAT, 128x128x1x1]
%onnx::Conv_755[FLOAT, 128x128x3x3]
%onnx::Conv_758[FLOAT, 128x128x1x1]
%onnx::Conv_761[FLOAT, 128x128x1x1]
%onnx::Conv_764[FLOAT, 128x128x3x3]
%onnx::Conv_767[FLOAT, 128x128x1x1]
%onnx::Conv_770[FLOAT, 128x128x3x3]
%onnx::Conv_773[FLOAT, 128x128x1x1]
%onnx::Conv_776[FLOAT, 128x128x1x1]
%onnx::Conv_779[FLOAT, 128x128x3x3]
%onnx::Conv_782[FLOAT, 128x128x1x1]
%onnx::Conv_785[FLOAT, 128x128x3x3]
%onnx::Conv_788[FLOAT, 128x128x1x1]
%onnx::Conv_791[FLOAT, 256x128x1x1]
%onnx::Conv_792[FLOAT, 256]
%onnx::Conv_794[FLOAT, 256x256x3x3]
%onnx::Conv_797[FLOAT, 256x128x1x1]
%onnx::Conv_800[FLOAT, 256x256x3x3]
%onnx::Conv_803[FLOAT, 256x128x1x1]
%onnx::Conv_806[FLOAT, 256x256x1x1]
%onnx::Conv_809[FLOAT, 256x256x3x3]
%onnx::Conv_812[FLOAT, 256x256x1x1]
%onnx::Conv_815[FLOAT, 256x256x3x3]
%onnx::Conv_818[FLOAT, 256x256x1x1]
%onnx::Conv_821[FLOAT, 256x256x1x1]
%onnx::Conv_824[FLOAT, 256x256x3x3]
%onnx::Conv_827[FLOAT, 256x256x1x1]
%onnx::Conv_830[FLOAT, 256x256x3x3]
%onnx::Conv_833[FLOAT, 256x256x1x1]
%onnx::Conv_836[FLOAT, 512x256x1x1]
%onnx::Conv_837[FLOAT, 512]
%onnx::Conv_839[FLOAT, 512x512x3x3]
%onnx::Conv_842[FLOAT, 512x256x1x1]
%onnx::Conv_845[FLOAT, 512x512x3x3]
%onnx::Conv_848[FLOAT, 512x256x1x1]
%onnx::Conv_851[FLOAT, 512x512x1x1]
%onnx::Conv_854[FLOAT, 512x512x3x3]
%onnx::Conv_857[FLOAT, 512x512x1x1]
%onnx::Conv_860[FLOAT, 512x512x3x3]
%onnx::Conv_863[FLOAT, 512x512x1x1]
%onnx::Conv_866[FLOAT, 512x512x1x1]
%onnx::Conv_869[FLOAT, 512x512x3x3]
%onnx::Conv_872[FLOAT, 512x512x1x1]
%onnx::Conv_875[FLOAT, 512x512x3x3]
%onnx::Conv_878[FLOAT, 512x512x1x1]
) {
%onnx::Conv_879 = Identity(%onnx::Conv_837)
%onnx::Conv_876 = Identity(%onnx::Conv_837)
%onnx::Conv_873 = Identity(%onnx::Conv_837)
%onnx::Conv_870 = Identity(%onnx::Conv_837)
%onnx::Conv_867 = Identity(%onnx::Conv_837)
%onnx::Conv_864 = Identity(%onnx::Conv_837)
%onnx::Conv_861 = Identity(%onnx::Conv_837)
%onnx::Conv_858 = Identity(%onnx::Conv_837)
%onnx::Conv_855 = Identity(%onnx::Conv_837)
%onnx::Conv_852 = Identity(%onnx::Conv_837)
%onnx::Conv_849 = Identity(%onnx::Conv_837)
%onnx::Conv_846 = Identity(%onnx::Conv_837)
%onnx::Conv_843 = Identity(%onnx::Conv_837)
%onnx::Conv_840 = Identity(%onnx::Conv_837)
%onnx::Conv_834 = Identity(%onnx::Conv_792)
%onnx::Conv_831 = Identity(%onnx::Conv_792)
%onnx::Conv_828 = Identity(%onnx::Conv_792)
%onnx::Conv_825 = Identity(%onnx::Conv_792)
%onnx::Conv_822 = Identity(%onnx::Conv_792)
%onnx::Conv_819 = Identity(%onnx::Conv_792)
%onnx::Conv_816 = Identity(%onnx::Conv_792)
%onnx::Conv_813 = Identity(%onnx::Conv_792)
%onnx::Conv_810 = Identity(%onnx::Conv_792)
%onnx::Conv_807 = Identity(%onnx::Conv_792)
%onnx::Conv_804 = Identity(%onnx::Conv_792)
%onnx::Conv_801 = Identity(%onnx::Conv_792)
%onnx::Conv_798 = Identity(%onnx::Conv_792)
%onnx::Conv_795 = Identity(%onnx::Conv_792)
%onnx::Conv_789 = Identity(%onnx::Conv_744)
%onnx::Conv_786 = Identity(%onnx::Conv_744)
%onnx::Conv_783 = Identity(%onnx::Conv_744)
%onnx::Conv_780 = Identity(%onnx::Conv_744)
%onnx::Conv_777 = Identity(%onnx::Conv_744)
%onnx::Conv_774 = Identity(%onnx::Conv_744)
%onnx::Conv_771 = Identity(%onnx::Conv_744)
%onnx::Conv_768 = Identity(%onnx::Conv_744)
%onnx::Conv_765 = Identity(%onnx::Conv_744)
%onnx::Conv_762 = Identity(%onnx::Conv_744)
%onnx::Conv_759 = Identity(%onnx::Conv_744)
%onnx::Conv_756 = Identity(%onnx::Conv_744)
%onnx::Conv_753 = Identity(%onnx::Conv_744)
%onnx::Conv_750 = Identity(%onnx::Conv_744)
%onnx::Conv_747 = Identity(%onnx::Conv_744)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_743, %onnx::Conv_744)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_746, %onnx::Conv_747)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_749, %onnx::Conv_750)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_752, %onnx::Conv_753)
%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.3/maxpool/MaxPool_output_0, %/layers.1/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/Add_2_output_0, %/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_755, %onnx::Conv_756)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_758, %onnx::Conv_759)
%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_761, %onnx::Conv_762)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_764, %onnx::Conv_765)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_767, %onnx::Conv_768)
%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.3/maxpool/MaxPool_output_0, %/layers.2/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/Add_2_output_0, %/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_770, %onnx::Conv_771)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_773, %onnx::Conv_774)
%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_776, %onnx::Conv_777)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_779, %onnx::Conv_780)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_782, %onnx::Conv_783)
%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.3/maxpool/MaxPool_output_0, %/layers.3/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/Add_2_output_0, %/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_785, %onnx::Conv_786)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_788, %onnx::Conv_789)
%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Add_4_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_797, %onnx::Conv_798)
%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.3/maxpool/MaxPool_output_0, %/layers.5/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/Add_2_output_0, %/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_800, %onnx::Conv_801)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_803, %onnx::Conv_804)
%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_806, %onnx::Conv_807)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_809, %onnx::Conv_810)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_812, %onnx::Conv_813)
%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.3/maxpool/MaxPool_output_0, %/layers.6/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/Add_2_output_0, %/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_815, %onnx::Conv_816)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_818, %onnx::Conv_819)
%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_821, %onnx::Conv_822)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_824, %onnx::Conv_825)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_827, %onnx::Conv_828)
%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.3/maxpool/MaxPool_output_0, %/layers.7/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/Add_2_output_0, %/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_830, %onnx::Conv_831)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_833, %onnx::Conv_834)
%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Add_4_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_836, %onnx::Conv_837)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_839, %onnx::Conv_840)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_842, %onnx::Conv_843)
%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.3/maxpool/MaxPool_output_0, %/layers.9/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/Add_2_output_0, %/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_845, %onnx::Conv_846)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_848, %onnx::Conv_849)
%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_851, %onnx::Conv_852)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_854, %onnx::Conv_855)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_857, %onnx::Conv_858)
%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.3/maxpool/MaxPool_output_0, %/layers.10/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/Add_2_output_0, %/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.3/maxpool/MaxPool_output_0, %/layers.11/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/Add_2_output_0, %/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Add_4_output_0)
%741 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %741
}
|
val_accuracy
| 93.609774
| 6,276,786,176
| 21,220,234
|
{'zcp_epe_nas': 92.5803014571585, 'zcp_fisher': 27.52068328857422, 'zcp_flops': 100428578816.0, 'zcp_grad_norm': 86.03299713134766, 'zcp_grasp': -4.34564208984375, 'zcp_jacov': -16.064400499948814, 'zcp_l2_norm': 1014.4686889648438, 'zcp_nwot': 231.56322738606062, 'zcp_params': 21220234.0, 'zcp_plain': 0.004420166835188, 'zcp_snip': 759.8167114257812, 'zcp_synflow': 109.60375274439713, 'zcp_zen': 106.68012237548828, 'zcp_val_accuracy': 0.9364984035491941}
| |
NASBench101_416515
|
NASBench101
|
416515
|
fbb47ed0179acaff51235455fc29316e
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_824[FLOAT, 128x3x3x3]
%onnx::Conv_825[FLOAT, 128]
%onnx::Conv_827[FLOAT, 128x128x1x1]
%onnx::Conv_830[FLOAT, 128x128x3x3]
%onnx::Conv_833[FLOAT, 128x128x1x1]
%onnx::Conv_836[FLOAT, 128x128x1x1]
%onnx::Conv_839[FLOAT, 128x128x1x1]
%onnx::Conv_842[FLOAT, 128x128x1x1]
%onnx::Conv_845[FLOAT, 128x128x1x1]
%onnx::Conv_848[FLOAT, 128x128x3x3]
%onnx::Conv_851[FLOAT, 128x128x1x1]
%onnx::Conv_854[FLOAT, 128x128x1x1]
%onnx::Conv_857[FLOAT, 128x128x1x1]
%onnx::Conv_860[FLOAT, 128x128x1x1]
%onnx::Conv_863[FLOAT, 128x128x1x1]
%onnx::Conv_866[FLOAT, 128x128x3x3]
%onnx::Conv_869[FLOAT, 128x128x1x1]
%onnx::Conv_872[FLOAT, 128x128x1x1]
%onnx::Conv_875[FLOAT, 128x128x1x1]
%onnx::Conv_878[FLOAT, 128x128x1x1]
%onnx::Conv_881[FLOAT, 256x128x1x1]
%onnx::Conv_882[FLOAT, 256]
%onnx::Conv_884[FLOAT, 256x256x3x3]
%onnx::Conv_887[FLOAT, 256x128x1x1]
%onnx::Conv_890[FLOAT, 256x256x1x1]
%onnx::Conv_893[FLOAT, 256x128x1x1]
%onnx::Conv_896[FLOAT, 256x256x1x1]
%onnx::Conv_899[FLOAT, 256x256x1x1]
%onnx::Conv_902[FLOAT, 256x256x3x3]
%onnx::Conv_905[FLOAT, 256x256x1x1]
%onnx::Conv_908[FLOAT, 256x256x1x1]
%onnx::Conv_911[FLOAT, 256x256x1x1]
%onnx::Conv_914[FLOAT, 256x256x1x1]
%onnx::Conv_917[FLOAT, 256x256x1x1]
%onnx::Conv_920[FLOAT, 256x256x3x3]
%onnx::Conv_923[FLOAT, 256x256x1x1]
%onnx::Conv_926[FLOAT, 256x256x1x1]
%onnx::Conv_929[FLOAT, 256x256x1x1]
%onnx::Conv_932[FLOAT, 256x256x1x1]
%onnx::Conv_935[FLOAT, 512x256x1x1]
%onnx::Conv_936[FLOAT, 512]
%onnx::Conv_938[FLOAT, 512x512x3x3]
%onnx::Conv_941[FLOAT, 512x256x1x1]
%onnx::Conv_944[FLOAT, 512x512x1x1]
%onnx::Conv_947[FLOAT, 512x256x1x1]
%onnx::Conv_950[FLOAT, 512x512x1x1]
%onnx::Conv_953[FLOAT, 512x512x1x1]
%onnx::Conv_956[FLOAT, 512x512x3x3]
%onnx::Conv_959[FLOAT, 512x512x1x1]
%onnx::Conv_962[FLOAT, 512x512x1x1]
%onnx::Conv_965[FLOAT, 512x512x1x1]
%onnx::Conv_968[FLOAT, 512x512x1x1]
%onnx::Conv_971[FLOAT, 512x512x1x1]
%onnx::Conv_974[FLOAT, 512x512x3x3]
%onnx::Conv_977[FLOAT, 512x512x1x1]
%onnx::Conv_980[FLOAT, 512x512x1x1]
%onnx::Conv_983[FLOAT, 512x512x1x1]
%onnx::Conv_986[FLOAT, 512x512x1x1]
) {
%onnx::Conv_987 = Identity(%onnx::Conv_936)
%onnx::Conv_984 = Identity(%onnx::Conv_936)
%onnx::Conv_981 = Identity(%onnx::Conv_936)
%onnx::Conv_978 = Identity(%onnx::Conv_936)
%onnx::Conv_975 = Identity(%onnx::Conv_936)
%onnx::Conv_972 = Identity(%onnx::Conv_936)
%onnx::Conv_969 = Identity(%onnx::Conv_936)
%onnx::Conv_966 = Identity(%onnx::Conv_936)
%onnx::Conv_963 = Identity(%onnx::Conv_936)
%onnx::Conv_960 = Identity(%onnx::Conv_936)
%onnx::Conv_957 = Identity(%onnx::Conv_936)
%onnx::Conv_954 = Identity(%onnx::Conv_936)
%onnx::Conv_951 = Identity(%onnx::Conv_936)
%onnx::Conv_948 = Identity(%onnx::Conv_936)
%onnx::Conv_945 = Identity(%onnx::Conv_936)
%onnx::Conv_942 = Identity(%onnx::Conv_936)
%onnx::Conv_939 = Identity(%onnx::Conv_936)
%onnx::Conv_933 = Identity(%onnx::Conv_882)
%onnx::Conv_930 = Identity(%onnx::Conv_882)
%onnx::Conv_927 = Identity(%onnx::Conv_882)
%onnx::Conv_924 = Identity(%onnx::Conv_882)
%onnx::Conv_921 = Identity(%onnx::Conv_882)
%onnx::Conv_918 = Identity(%onnx::Conv_882)
%onnx::Conv_915 = Identity(%onnx::Conv_882)
%onnx::Conv_912 = Identity(%onnx::Conv_882)
%onnx::Conv_909 = Identity(%onnx::Conv_882)
%onnx::Conv_906 = Identity(%onnx::Conv_882)
%onnx::Conv_903 = Identity(%onnx::Conv_882)
%onnx::Conv_900 = Identity(%onnx::Conv_882)
%onnx::Conv_897 = Identity(%onnx::Conv_882)
%onnx::Conv_894 = Identity(%onnx::Conv_882)
%onnx::Conv_891 = Identity(%onnx::Conv_882)
%onnx::Conv_888 = Identity(%onnx::Conv_882)
%onnx::Conv_885 = Identity(%onnx::Conv_882)
%onnx::Conv_879 = Identity(%onnx::Conv_825)
%onnx::Conv_876 = Identity(%onnx::Conv_825)
%onnx::Conv_873 = Identity(%onnx::Conv_825)
%onnx::Conv_870 = Identity(%onnx::Conv_825)
%onnx::Conv_867 = Identity(%onnx::Conv_825)
%onnx::Conv_864 = Identity(%onnx::Conv_825)
%onnx::Conv_861 = Identity(%onnx::Conv_825)
%onnx::Conv_858 = Identity(%onnx::Conv_825)
%onnx::Conv_855 = Identity(%onnx::Conv_825)
%onnx::Conv_852 = Identity(%onnx::Conv_825)
%onnx::Conv_849 = Identity(%onnx::Conv_825)
%onnx::Conv_846 = Identity(%onnx::Conv_825)
%onnx::Conv_843 = Identity(%onnx::Conv_825)
%onnx::Conv_840 = Identity(%onnx::Conv_825)
%onnx::Conv_837 = Identity(%onnx::Conv_825)
%onnx::Conv_834 = Identity(%onnx::Conv_825)
%onnx::Conv_831 = Identity(%onnx::Conv_825)
%onnx::Conv_828 = Identity(%onnx::Conv_825)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_824, %onnx::Conv_825)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_827, %onnx::Conv_828)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_830, %onnx::Conv_831)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_833, %onnx::Conv_834)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/Add_1_output_0, %/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_836, %onnx::Conv_837)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_839, %onnx::Conv_840)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/Add_3_output_0, %/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_842, %onnx::Conv_843)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_845, %onnx::Conv_846)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_848, %onnx::Conv_849)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_851, %onnx::Conv_852)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/Add_1_output_0, %/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_854, %onnx::Conv_855)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_857, %onnx::Conv_858)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/Add_3_output_0, %/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/Add_1_output_0, %/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/Add_3_output_0, %/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_4_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/Add_1_output_0, %/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/Add_3_output_0, %/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/Add_1_output_0, %/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_908, %onnx::Conv_909)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_911, %onnx::Conv_912)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/Add_3_output_0, %/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_914, %onnx::Conv_915)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_917, %onnx::Conv_918)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_920, %onnx::Conv_921)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_923, %onnx::Conv_924)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/Add_1_output_0, %/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_926, %onnx::Conv_927)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_929, %onnx::Conv_930)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/Add_3_output_0, %/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_4_output_0, %onnx::Conv_932, %onnx::Conv_933)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_935, %onnx::Conv_936)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_938, %onnx::Conv_939)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_941, %onnx::Conv_942)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/Add_1_output_0, %/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_944, %onnx::Conv_945)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_947, %onnx::Conv_948)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/Add_3_output_0, %/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_950, %onnx::Conv_951)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_953, %onnx::Conv_954)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_956, %onnx::Conv_957)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_959, %onnx::Conv_960)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/Add_1_output_0, %/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_962, %onnx::Conv_963)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_965, %onnx::Conv_966)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/Add_3_output_0, %/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_968, %onnx::Conv_969)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_974, %onnx::Conv_975)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_977, %onnx::Conv_978)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/Add_1_output_0, %/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/Add_3_output_0, %/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_4_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%822 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %822
}
|
val_accuracy
| 93.419468
| 4,168,361,984
| 14,000,266
|
{'zcp_epe_nas': 88.0178286306825, 'zcp_fisher': 63.25199890136719, 'zcp_flops': 66693791744.0, 'zcp_grad_norm': 148.4976348876953, 'zcp_grasp': 17.016357421875, 'zcp_jacov': -16.059048206375103, 'zcp_l2_norm': 1226.176513671875, 'zcp_nwot': 235.4915031433725, 'zcp_params': 14000266.0, 'zcp_plain': 0.002089280635118, 'zcp_snip': 1149.6990966796875, 'zcp_synflow': 120.06707320802892, 'zcp_zen': 100.7274169921875, 'zcp_val_accuracy': 0.93359375}
| |
NASBench101_159249
|
NASBench101
|
159249
|
606748918116257a04cd9702710b1426
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_761[FLOAT, 128x3x3x3]
%onnx::Conv_762[FLOAT, 128]
%onnx::Conv_764[FLOAT, 64x128x1x1]
%onnx::Conv_765[FLOAT, 64]
%onnx::Conv_767[FLOAT, 64x64x1x1]
%onnx::Conv_770[FLOAT, 64x64x1x1]
%onnx::Conv_773[FLOAT, 64x128x1x1]
%onnx::Conv_776[FLOAT, 64x64x1x1]
%onnx::Conv_779[FLOAT, 64x128x1x1]
%onnx::Conv_782[FLOAT, 64x64x1x1]
%onnx::Conv_785[FLOAT, 64x64x1x1]
%onnx::Conv_788[FLOAT, 64x128x1x1]
%onnx::Conv_791[FLOAT, 64x64x1x1]
%onnx::Conv_794[FLOAT, 64x128x1x1]
%onnx::Conv_797[FLOAT, 64x64x1x1]
%onnx::Conv_800[FLOAT, 64x64x1x1]
%onnx::Conv_803[FLOAT, 64x128x1x1]
%onnx::Conv_806[FLOAT, 64x64x1x1]
%onnx::Conv_809[FLOAT, 128x128x1x1]
%onnx::Conv_812[FLOAT, 128x128x1x1]
%onnx::Conv_815[FLOAT, 128x128x1x1]
%onnx::Conv_818[FLOAT, 128x128x1x1]
%onnx::Conv_821[FLOAT, 128x128x1x1]
%onnx::Conv_824[FLOAT, 128x256x1x1]
%onnx::Conv_827[FLOAT, 128x128x1x1]
%onnx::Conv_830[FLOAT, 128x128x1x1]
%onnx::Conv_833[FLOAT, 128x256x1x1]
%onnx::Conv_836[FLOAT, 128x128x1x1]
%onnx::Conv_839[FLOAT, 128x256x1x1]
%onnx::Conv_842[FLOAT, 128x128x1x1]
%onnx::Conv_845[FLOAT, 128x128x1x1]
%onnx::Conv_848[FLOAT, 128x256x1x1]
%onnx::Conv_851[FLOAT, 128x128x1x1]
%onnx::Conv_854[FLOAT, 256x256x1x1]
%onnx::Conv_855[FLOAT, 256]
%onnx::Conv_857[FLOAT, 256x256x1x1]
%onnx::Conv_860[FLOAT, 256x256x1x1]
%onnx::Conv_863[FLOAT, 256x256x1x1]
%onnx::Conv_866[FLOAT, 256x256x1x1]
%onnx::Conv_869[FLOAT, 256x512x1x1]
%onnx::Conv_872[FLOAT, 256x256x1x1]
%onnx::Conv_875[FLOAT, 256x256x1x1]
%onnx::Conv_878[FLOAT, 256x512x1x1]
%onnx::Conv_881[FLOAT, 256x256x1x1]
%onnx::Conv_884[FLOAT, 256x512x1x1]
%onnx::Conv_887[FLOAT, 256x256x1x1]
%onnx::Conv_890[FLOAT, 256x256x1x1]
%onnx::Conv_893[FLOAT, 256x512x1x1]
%onnx::Conv_896[FLOAT, 256x256x1x1]
) {
%onnx::Conv_897 = Identity(%onnx::Conv_855)
%onnx::Conv_894 = Identity(%onnx::Conv_855)
%onnx::Conv_891 = Identity(%onnx::Conv_855)
%onnx::Conv_888 = Identity(%onnx::Conv_855)
%onnx::Conv_885 = Identity(%onnx::Conv_855)
%onnx::Conv_882 = Identity(%onnx::Conv_855)
%onnx::Conv_879 = Identity(%onnx::Conv_855)
%onnx::Conv_876 = Identity(%onnx::Conv_855)
%onnx::Conv_873 = Identity(%onnx::Conv_855)
%onnx::Conv_870 = Identity(%onnx::Conv_855)
%onnx::Conv_867 = Identity(%onnx::Conv_855)
%onnx::Conv_864 = Identity(%onnx::Conv_855)
%onnx::Conv_861 = Identity(%onnx::Conv_855)
%onnx::Conv_858 = Identity(%onnx::Conv_855)
%onnx::Conv_852 = Identity(%onnx::Conv_762)
%onnx::Conv_849 = Identity(%onnx::Conv_762)
%onnx::Conv_846 = Identity(%onnx::Conv_762)
%onnx::Conv_843 = Identity(%onnx::Conv_762)
%onnx::Conv_840 = Identity(%onnx::Conv_762)
%onnx::Conv_837 = Identity(%onnx::Conv_762)
%onnx::Conv_834 = Identity(%onnx::Conv_762)
%onnx::Conv_831 = Identity(%onnx::Conv_762)
%onnx::Conv_828 = Identity(%onnx::Conv_762)
%onnx::Conv_825 = Identity(%onnx::Conv_762)
%onnx::Conv_822 = Identity(%onnx::Conv_762)
%onnx::Conv_819 = Identity(%onnx::Conv_762)
%onnx::Conv_816 = Identity(%onnx::Conv_762)
%onnx::Conv_813 = Identity(%onnx::Conv_762)
%onnx::Conv_810 = Identity(%onnx::Conv_762)
%onnx::Conv_807 = Identity(%onnx::Conv_765)
%onnx::Conv_804 = Identity(%onnx::Conv_765)
%onnx::Conv_801 = Identity(%onnx::Conv_765)
%onnx::Conv_798 = Identity(%onnx::Conv_765)
%onnx::Conv_795 = Identity(%onnx::Conv_765)
%onnx::Conv_792 = Identity(%onnx::Conv_765)
%onnx::Conv_789 = Identity(%onnx::Conv_765)
%onnx::Conv_786 = Identity(%onnx::Conv_765)
%onnx::Conv_783 = Identity(%onnx::Conv_765)
%onnx::Conv_780 = Identity(%onnx::Conv_765)
%onnx::Conv_777 = Identity(%onnx::Conv_765)
%onnx::Conv_774 = Identity(%onnx::Conv_765)
%onnx::Conv_771 = Identity(%onnx::Conv_765)
%onnx::Conv_768 = Identity(%onnx::Conv_765)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_761, %onnx::Conv_762)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_764, %onnx::Conv_765)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_767, %onnx::Conv_768)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_770, %onnx::Conv_771)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/Add_2_output_0, %/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0)
%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_773, %onnx::Conv_774)
%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.4/maxpool/MaxPool_output_0, %/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_776, %onnx::Conv_777)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_779, %onnx::Conv_780)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_782, %onnx::Conv_783)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_785, %onnx::Conv_786)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/Add_2_output_0, %/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0)
%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_788, %onnx::Conv_789)
%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.4/maxpool/MaxPool_output_0, %/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_797, %onnx::Conv_798)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_800, %onnx::Conv_801)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/Add_2_output_0, %/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0)
%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_803, %onnx::Conv_804)
%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.4/maxpool/MaxPool_output_0, %/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_4_output_0, %onnx::Conv_806, %onnx::Conv_807)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_809, %onnx::Conv_810)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_812, %onnx::Conv_813)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_815, %onnx::Conv_816)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/Add_2_output_0, %/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0)
%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_818, %onnx::Conv_819)
%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.4/maxpool/MaxPool_output_0, %/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_821, %onnx::Conv_822)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_824, %onnx::Conv_825)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_827, %onnx::Conv_828)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_830, %onnx::Conv_831)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/Add_2_output_0, %/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0)
%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_833, %onnx::Conv_834)
%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.4/maxpool/MaxPool_output_0, %/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_836, %onnx::Conv_837)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_839, %onnx::Conv_840)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_842, %onnx::Conv_843)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_845, %onnx::Conv_846)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/Add_2_output_0, %/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0)
%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_848, %onnx::Conv_849)
%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.4/maxpool/MaxPool_output_0, %/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_4_output_0, %onnx::Conv_851, %onnx::Conv_852)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_854, %onnx::Conv_855)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_857, %onnx::Conv_858)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/Add_2_output_0, %/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0)
%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.4/maxpool/MaxPool_output_0, %/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/Add_2_output_0, %/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0)
%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.4/maxpool/MaxPool_output_0, %/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/Add_2_output_0, %/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0)
%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.4/maxpool/MaxPool_output_0, %/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_4_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%759 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %759
}
|
val_accuracy
| 89.583331
| 516,827,136
| 1,664,778
|
{'zcp_epe_nas': 112.43778642240851, 'zcp_fisher': 35.052059173583984, 'zcp_flops': 8269234176.0, 'zcp_grad_norm': 109.61199951171875, 'zcp_grasp': 14.481689453125, 'zcp_jacov': -16.066334693534195, 'zcp_l2_norm': 845.0563354492188, 'zcp_nwot': 222.0986723499848, 'zcp_params': 1664778.0, 'zcp_plain': -0.026139056310057, 'zcp_snip': 549.0386352539062, 'zcp_synflow': 98.3857966935803, 'zcp_zen': 70.08622741699219, 'zcp_val_accuracy': 0.909955918788909}
| |
NASBench101_419980
|
NASBench101
|
419980
|
fdc928e379ba220ada2874f510ab5f91
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_887[FLOAT, 128x3x3x3]
%onnx::Conv_888[FLOAT, 128]
%onnx::Conv_890[FLOAT, 128x128x1x1]
%onnx::Conv_893[FLOAT, 128x128x1x1]
%onnx::Conv_896[FLOAT, 128x128x3x3]
%onnx::Conv_899[FLOAT, 128x128x1x1]
%onnx::Conv_902[FLOAT, 128x128x1x1]
%onnx::Conv_905[FLOAT, 128x128x3x3]
%onnx::Conv_908[FLOAT, 128x128x1x1]
%onnx::Conv_911[FLOAT, 128x128x1x1]
%onnx::Conv_914[FLOAT, 128x128x3x3]
%onnx::Conv_917[FLOAT, 128x128x1x1]
%onnx::Conv_920[FLOAT, 128x128x1x1]
%onnx::Conv_923[FLOAT, 128x128x3x3]
%onnx::Conv_926[FLOAT, 128x128x1x1]
%onnx::Conv_929[FLOAT, 128x128x1x1]
%onnx::Conv_932[FLOAT, 128x128x3x3]
%onnx::Conv_935[FLOAT, 128x128x1x1]
%onnx::Conv_938[FLOAT, 128x128x1x1]
%onnx::Conv_941[FLOAT, 128x128x3x3]
%onnx::Conv_944[FLOAT, 256x128x1x1]
%onnx::Conv_945[FLOAT, 256]
%onnx::Conv_947[FLOAT, 256x128x1x1]
%onnx::Conv_950[FLOAT, 256x256x3x3]
%onnx::Conv_953[FLOAT, 256x128x1x1]
%onnx::Conv_956[FLOAT, 256x128x1x1]
%onnx::Conv_959[FLOAT, 256x256x3x3]
%onnx::Conv_962[FLOAT, 256x256x1x1]
%onnx::Conv_965[FLOAT, 256x256x1x1]
%onnx::Conv_968[FLOAT, 256x256x3x3]
%onnx::Conv_971[FLOAT, 256x256x1x1]
%onnx::Conv_974[FLOAT, 256x256x1x1]
%onnx::Conv_977[FLOAT, 256x256x3x3]
%onnx::Conv_980[FLOAT, 256x256x1x1]
%onnx::Conv_983[FLOAT, 256x256x1x1]
%onnx::Conv_986[FLOAT, 256x256x3x3]
%onnx::Conv_989[FLOAT, 256x256x1x1]
%onnx::Conv_992[FLOAT, 256x256x1x1]
%onnx::Conv_995[FLOAT, 256x256x3x3]
%onnx::Conv_998[FLOAT, 512x256x1x1]
%onnx::Conv_999[FLOAT, 512]
%onnx::Conv_1001[FLOAT, 512x256x1x1]
%onnx::Conv_1004[FLOAT, 512x512x3x3]
%onnx::Conv_1007[FLOAT, 512x256x1x1]
%onnx::Conv_1010[FLOAT, 512x256x1x1]
%onnx::Conv_1013[FLOAT, 512x512x3x3]
%onnx::Conv_1016[FLOAT, 512x512x1x1]
%onnx::Conv_1019[FLOAT, 512x512x1x1]
%onnx::Conv_1022[FLOAT, 512x512x3x3]
%onnx::Conv_1025[FLOAT, 512x512x1x1]
%onnx::Conv_1028[FLOAT, 512x512x1x1]
%onnx::Conv_1031[FLOAT, 512x512x3x3]
%onnx::Conv_1034[FLOAT, 512x512x1x1]
%onnx::Conv_1037[FLOAT, 512x512x1x1]
%onnx::Conv_1040[FLOAT, 512x512x3x3]
%onnx::Conv_1043[FLOAT, 512x512x1x1]
%onnx::Conv_1046[FLOAT, 512x512x1x1]
%onnx::Conv_1049[FLOAT, 512x512x3x3]
) {
%onnx::Conv_1050 = Identity(%onnx::Conv_999)
%onnx::Conv_1047 = Identity(%onnx::Conv_999)
%onnx::Conv_1044 = Identity(%onnx::Conv_999)
%onnx::Conv_1041 = Identity(%onnx::Conv_999)
%onnx::Conv_1038 = Identity(%onnx::Conv_999)
%onnx::Conv_1035 = Identity(%onnx::Conv_999)
%onnx::Conv_1032 = Identity(%onnx::Conv_999)
%onnx::Conv_1029 = Identity(%onnx::Conv_999)
%onnx::Conv_1026 = Identity(%onnx::Conv_999)
%onnx::Conv_1023 = Identity(%onnx::Conv_999)
%onnx::Conv_1020 = Identity(%onnx::Conv_999)
%onnx::Conv_1017 = Identity(%onnx::Conv_999)
%onnx::Conv_1014 = Identity(%onnx::Conv_999)
%onnx::Conv_1011 = Identity(%onnx::Conv_999)
%onnx::Conv_1008 = Identity(%onnx::Conv_999)
%onnx::Conv_1005 = Identity(%onnx::Conv_999)
%onnx::Conv_1002 = Identity(%onnx::Conv_999)
%onnx::Conv_996 = Identity(%onnx::Conv_945)
%onnx::Conv_993 = Identity(%onnx::Conv_945)
%onnx::Conv_990 = Identity(%onnx::Conv_945)
%onnx::Conv_987 = Identity(%onnx::Conv_945)
%onnx::Conv_984 = Identity(%onnx::Conv_945)
%onnx::Conv_981 = Identity(%onnx::Conv_945)
%onnx::Conv_978 = Identity(%onnx::Conv_945)
%onnx::Conv_975 = Identity(%onnx::Conv_945)
%onnx::Conv_972 = Identity(%onnx::Conv_945)
%onnx::Conv_969 = Identity(%onnx::Conv_945)
%onnx::Conv_966 = Identity(%onnx::Conv_945)
%onnx::Conv_963 = Identity(%onnx::Conv_945)
%onnx::Conv_960 = Identity(%onnx::Conv_945)
%onnx::Conv_957 = Identity(%onnx::Conv_945)
%onnx::Conv_954 = Identity(%onnx::Conv_945)
%onnx::Conv_951 = Identity(%onnx::Conv_945)
%onnx::Conv_948 = Identity(%onnx::Conv_945)
%onnx::Conv_942 = Identity(%onnx::Conv_888)
%onnx::Conv_939 = Identity(%onnx::Conv_888)
%onnx::Conv_936 = Identity(%onnx::Conv_888)
%onnx::Conv_933 = Identity(%onnx::Conv_888)
%onnx::Conv_930 = Identity(%onnx::Conv_888)
%onnx::Conv_927 = Identity(%onnx::Conv_888)
%onnx::Conv_924 = Identity(%onnx::Conv_888)
%onnx::Conv_921 = Identity(%onnx::Conv_888)
%onnx::Conv_918 = Identity(%onnx::Conv_888)
%onnx::Conv_915 = Identity(%onnx::Conv_888)
%onnx::Conv_912 = Identity(%onnx::Conv_888)
%onnx::Conv_909 = Identity(%onnx::Conv_888)
%onnx::Conv_906 = Identity(%onnx::Conv_888)
%onnx::Conv_903 = Identity(%onnx::Conv_888)
%onnx::Conv_900 = Identity(%onnx::Conv_888)
%onnx::Conv_897 = Identity(%onnx::Conv_888)
%onnx::Conv_894 = Identity(%onnx::Conv_888)
%onnx::Conv_891 = Identity(%onnx::Conv_888)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_887, %onnx::Conv_888)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/Add_3_output_0, %/layers.1/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.1/Add_5_output_0 = Add(%/layers.1/Add_4_output_0, %/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_5_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_6_output_0 = Add(%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_6_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_908, %onnx::Conv_909)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_911, %onnx::Conv_912)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_914, %onnx::Conv_915)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_917, %onnx::Conv_918)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_920, %onnx::Conv_921)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/Add_3_output_0, %/layers.2/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.2/Add_5_output_0 = Add(%/layers.2/Add_4_output_0, %/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_5_output_0, %onnx::Conv_923, %onnx::Conv_924)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_6_output_0 = Add(%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_6_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_926, %onnx::Conv_927)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_929, %onnx::Conv_930)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_932, %onnx::Conv_933)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_935, %onnx::Conv_936)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_938, %onnx::Conv_939)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/Add_3_output_0, %/layers.3/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.3/Add_5_output_0 = Add(%/layers.3/Add_4_output_0, %/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_5_output_0, %onnx::Conv_941, %onnx::Conv_942)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_6_output_0 = Add(%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_6_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_944, %onnx::Conv_945)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_947, %onnx::Conv_948)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_950, %onnx::Conv_951)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_953, %onnx::Conv_954)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_956, %onnx::Conv_957)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/Add_3_output_0, %/layers.5/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.5/Add_5_output_0 = Add(%/layers.5/Add_4_output_0, %/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_5_output_0, %onnx::Conv_959, %onnx::Conv_960)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_6_output_0 = Add(%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_6_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_962, %onnx::Conv_963)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_965, %onnx::Conv_966)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_968, %onnx::Conv_969)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_974, %onnx::Conv_975)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/Add_3_output_0, %/layers.6/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.6/Add_5_output_0 = Add(%/layers.6/Add_4_output_0, %/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_5_output_0, %onnx::Conv_977, %onnx::Conv_978)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_6_output_0 = Add(%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_6_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/Add_3_output_0, %/layers.7/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.7/Add_5_output_0 = Add(%/layers.7/Add_4_output_0, %/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_5_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_6_output_0 = Add(%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_6_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/Add_3_output_0, %/layers.9/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.9/Add_5_output_0 = Add(%/layers.9/Add_4_output_0, %/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_5_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_6_output_0 = Add(%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_6_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_1025, %onnx::Conv_1026)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_1028, %onnx::Conv_1029)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/Add_3_output_0, %/layers.10/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.10/Add_5_output_0 = Add(%/layers.10/Add_4_output_0, %/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_5_output_0, %onnx::Conv_1031, %onnx::Conv_1032)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_6_output_0 = Add(%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_6_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_1034, %onnx::Conv_1035)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_1037, %onnx::Conv_1038)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_1040, %onnx::Conv_1041)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_1043, %onnx::Conv_1044)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_1046, %onnx::Conv_1047)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/Add_3_output_0, %/layers.11/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.11/Add_5_output_0 = Add(%/layers.11/Add_4_output_0, %/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_5_output_0, %onnx::Conv_1049, %onnx::Conv_1050)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_6_output_0 = Add(%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_6_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/vertex_op.5/maxpool/MaxPool_output_0)
%885 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %885
}
|
val_accuracy
| 91.726762
| 6,550,726,656
| 22,093,962
|
{'zcp_epe_nas': 83.96426850185898, 'zcp_fisher': 20.676435470581055, 'zcp_flops': 104811626496.0, 'zcp_grad_norm': 87.9381103515625, 'zcp_grasp': -10.8192138671875, 'zcp_jacov': -16.043265469044705, 'zcp_l2_norm': 1209.930419921875, 'zcp_nwot': 234.31919810017996, 'zcp_params': 22093962.0, 'zcp_plain': 0.008341553620994, 'zcp_snip': 829.118408203125, 'zcp_synflow': 103.72930937890187, 'zcp_zen': 123.25053405761719, 'zcp_val_accuracy': 0.9288862347602841}
| |
NASBench101_117730
|
NASBench101
|
117730
|
4711da2ee57ed3db4afaefc95ade4b03
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_830[FLOAT, 128x3x3x3]
%onnx::Conv_831[FLOAT, 128]
%onnx::Conv_833[FLOAT, 43x128x1x1]
%onnx::Conv_834[FLOAT, 43]
%onnx::Conv_836[FLOAT, 43x43x3x3]
%onnx::Conv_839[FLOAT, 43x128x1x1]
%onnx::Conv_842[FLOAT, 43x43x1x1]
%onnx::Conv_845[FLOAT, 43x43x3x3]
%onnx::Conv_848[FLOAT, 43x128x1x1]
%onnx::Conv_851[FLOAT, 43x43x3x3]
%onnx::Conv_854[FLOAT, 43x128x1x1]
%onnx::Conv_857[FLOAT, 43x43x1x1]
%onnx::Conv_860[FLOAT, 43x43x3x3]
%onnx::Conv_863[FLOAT, 43x128x1x1]
%onnx::Conv_866[FLOAT, 43x43x3x3]
%onnx::Conv_869[FLOAT, 43x128x1x1]
%onnx::Conv_872[FLOAT, 43x43x1x1]
%onnx::Conv_875[FLOAT, 43x43x3x3]
%onnx::Conv_878[FLOAT, 86x128x1x1]
%onnx::Conv_879[FLOAT, 86]
%onnx::Conv_881[FLOAT, 86x86x3x3]
%onnx::Conv_884[FLOAT, 86x128x1x1]
%onnx::Conv_887[FLOAT, 86x86x1x1]
%onnx::Conv_890[FLOAT, 85x85x3x3]
%onnx::Conv_891[FLOAT, 85]
%onnx::Conv_893[FLOAT, 86x256x1x1]
%onnx::Conv_896[FLOAT, 86x86x3x3]
%onnx::Conv_899[FLOAT, 86x256x1x1]
%onnx::Conv_902[FLOAT, 86x86x1x1]
%onnx::Conv_905[FLOAT, 85x85x3x3]
%onnx::Conv_908[FLOAT, 86x256x1x1]
%onnx::Conv_911[FLOAT, 86x86x3x3]
%onnx::Conv_914[FLOAT, 86x256x1x1]
%onnx::Conv_917[FLOAT, 86x86x1x1]
%onnx::Conv_920[FLOAT, 85x85x3x3]
%onnx::Conv_923[FLOAT, 171x256x1x1]
%onnx::Conv_924[FLOAT, 171]
%onnx::Conv_926[FLOAT, 171x171x3x3]
%onnx::Conv_929[FLOAT, 171x256x1x1]
%onnx::Conv_932[FLOAT, 171x171x1x1]
%onnx::Conv_935[FLOAT, 171x171x3x3]
%onnx::Conv_938[FLOAT, 171x512x1x1]
%onnx::Conv_941[FLOAT, 171x171x3x3]
%onnx::Conv_944[FLOAT, 171x512x1x1]
%onnx::Conv_947[FLOAT, 171x171x1x1]
%onnx::Conv_950[FLOAT, 171x171x3x3]
%onnx::Conv_953[FLOAT, 171x512x1x1]
%onnx::Conv_956[FLOAT, 171x171x3x3]
%onnx::Conv_959[FLOAT, 171x512x1x1]
%onnx::Conv_962[FLOAT, 171x171x1x1]
%onnx::Conv_965[FLOAT, 171x171x3x3]
) {
%onnx::Conv_966 = Identity(%onnx::Conv_924)
%onnx::Conv_963 = Identity(%onnx::Conv_924)
%onnx::Conv_960 = Identity(%onnx::Conv_924)
%onnx::Conv_957 = Identity(%onnx::Conv_924)
%onnx::Conv_954 = Identity(%onnx::Conv_924)
%onnx::Conv_951 = Identity(%onnx::Conv_924)
%onnx::Conv_948 = Identity(%onnx::Conv_924)
%onnx::Conv_945 = Identity(%onnx::Conv_924)
%onnx::Conv_942 = Identity(%onnx::Conv_924)
%onnx::Conv_939 = Identity(%onnx::Conv_924)
%onnx::Conv_936 = Identity(%onnx::Conv_924)
%onnx::Conv_933 = Identity(%onnx::Conv_924)
%onnx::Conv_930 = Identity(%onnx::Conv_924)
%onnx::Conv_927 = Identity(%onnx::Conv_924)
%onnx::Conv_921 = Identity(%onnx::Conv_891)
%onnx::Conv_918 = Identity(%onnx::Conv_879)
%onnx::Conv_915 = Identity(%onnx::Conv_879)
%onnx::Conv_912 = Identity(%onnx::Conv_879)
%onnx::Conv_909 = Identity(%onnx::Conv_879)
%onnx::Conv_906 = Identity(%onnx::Conv_891)
%onnx::Conv_903 = Identity(%onnx::Conv_879)
%onnx::Conv_900 = Identity(%onnx::Conv_879)
%onnx::Conv_897 = Identity(%onnx::Conv_879)
%onnx::Conv_894 = Identity(%onnx::Conv_879)
%onnx::Conv_888 = Identity(%onnx::Conv_879)
%onnx::Conv_885 = Identity(%onnx::Conv_879)
%onnx::Conv_882 = Identity(%onnx::Conv_879)
%onnx::Conv_876 = Identity(%onnx::Conv_834)
%onnx::Conv_873 = Identity(%onnx::Conv_834)
%onnx::Conv_870 = Identity(%onnx::Conv_834)
%onnx::Conv_867 = Identity(%onnx::Conv_834)
%onnx::Conv_864 = Identity(%onnx::Conv_834)
%onnx::Conv_861 = Identity(%onnx::Conv_834)
%onnx::Conv_858 = Identity(%onnx::Conv_834)
%onnx::Conv_855 = Identity(%onnx::Conv_834)
%onnx::Conv_852 = Identity(%onnx::Conv_834)
%onnx::Conv_849 = Identity(%onnx::Conv_834)
%onnx::Conv_846 = Identity(%onnx::Conv_834)
%onnx::Conv_843 = Identity(%onnx::Conv_834)
%onnx::Conv_840 = Identity(%onnx::Conv_834)
%onnx::Conv_837 = Identity(%onnx::Conv_834)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_830, %onnx::Conv_831)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_833, %onnx::Conv_834)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_836, %onnx::Conv_837)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_839, %onnx::Conv_840)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/Add_1_output_0, %/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_842, %onnx::Conv_843)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_845, %onnx::Conv_846)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.1/Slice_output_0 = Slice(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_4_output_0, %/layers.1/Constant_5_output_0, %/layers.1/Constant_3_output_0, %/layers.1/Constant_6_output_0)
%/layers.1/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_4_output_0 = Add(%/layers.1/Slice_output_0, %/layers.1/Constant_7_output_0)
%/layers.1/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_4_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_848, %onnx::Conv_849)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_851, %onnx::Conv_852)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_854, %onnx::Conv_855)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/Add_1_output_0, %/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_857, %onnx::Conv_858)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.2/Slice_output_0 = Slice(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_4_output_0, %/layers.2/Constant_5_output_0, %/layers.2/Constant_3_output_0, %/layers.2/Constant_6_output_0)
%/layers.2/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_4_output_0 = Add(%/layers.2/Slice_output_0, %/layers.2/Constant_7_output_0)
%/layers.2/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_4_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/Add_1_output_0, %/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.3/Slice_output_0 = Slice(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_4_output_0, %/layers.3/Constant_5_output_0, %/layers.3/Constant_3_output_0, %/layers.3/Constant_6_output_0)
%/layers.3/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_4_output_0 = Add(%/layers.3/Slice_output_0, %/layers.3/Constant_7_output_0)
%/layers.3/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_4_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/Add_1_output_0, %/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.5/Slice_output_0 = Slice(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0, %/layers.5/Constant_4_output_0, %/layers.5/Constant_2_output_0, %/layers.5/Constant_5_output_0)
%/layers.5/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/Slice_output_0, %/layers.5/Constant_6_output_0)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_10_output_0 = Constant[value = <Tensor>]()
%/layers.5/Slice_1_output_0 = Slice(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_8_output_0, %/layers.5/Constant_9_output_0, %/layers.5/Constant_7_output_0, %/layers.5/Constant_10_output_0)
%/layers.5/Constant_11_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_4_output_0 = Add(%/layers.5/Slice_1_output_0, %/layers.5/Constant_11_output_0)
%/layers.5/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_4_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/Add_1_output_0, %/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.6/Slice_output_0 = Slice(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0, %/layers.6/Constant_4_output_0, %/layers.6/Constant_2_output_0, %/layers.6/Constant_5_output_0)
%/layers.6/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/Slice_output_0, %/layers.6/Constant_6_output_0)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_10_output_0 = Constant[value = <Tensor>]()
%/layers.6/Slice_1_output_0 = Slice(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_8_output_0, %/layers.6/Constant_9_output_0, %/layers.6/Constant_7_output_0, %/layers.6/Constant_10_output_0)
%/layers.6/Constant_11_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_4_output_0 = Add(%/layers.6/Slice_1_output_0, %/layers.6/Constant_11_output_0)
%/layers.6/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_4_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_908, %onnx::Conv_909)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_911, %onnx::Conv_912)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_914, %onnx::Conv_915)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/Add_1_output_0, %/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_917, %onnx::Conv_918)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.7/Slice_output_0 = Slice(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0, %/layers.7/Constant_4_output_0, %/layers.7/Constant_2_output_0, %/layers.7/Constant_5_output_0)
%/layers.7/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/Slice_output_0, %/layers.7/Constant_6_output_0)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_920, %onnx::Conv_921)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_10_output_0 = Constant[value = <Tensor>]()
%/layers.7/Slice_1_output_0 = Slice(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_8_output_0, %/layers.7/Constant_9_output_0, %/layers.7/Constant_7_output_0, %/layers.7/Constant_10_output_0)
%/layers.7/Constant_11_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_4_output_0 = Add(%/layers.7/Slice_1_output_0, %/layers.7/Constant_11_output_0)
%/layers.7/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_4_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_923, %onnx::Conv_924)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_926, %onnx::Conv_927)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_929, %onnx::Conv_930)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/Add_1_output_0, %/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_932, %onnx::Conv_933)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_935, %onnx::Conv_936)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.9/Slice_output_0 = Slice(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_4_output_0, %/layers.9/Constant_5_output_0, %/layers.9/Constant_3_output_0, %/layers.9/Constant_6_output_0)
%/layers.9/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_4_output_0 = Add(%/layers.9/Slice_output_0, %/layers.9/Constant_7_output_0)
%/layers.9/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_4_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_938, %onnx::Conv_939)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_941, %onnx::Conv_942)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_944, %onnx::Conv_945)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/Add_1_output_0, %/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_947, %onnx::Conv_948)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_950, %onnx::Conv_951)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.10/Slice_output_0 = Slice(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_4_output_0, %/layers.10/Constant_5_output_0, %/layers.10/Constant_3_output_0, %/layers.10/Constant_6_output_0)
%/layers.10/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_4_output_0 = Add(%/layers.10/Slice_output_0, %/layers.10/Constant_7_output_0)
%/layers.10/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_4_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_953, %onnx::Conv_954)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_956, %onnx::Conv_957)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_959, %onnx::Conv_960)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/Add_1_output_0, %/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_962, %onnx::Conv_963)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_965, %onnx::Conv_966)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.11/Slice_output_0 = Slice(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_4_output_0, %/layers.11/Constant_5_output_0, %/layers.11/Constant_3_output_0, %/layers.11/Constant_6_output_0)
%/layers.11/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_4_output_0 = Add(%/layers.11/Slice_output_0, %/layers.11/Constant_7_output_0)
%/layers.11/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_4_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.5/maxpool/MaxPool_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%828 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %828
}
|
val_accuracy
| 91.786861
| 840,065,152
| 2,787,785
|
{'zcp_epe_nas': 78.17051299315162, 'zcp_fisher': 53.00083923339844, 'zcp_flops': 13441042432.0, 'zcp_grad_norm': 134.6407012939453, 'zcp_grasp': 12.382080078125, 'zcp_jacov': -16.055296076223847, 'zcp_l2_norm': 762.9549560546875, 'zcp_nwot': 215.66545965347333, 'zcp_params': 2787785.0, 'zcp_plain': 0.027932440862059003, 'zcp_snip': 675.92724609375, 'zcp_synflow': 110.51715801833173, 'zcp_zen': 82.24288177490234, 'zcp_val_accuracy': 0.922475934028625}
| |
NASBench101_307266
|
NASBench101
|
307266
|
b9e87bb03dcee1166e93435c9742a1dd
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_662[FLOAT, 128x3x3x3]
%onnx::Conv_663[FLOAT, 128]
%onnx::Conv_665[FLOAT, 128x128x1x1]
%onnx::Conv_668[FLOAT, 128x128x1x1]
%onnx::Conv_671[FLOAT, 128x128x1x1]
%onnx::Conv_674[FLOAT, 128x128x1x1]
%onnx::Conv_677[FLOAT, 128x128x1x1]
%onnx::Conv_680[FLOAT, 128x128x1x1]
%onnx::Conv_683[FLOAT, 128x128x1x1]
%onnx::Conv_686[FLOAT, 128x128x1x1]
%onnx::Conv_689[FLOAT, 128x128x1x1]
%onnx::Conv_692[FLOAT, 128x128x1x1]
%onnx::Conv_695[FLOAT, 128x128x1x1]
%onnx::Conv_698[FLOAT, 128x128x1x1]
%onnx::Conv_701[FLOAT, 256x128x1x1]
%onnx::Conv_702[FLOAT, 256]
%onnx::Conv_704[FLOAT, 256x128x1x1]
%onnx::Conv_707[FLOAT, 256x128x1x1]
%onnx::Conv_710[FLOAT, 256x256x1x1]
%onnx::Conv_713[FLOAT, 256x256x1x1]
%onnx::Conv_716[FLOAT, 256x256x1x1]
%onnx::Conv_719[FLOAT, 256x256x1x1]
%onnx::Conv_722[FLOAT, 256x256x1x1]
%onnx::Conv_725[FLOAT, 256x256x1x1]
%onnx::Conv_728[FLOAT, 256x256x1x1]
%onnx::Conv_731[FLOAT, 256x256x1x1]
%onnx::Conv_734[FLOAT, 256x256x1x1]
%onnx::Conv_737[FLOAT, 512x256x1x1]
%onnx::Conv_738[FLOAT, 512]
%onnx::Conv_740[FLOAT, 512x256x1x1]
%onnx::Conv_743[FLOAT, 512x256x1x1]
%onnx::Conv_746[FLOAT, 512x512x1x1]
%onnx::Conv_749[FLOAT, 512x512x1x1]
%onnx::Conv_752[FLOAT, 512x512x1x1]
%onnx::Conv_755[FLOAT, 512x512x1x1]
%onnx::Conv_758[FLOAT, 512x512x1x1]
%onnx::Conv_761[FLOAT, 512x512x1x1]
%onnx::Conv_764[FLOAT, 512x512x1x1]
%onnx::Conv_767[FLOAT, 512x512x1x1]
%onnx::Conv_770[FLOAT, 512x512x1x1]
) {
%onnx::Conv_771 = Identity(%onnx::Conv_738)
%onnx::Conv_768 = Identity(%onnx::Conv_738)
%onnx::Conv_765 = Identity(%onnx::Conv_738)
%onnx::Conv_762 = Identity(%onnx::Conv_738)
%onnx::Conv_759 = Identity(%onnx::Conv_738)
%onnx::Conv_756 = Identity(%onnx::Conv_738)
%onnx::Conv_753 = Identity(%onnx::Conv_738)
%onnx::Conv_750 = Identity(%onnx::Conv_738)
%onnx::Conv_747 = Identity(%onnx::Conv_738)
%onnx::Conv_744 = Identity(%onnx::Conv_738)
%onnx::Conv_741 = Identity(%onnx::Conv_738)
%onnx::Conv_735 = Identity(%onnx::Conv_702)
%onnx::Conv_732 = Identity(%onnx::Conv_702)
%onnx::Conv_729 = Identity(%onnx::Conv_702)
%onnx::Conv_726 = Identity(%onnx::Conv_702)
%onnx::Conv_723 = Identity(%onnx::Conv_702)
%onnx::Conv_720 = Identity(%onnx::Conv_702)
%onnx::Conv_717 = Identity(%onnx::Conv_702)
%onnx::Conv_714 = Identity(%onnx::Conv_702)
%onnx::Conv_711 = Identity(%onnx::Conv_702)
%onnx::Conv_708 = Identity(%onnx::Conv_702)
%onnx::Conv_705 = Identity(%onnx::Conv_702)
%onnx::Conv_699 = Identity(%onnx::Conv_663)
%onnx::Conv_696 = Identity(%onnx::Conv_663)
%onnx::Conv_693 = Identity(%onnx::Conv_663)
%onnx::Conv_690 = Identity(%onnx::Conv_663)
%onnx::Conv_687 = Identity(%onnx::Conv_663)
%onnx::Conv_684 = Identity(%onnx::Conv_663)
%onnx::Conv_681 = Identity(%onnx::Conv_663)
%onnx::Conv_678 = Identity(%onnx::Conv_663)
%onnx::Conv_675 = Identity(%onnx::Conv_663)
%onnx::Conv_672 = Identity(%onnx::Conv_663)
%onnx::Conv_669 = Identity(%onnx::Conv_663)
%onnx::Conv_666 = Identity(%onnx::Conv_663)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_662, %onnx::Conv_663)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_665, %onnx::Conv_666)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_668, %onnx::Conv_669)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_671, %onnx::Conv_672)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/Add_2_output_0, %/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_674, %onnx::Conv_675)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_4_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_677, %onnx::Conv_678)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_680, %onnx::Conv_681)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_683, %onnx::Conv_684)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/Add_2_output_0, %/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_686, %onnx::Conv_687)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_4_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_689, %onnx::Conv_690)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_692, %onnx::Conv_693)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_695, %onnx::Conv_696)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/Add_2_output_0, %/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_698, %onnx::Conv_699)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_4_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_701, %onnx::Conv_702)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_704, %onnx::Conv_705)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_707, %onnx::Conv_708)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/Add_2_output_0, %/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_710, %onnx::Conv_711)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_4_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_713, %onnx::Conv_714)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_716, %onnx::Conv_717)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_719, %onnx::Conv_720)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/Add_2_output_0, %/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_722, %onnx::Conv_723)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_4_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_725, %onnx::Conv_726)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_728, %onnx::Conv_729)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_731, %onnx::Conv_732)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/Add_2_output_0, %/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_734, %onnx::Conv_735)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_4_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_737, %onnx::Conv_738)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_740, %onnx::Conv_741)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_743, %onnx::Conv_744)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/Add_2_output_0, %/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_746, %onnx::Conv_747)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_4_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_749, %onnx::Conv_750)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_752, %onnx::Conv_753)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_755, %onnx::Conv_756)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/Add_2_output_0, %/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_758, %onnx::Conv_759)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_4_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_761, %onnx::Conv_762)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_764, %onnx::Conv_765)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_767, %onnx::Conv_768)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/Add_2_output_0, %/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_770, %onnx::Conv_771)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_4_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/vertex_op.5/maxpool/MaxPool_output_0)
%660 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %660
}
|
val_accuracy
| 89.152646
| 1,137,453,056
| 3,667,594
|
{'zcp_epe_nas': 131.51864394103876, 'zcp_fisher': 11.560861587524414, 'zcp_flops': 18199248896.0, 'zcp_grad_norm': 74.5648193359375, 'zcp_grasp': -3.93060302734375, 'zcp_jacov': -16.048827316522654, 'zcp_l2_norm': 803.1466674804688, 'zcp_nwot': 228.9122483976611, 'zcp_params': 3667594.0, 'zcp_plain': 0.09805624932050701, 'zcp_snip': 520.6546630859375, 'zcp_synflow': 67.33357855716555, 'zcp_zen': 75.44727325439453, 'zcp_val_accuracy': 0.9287860393524171}
| |
NASBench101_360581
|
NASBench101
|
360581
|
d9f44c657c6cb7e8af947a2f42475a5f
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_563[FLOAT, 128x3x3x3]
%onnx::Conv_564[FLOAT, 128]
%onnx::Conv_566[FLOAT, 128x128x1x1]
%onnx::Conv_569[FLOAT, 128x128x1x1]
%onnx::Conv_572[FLOAT, 128x128x1x1]
%onnx::Conv_575[FLOAT, 128x128x1x1]
%onnx::Conv_578[FLOAT, 128x128x1x1]
%onnx::Conv_581[FLOAT, 128x128x1x1]
%onnx::Conv_584[FLOAT, 128x128x1x1]
%onnx::Conv_587[FLOAT, 128x128x1x1]
%onnx::Conv_590[FLOAT, 128x128x1x1]
%onnx::Conv_593[FLOAT, 256x128x1x1]
%onnx::Conv_594[FLOAT, 256]
%onnx::Conv_596[FLOAT, 256x256x1x1]
%onnx::Conv_599[FLOAT, 256x128x1x1]
%onnx::Conv_602[FLOAT, 256x256x1x1]
%onnx::Conv_605[FLOAT, 256x256x1x1]
%onnx::Conv_608[FLOAT, 256x256x1x1]
%onnx::Conv_611[FLOAT, 256x256x1x1]
%onnx::Conv_614[FLOAT, 256x256x1x1]
%onnx::Conv_617[FLOAT, 256x256x1x1]
%onnx::Conv_620[FLOAT, 512x256x1x1]
%onnx::Conv_621[FLOAT, 512]
%onnx::Conv_623[FLOAT, 512x512x1x1]
%onnx::Conv_626[FLOAT, 512x256x1x1]
%onnx::Conv_629[FLOAT, 512x512x1x1]
%onnx::Conv_632[FLOAT, 512x512x1x1]
%onnx::Conv_635[FLOAT, 512x512x1x1]
%onnx::Conv_638[FLOAT, 512x512x1x1]
%onnx::Conv_641[FLOAT, 512x512x1x1]
%onnx::Conv_644[FLOAT, 512x512x1x1]
) {
%onnx::Conv_645 = Identity(%onnx::Conv_621)
%onnx::Conv_642 = Identity(%onnx::Conv_621)
%onnx::Conv_639 = Identity(%onnx::Conv_621)
%onnx::Conv_636 = Identity(%onnx::Conv_621)
%onnx::Conv_633 = Identity(%onnx::Conv_621)
%onnx::Conv_630 = Identity(%onnx::Conv_621)
%onnx::Conv_627 = Identity(%onnx::Conv_621)
%onnx::Conv_624 = Identity(%onnx::Conv_621)
%onnx::Conv_618 = Identity(%onnx::Conv_594)
%onnx::Conv_615 = Identity(%onnx::Conv_594)
%onnx::Conv_612 = Identity(%onnx::Conv_594)
%onnx::Conv_609 = Identity(%onnx::Conv_594)
%onnx::Conv_606 = Identity(%onnx::Conv_594)
%onnx::Conv_603 = Identity(%onnx::Conv_594)
%onnx::Conv_600 = Identity(%onnx::Conv_594)
%onnx::Conv_597 = Identity(%onnx::Conv_594)
%onnx::Conv_591 = Identity(%onnx::Conv_564)
%onnx::Conv_588 = Identity(%onnx::Conv_564)
%onnx::Conv_585 = Identity(%onnx::Conv_564)
%onnx::Conv_582 = Identity(%onnx::Conv_564)
%onnx::Conv_579 = Identity(%onnx::Conv_564)
%onnx::Conv_576 = Identity(%onnx::Conv_564)
%onnx::Conv_573 = Identity(%onnx::Conv_564)
%onnx::Conv_570 = Identity(%onnx::Conv_564)
%onnx::Conv_567 = Identity(%onnx::Conv_564)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_563, %onnx::Conv_564)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_566, %onnx::Conv_567)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_569, %onnx::Conv_570)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_572, %onnx::Conv_573)
%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/Add_2_output_0, %/layers.1/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/Add_3_output_0, %/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_4_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_575, %onnx::Conv_576)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_578, %onnx::Conv_579)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_581, %onnx::Conv_582)
%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/Add_2_output_0, %/layers.2/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/Add_3_output_0, %/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_4_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_584, %onnx::Conv_585)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_587, %onnx::Conv_588)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_590, %onnx::Conv_591)
%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/Add_2_output_0, %/layers.3/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/Add_3_output_0, %/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_4_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_593, %onnx::Conv_594)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_596, %onnx::Conv_597)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_599, %onnx::Conv_600)
%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/Add_2_output_0, %/layers.5/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/Add_3_output_0, %/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_4_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_602, %onnx::Conv_603)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_605, %onnx::Conv_606)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_608, %onnx::Conv_609)
%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/Add_2_output_0, %/layers.6/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/Add_3_output_0, %/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_4_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_611, %onnx::Conv_612)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_614, %onnx::Conv_615)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_617, %onnx::Conv_618)
%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/Add_2_output_0, %/layers.7/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/Add_3_output_0, %/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_4_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_620, %onnx::Conv_621)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_623, %onnx::Conv_624)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_626, %onnx::Conv_627)
%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/Add_2_output_0, %/layers.9/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/Add_3_output_0, %/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_4_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_629, %onnx::Conv_630)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_632, %onnx::Conv_633)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_635, %onnx::Conv_636)
%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/Add_2_output_0, %/layers.10/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/Add_3_output_0, %/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_4_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_638, %onnx::Conv_639)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_641, %onnx::Conv_642)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_644, %onnx::Conv_645)
%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/Add_2_output_0, %/layers.11/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/Add_3_output_0, %/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_4_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/vertex_op.5/maxpool/MaxPool_output_0)
%561 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %561
}
|
val_accuracy
| 87.880611
| 863,512,576
| 2,793,866
|
{'zcp_epe_nas': 153.4664463424414, 'zcp_fisher': 33.20349884033203, 'zcp_flops': 13816201216.0, 'zcp_grad_norm': 113.41661071777344, 'zcp_grasp': -117.37548828125, 'zcp_jacov': -16.048166452287816, 'zcp_l2_norm': 607.0596313476562, 'zcp_nwot': 224.42371454532963, 'zcp_params': 2793866.0, 'zcp_plain': 0.46190971136093106, 'zcp_snip': 763.4400024414062, 'zcp_synflow': 63.98364564590116, 'zcp_zen': 62.297462463378906, 'zcp_val_accuracy': 0.890224337577819}
| |
NASBench101_78448
|
NASBench101
|
78448
|
2f8fc3abf45df76f1f870aa58c2c823d
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_860[FLOAT, 128x3x3x3]
%onnx::Conv_861[FLOAT, 128]
%onnx::Conv_863[FLOAT, 128x128x1x1]
%onnx::Conv_866[FLOAT, 128x128x3x3]
%onnx::Conv_869[FLOAT, 128x128x1x1]
%onnx::Conv_872[FLOAT, 128x128x1x1]
%onnx::Conv_875[FLOAT, 128x128x1x1]
%onnx::Conv_878[FLOAT, 128x128x3x3]
%onnx::Conv_881[FLOAT, 128x128x1x1]
%onnx::Conv_884[FLOAT, 128x128x3x3]
%onnx::Conv_887[FLOAT, 128x128x1x1]
%onnx::Conv_890[FLOAT, 128x128x1x1]
%onnx::Conv_893[FLOAT, 128x128x1x1]
%onnx::Conv_896[FLOAT, 128x128x3x3]
%onnx::Conv_899[FLOAT, 128x128x1x1]
%onnx::Conv_902[FLOAT, 128x128x3x3]
%onnx::Conv_905[FLOAT, 128x128x1x1]
%onnx::Conv_908[FLOAT, 128x128x1x1]
%onnx::Conv_911[FLOAT, 128x128x1x1]
%onnx::Conv_914[FLOAT, 128x128x3x3]
%onnx::Conv_917[FLOAT, 256x128x1x1]
%onnx::Conv_918[FLOAT, 256]
%onnx::Conv_920[FLOAT, 256x256x3x3]
%onnx::Conv_923[FLOAT, 256x128x1x1]
%onnx::Conv_926[FLOAT, 256x256x1x1]
%onnx::Conv_929[FLOAT, 256x128x1x1]
%onnx::Conv_932[FLOAT, 256x256x3x3]
%onnx::Conv_935[FLOAT, 256x256x1x1]
%onnx::Conv_938[FLOAT, 256x256x3x3]
%onnx::Conv_941[FLOAT, 256x256x1x1]
%onnx::Conv_944[FLOAT, 256x256x1x1]
%onnx::Conv_947[FLOAT, 256x256x1x1]
%onnx::Conv_950[FLOAT, 256x256x3x3]
%onnx::Conv_953[FLOAT, 256x256x1x1]
%onnx::Conv_956[FLOAT, 256x256x3x3]
%onnx::Conv_959[FLOAT, 256x256x1x1]
%onnx::Conv_962[FLOAT, 256x256x1x1]
%onnx::Conv_965[FLOAT, 256x256x1x1]
%onnx::Conv_968[FLOAT, 256x256x3x3]
%onnx::Conv_971[FLOAT, 512x256x1x1]
%onnx::Conv_972[FLOAT, 512]
%onnx::Conv_974[FLOAT, 512x512x3x3]
%onnx::Conv_977[FLOAT, 512x256x1x1]
%onnx::Conv_980[FLOAT, 512x512x1x1]
%onnx::Conv_983[FLOAT, 512x256x1x1]
%onnx::Conv_986[FLOAT, 512x512x3x3]
%onnx::Conv_989[FLOAT, 512x512x1x1]
%onnx::Conv_992[FLOAT, 512x512x3x3]
%onnx::Conv_995[FLOAT, 512x512x1x1]
%onnx::Conv_998[FLOAT, 512x512x1x1]
%onnx::Conv_1001[FLOAT, 512x512x1x1]
%onnx::Conv_1004[FLOAT, 512x512x3x3]
%onnx::Conv_1007[FLOAT, 512x512x1x1]
%onnx::Conv_1010[FLOAT, 512x512x3x3]
%onnx::Conv_1013[FLOAT, 512x512x1x1]
%onnx::Conv_1016[FLOAT, 512x512x1x1]
%onnx::Conv_1019[FLOAT, 512x512x1x1]
%onnx::Conv_1022[FLOAT, 512x512x3x3]
) {
%onnx::Conv_1023 = Identity(%onnx::Conv_972)
%onnx::Conv_1020 = Identity(%onnx::Conv_972)
%onnx::Conv_1017 = Identity(%onnx::Conv_972)
%onnx::Conv_1014 = Identity(%onnx::Conv_972)
%onnx::Conv_1011 = Identity(%onnx::Conv_972)
%onnx::Conv_1008 = Identity(%onnx::Conv_972)
%onnx::Conv_1005 = Identity(%onnx::Conv_972)
%onnx::Conv_1002 = Identity(%onnx::Conv_972)
%onnx::Conv_999 = Identity(%onnx::Conv_972)
%onnx::Conv_996 = Identity(%onnx::Conv_972)
%onnx::Conv_993 = Identity(%onnx::Conv_972)
%onnx::Conv_990 = Identity(%onnx::Conv_972)
%onnx::Conv_987 = Identity(%onnx::Conv_972)
%onnx::Conv_984 = Identity(%onnx::Conv_972)
%onnx::Conv_981 = Identity(%onnx::Conv_972)
%onnx::Conv_978 = Identity(%onnx::Conv_972)
%onnx::Conv_975 = Identity(%onnx::Conv_972)
%onnx::Conv_969 = Identity(%onnx::Conv_918)
%onnx::Conv_966 = Identity(%onnx::Conv_918)
%onnx::Conv_963 = Identity(%onnx::Conv_918)
%onnx::Conv_960 = Identity(%onnx::Conv_918)
%onnx::Conv_957 = Identity(%onnx::Conv_918)
%onnx::Conv_954 = Identity(%onnx::Conv_918)
%onnx::Conv_951 = Identity(%onnx::Conv_918)
%onnx::Conv_948 = Identity(%onnx::Conv_918)
%onnx::Conv_945 = Identity(%onnx::Conv_918)
%onnx::Conv_942 = Identity(%onnx::Conv_918)
%onnx::Conv_939 = Identity(%onnx::Conv_918)
%onnx::Conv_936 = Identity(%onnx::Conv_918)
%onnx::Conv_933 = Identity(%onnx::Conv_918)
%onnx::Conv_930 = Identity(%onnx::Conv_918)
%onnx::Conv_927 = Identity(%onnx::Conv_918)
%onnx::Conv_924 = Identity(%onnx::Conv_918)
%onnx::Conv_921 = Identity(%onnx::Conv_918)
%onnx::Conv_915 = Identity(%onnx::Conv_861)
%onnx::Conv_912 = Identity(%onnx::Conv_861)
%onnx::Conv_909 = Identity(%onnx::Conv_861)
%onnx::Conv_906 = Identity(%onnx::Conv_861)
%onnx::Conv_903 = Identity(%onnx::Conv_861)
%onnx::Conv_900 = Identity(%onnx::Conv_861)
%onnx::Conv_897 = Identity(%onnx::Conv_861)
%onnx::Conv_894 = Identity(%onnx::Conv_861)
%onnx::Conv_891 = Identity(%onnx::Conv_861)
%onnx::Conv_888 = Identity(%onnx::Conv_861)
%onnx::Conv_885 = Identity(%onnx::Conv_861)
%onnx::Conv_882 = Identity(%onnx::Conv_861)
%onnx::Conv_879 = Identity(%onnx::Conv_861)
%onnx::Conv_876 = Identity(%onnx::Conv_861)
%onnx::Conv_873 = Identity(%onnx::Conv_861)
%onnx::Conv_870 = Identity(%onnx::Conv_861)
%onnx::Conv_867 = Identity(%onnx::Conv_861)
%onnx::Conv_864 = Identity(%onnx::Conv_861)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_860, %onnx::Conv_861)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/Add_3_output_0, %/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_4_output_0)
%/layers.1/Add_5_output_0 = Add(%/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %/layers.1/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_5_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/Add_3_output_0, %/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_4_output_0)
%/layers.2/Add_5_output_0 = Add(%/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %/layers.2/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_5_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_908, %onnx::Conv_909)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_911, %onnx::Conv_912)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/Add_3_output_0, %/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_4_output_0)
%/layers.3/Add_5_output_0 = Add(%/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %/layers.3/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_5_output_0, %onnx::Conv_914, %onnx::Conv_915)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_917, %onnx::Conv_918)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_920, %onnx::Conv_921)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_923, %onnx::Conv_924)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_926, %onnx::Conv_927)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_929, %onnx::Conv_930)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/Add_3_output_0, %/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_4_output_0)
%/layers.5/Add_5_output_0 = Add(%/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %/layers.5/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_5_output_0, %onnx::Conv_932, %onnx::Conv_933)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_935, %onnx::Conv_936)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_938, %onnx::Conv_939)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_941, %onnx::Conv_942)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_944, %onnx::Conv_945)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_947, %onnx::Conv_948)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/Add_3_output_0, %/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_4_output_0)
%/layers.6/Add_5_output_0 = Add(%/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %/layers.6/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_5_output_0, %onnx::Conv_950, %onnx::Conv_951)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_953, %onnx::Conv_954)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_956, %onnx::Conv_957)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_959, %onnx::Conv_960)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_962, %onnx::Conv_963)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_965, %onnx::Conv_966)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/Add_3_output_0, %/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_4_output_0)
%/layers.7/Add_5_output_0 = Add(%/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %/layers.7/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_5_output_0, %onnx::Conv_968, %onnx::Conv_969)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_974, %onnx::Conv_975)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_977, %onnx::Conv_978)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/Add_3_output_0, %/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_4_output_0)
%/layers.9/Add_5_output_0 = Add(%/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %/layers.9/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_5_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/Add_3_output_0, %/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_4_output_0)
%/layers.10/Add_5_output_0 = Add(%/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %/layers.10/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_5_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/Add_3_output_0, %/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_4_output_0)
%/layers.11/Add_5_output_0 = Add(%/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %/layers.11/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_5_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%858 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %858
}
|
val_accuracy
| 92.477965
| 6,584,281,088
| 22,257,802
|
{'zcp_epe_nas': 86.31888145550569, 'zcp_fisher': 46.62611389160156, 'zcp_flops': 105348497408.0, 'zcp_grad_norm': 116.17855834960938, 'zcp_grasp': -10.4940185546875, 'zcp_jacov': -16.058667419864662, 'zcp_l2_norm': 1225.9224853515625, 'zcp_nwot': 234.0327698198668, 'zcp_params': 22257802.0, 'zcp_plain': 0.041134878993034, 'zcp_snip': 961.6873168945312, 'zcp_synflow': 130.25888415697938, 'zcp_zen': 114.28446960449219, 'zcp_val_accuracy': 0.891426265239715}
| |
NASBench101_332723
|
NASBench101
|
332723
|
c93e890624e49208e3fb3910e7d76c23
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_1085[FLOAT, 128x3x3x3]
%onnx::Conv_1086[FLOAT, 128]
%onnx::Conv_1088[FLOAT, 64x128x1x1]
%onnx::Conv_1089[FLOAT, 64]
%onnx::Conv_1091[FLOAT, 64x128x1x1]
%onnx::Conv_1094[FLOAT, 64x64x3x3]
%onnx::Conv_1097[FLOAT, 64x128x1x1]
%onnx::Conv_1100[FLOAT, 64x64x1x1]
%onnx::Conv_1103[FLOAT, 64x64x1x1]
%onnx::Conv_1106[FLOAT, 64x128x1x1]
%onnx::Conv_1109[FLOAT, 64x64x3x3]
%onnx::Conv_1112[FLOAT, 64x128x1x1]
%onnx::Conv_1115[FLOAT, 64x128x1x1]
%onnx::Conv_1118[FLOAT, 64x64x3x3]
%onnx::Conv_1121[FLOAT, 64x128x1x1]
%onnx::Conv_1124[FLOAT, 64x64x1x1]
%onnx::Conv_1127[FLOAT, 64x64x1x1]
%onnx::Conv_1130[FLOAT, 64x128x1x1]
%onnx::Conv_1133[FLOAT, 64x64x3x3]
%onnx::Conv_1136[FLOAT, 64x128x1x1]
%onnx::Conv_1139[FLOAT, 64x128x1x1]
%onnx::Conv_1142[FLOAT, 64x64x3x3]
%onnx::Conv_1145[FLOAT, 64x128x1x1]
%onnx::Conv_1148[FLOAT, 64x64x1x1]
%onnx::Conv_1151[FLOAT, 64x64x1x1]
%onnx::Conv_1154[FLOAT, 64x128x1x1]
%onnx::Conv_1157[FLOAT, 64x64x3x3]
%onnx::Conv_1160[FLOAT, 128x128x1x1]
%onnx::Conv_1163[FLOAT, 128x128x1x1]
%onnx::Conv_1166[FLOAT, 128x128x3x3]
%onnx::Conv_1169[FLOAT, 128x128x1x1]
%onnx::Conv_1172[FLOAT, 128x128x1x1]
%onnx::Conv_1175[FLOAT, 128x128x1x1]
%onnx::Conv_1178[FLOAT, 128x128x1x1]
%onnx::Conv_1181[FLOAT, 128x128x3x3]
%onnx::Conv_1184[FLOAT, 128x256x1x1]
%onnx::Conv_1187[FLOAT, 128x256x1x1]
%onnx::Conv_1190[FLOAT, 128x128x3x3]
%onnx::Conv_1193[FLOAT, 128x256x1x1]
%onnx::Conv_1196[FLOAT, 128x128x1x1]
%onnx::Conv_1199[FLOAT, 128x128x1x1]
%onnx::Conv_1202[FLOAT, 128x256x1x1]
%onnx::Conv_1205[FLOAT, 128x128x3x3]
%onnx::Conv_1208[FLOAT, 128x256x1x1]
%onnx::Conv_1211[FLOAT, 128x256x1x1]
%onnx::Conv_1214[FLOAT, 128x128x3x3]
%onnx::Conv_1217[FLOAT, 128x256x1x1]
%onnx::Conv_1220[FLOAT, 128x128x1x1]
%onnx::Conv_1223[FLOAT, 128x128x1x1]
%onnx::Conv_1226[FLOAT, 128x256x1x1]
%onnx::Conv_1229[FLOAT, 128x128x3x3]
%onnx::Conv_1232[FLOAT, 256x256x1x1]
%onnx::Conv_1233[FLOAT, 256]
%onnx::Conv_1235[FLOAT, 256x256x1x1]
%onnx::Conv_1238[FLOAT, 256x256x3x3]
%onnx::Conv_1241[FLOAT, 256x256x1x1]
%onnx::Conv_1244[FLOAT, 256x256x1x1]
%onnx::Conv_1247[FLOAT, 256x256x1x1]
%onnx::Conv_1250[FLOAT, 256x256x1x1]
%onnx::Conv_1253[FLOAT, 256x256x3x3]
%onnx::Conv_1256[FLOAT, 256x512x1x1]
%onnx::Conv_1259[FLOAT, 256x512x1x1]
%onnx::Conv_1262[FLOAT, 256x256x3x3]
%onnx::Conv_1265[FLOAT, 256x512x1x1]
%onnx::Conv_1268[FLOAT, 256x256x1x1]
%onnx::Conv_1271[FLOAT, 256x256x1x1]
%onnx::Conv_1274[FLOAT, 256x512x1x1]
%onnx::Conv_1277[FLOAT, 256x256x3x3]
%onnx::Conv_1280[FLOAT, 256x512x1x1]
%onnx::Conv_1283[FLOAT, 256x512x1x1]
%onnx::Conv_1286[FLOAT, 256x256x3x3]
%onnx::Conv_1289[FLOAT, 256x512x1x1]
%onnx::Conv_1292[FLOAT, 256x256x1x1]
%onnx::Conv_1295[FLOAT, 256x256x1x1]
%onnx::Conv_1298[FLOAT, 256x512x1x1]
%onnx::Conv_1301[FLOAT, 256x256x3x3]
) {
%onnx::Conv_1302 = Identity(%onnx::Conv_1233)
%onnx::Conv_1299 = Identity(%onnx::Conv_1233)
%onnx::Conv_1296 = Identity(%onnx::Conv_1233)
%onnx::Conv_1293 = Identity(%onnx::Conv_1233)
%onnx::Conv_1290 = Identity(%onnx::Conv_1233)
%onnx::Conv_1287 = Identity(%onnx::Conv_1233)
%onnx::Conv_1284 = Identity(%onnx::Conv_1233)
%onnx::Conv_1281 = Identity(%onnx::Conv_1233)
%onnx::Conv_1278 = Identity(%onnx::Conv_1233)
%onnx::Conv_1275 = Identity(%onnx::Conv_1233)
%onnx::Conv_1272 = Identity(%onnx::Conv_1233)
%onnx::Conv_1269 = Identity(%onnx::Conv_1233)
%onnx::Conv_1266 = Identity(%onnx::Conv_1233)
%onnx::Conv_1263 = Identity(%onnx::Conv_1233)
%onnx::Conv_1260 = Identity(%onnx::Conv_1233)
%onnx::Conv_1257 = Identity(%onnx::Conv_1233)
%onnx::Conv_1254 = Identity(%onnx::Conv_1233)
%onnx::Conv_1251 = Identity(%onnx::Conv_1233)
%onnx::Conv_1248 = Identity(%onnx::Conv_1233)
%onnx::Conv_1245 = Identity(%onnx::Conv_1233)
%onnx::Conv_1242 = Identity(%onnx::Conv_1233)
%onnx::Conv_1239 = Identity(%onnx::Conv_1233)
%onnx::Conv_1236 = Identity(%onnx::Conv_1233)
%onnx::Conv_1230 = Identity(%onnx::Conv_1086)
%onnx::Conv_1227 = Identity(%onnx::Conv_1086)
%onnx::Conv_1224 = Identity(%onnx::Conv_1086)
%onnx::Conv_1221 = Identity(%onnx::Conv_1086)
%onnx::Conv_1218 = Identity(%onnx::Conv_1086)
%onnx::Conv_1215 = Identity(%onnx::Conv_1086)
%onnx::Conv_1212 = Identity(%onnx::Conv_1086)
%onnx::Conv_1209 = Identity(%onnx::Conv_1086)
%onnx::Conv_1206 = Identity(%onnx::Conv_1086)
%onnx::Conv_1203 = Identity(%onnx::Conv_1086)
%onnx::Conv_1200 = Identity(%onnx::Conv_1086)
%onnx::Conv_1197 = Identity(%onnx::Conv_1086)
%onnx::Conv_1194 = Identity(%onnx::Conv_1086)
%onnx::Conv_1191 = Identity(%onnx::Conv_1086)
%onnx::Conv_1188 = Identity(%onnx::Conv_1086)
%onnx::Conv_1185 = Identity(%onnx::Conv_1086)
%onnx::Conv_1182 = Identity(%onnx::Conv_1086)
%onnx::Conv_1179 = Identity(%onnx::Conv_1086)
%onnx::Conv_1176 = Identity(%onnx::Conv_1086)
%onnx::Conv_1173 = Identity(%onnx::Conv_1086)
%onnx::Conv_1170 = Identity(%onnx::Conv_1086)
%onnx::Conv_1167 = Identity(%onnx::Conv_1086)
%onnx::Conv_1164 = Identity(%onnx::Conv_1086)
%onnx::Conv_1161 = Identity(%onnx::Conv_1086)
%onnx::Conv_1158 = Identity(%onnx::Conv_1089)
%onnx::Conv_1155 = Identity(%onnx::Conv_1089)
%onnx::Conv_1152 = Identity(%onnx::Conv_1089)
%onnx::Conv_1149 = Identity(%onnx::Conv_1089)
%onnx::Conv_1146 = Identity(%onnx::Conv_1089)
%onnx::Conv_1143 = Identity(%onnx::Conv_1089)
%onnx::Conv_1140 = Identity(%onnx::Conv_1089)
%onnx::Conv_1137 = Identity(%onnx::Conv_1089)
%onnx::Conv_1134 = Identity(%onnx::Conv_1089)
%onnx::Conv_1131 = Identity(%onnx::Conv_1089)
%onnx::Conv_1128 = Identity(%onnx::Conv_1089)
%onnx::Conv_1125 = Identity(%onnx::Conv_1089)
%onnx::Conv_1122 = Identity(%onnx::Conv_1089)
%onnx::Conv_1119 = Identity(%onnx::Conv_1089)
%onnx::Conv_1116 = Identity(%onnx::Conv_1089)
%onnx::Conv_1113 = Identity(%onnx::Conv_1089)
%onnx::Conv_1110 = Identity(%onnx::Conv_1089)
%onnx::Conv_1107 = Identity(%onnx::Conv_1089)
%onnx::Conv_1104 = Identity(%onnx::Conv_1089)
%onnx::Conv_1101 = Identity(%onnx::Conv_1089)
%onnx::Conv_1098 = Identity(%onnx::Conv_1089)
%onnx::Conv_1095 = Identity(%onnx::Conv_1089)
%onnx::Conv_1092 = Identity(%onnx::Conv_1089)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_1085, %onnx::Conv_1086)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1088, %onnx::Conv_1089)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1091, %onnx::Conv_1092)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_1094, %onnx::Conv_1095)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1097, %onnx::Conv_1098)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/Add_2_output_0, %/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_1100, %onnx::Conv_1101)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_1103, %onnx::Conv_1104)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1106, %onnx::Conv_1107)
%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_5_output_0 = Add(%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_4_output_0)
%/layers.1/Add_6_output_0 = Add(%/layers.1/Add_5_output_0, %/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_6_output_0, %onnx::Conv_1109, %onnx::Conv_1110)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_1112, %onnx::Conv_1113)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_1115, %onnx::Conv_1116)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_1118, %onnx::Conv_1119)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_1121, %onnx::Conv_1122)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/Add_2_output_0, %/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_1124, %onnx::Conv_1125)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_1127, %onnx::Conv_1128)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_1130, %onnx::Conv_1131)
%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_5_output_0 = Add(%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_4_output_0)
%/layers.2/Add_6_output_0 = Add(%/layers.2/Add_5_output_0, %/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_6_output_0, %onnx::Conv_1133, %onnx::Conv_1134)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_1136, %onnx::Conv_1137)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_1139, %onnx::Conv_1140)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_1142, %onnx::Conv_1143)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_1145, %onnx::Conv_1146)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/Add_2_output_0, %/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_1148, %onnx::Conv_1149)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_4_output_0, %onnx::Conv_1151, %onnx::Conv_1152)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_1154, %onnx::Conv_1155)
%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_5_output_0 = Add(%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_4_output_0)
%/layers.3/Add_6_output_0 = Add(%/layers.3/Add_5_output_0, %/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_6_output_0, %onnx::Conv_1157, %onnx::Conv_1158)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1160, %onnx::Conv_1161)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1163, %onnx::Conv_1164)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_1166, %onnx::Conv_1167)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1169, %onnx::Conv_1170)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/Add_2_output_0, %/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_1172, %onnx::Conv_1173)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_1175, %onnx::Conv_1176)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1178, %onnx::Conv_1179)
%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_5_output_0 = Add(%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_4_output_0)
%/layers.5/Add_6_output_0 = Add(%/layers.5/Add_5_output_0, %/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_6_output_0, %onnx::Conv_1181, %onnx::Conv_1182)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_1184, %onnx::Conv_1185)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_1187, %onnx::Conv_1188)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_1190, %onnx::Conv_1191)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_1193, %onnx::Conv_1194)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/Add_2_output_0, %/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_1196, %onnx::Conv_1197)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_1199, %onnx::Conv_1200)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_1202, %onnx::Conv_1203)
%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_5_output_0 = Add(%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_4_output_0)
%/layers.6/Add_6_output_0 = Add(%/layers.6/Add_5_output_0, %/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_6_output_0, %onnx::Conv_1205, %onnx::Conv_1206)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_1208, %onnx::Conv_1209)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_1211, %onnx::Conv_1212)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_1214, %onnx::Conv_1215)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_1217, %onnx::Conv_1218)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/Add_2_output_0, %/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_1220, %onnx::Conv_1221)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_4_output_0, %onnx::Conv_1223, %onnx::Conv_1224)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_1226, %onnx::Conv_1227)
%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_5_output_0 = Add(%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_4_output_0)
%/layers.7/Add_6_output_0 = Add(%/layers.7/Add_5_output_0, %/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_6_output_0, %onnx::Conv_1229, %onnx::Conv_1230)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1232, %onnx::Conv_1233)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1235, %onnx::Conv_1236)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_1238, %onnx::Conv_1239)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1241, %onnx::Conv_1242)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/Add_2_output_0, %/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_1244, %onnx::Conv_1245)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_1247, %onnx::Conv_1248)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1250, %onnx::Conv_1251)
%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_5_output_0 = Add(%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_4_output_0)
%/layers.9/Add_6_output_0 = Add(%/layers.9/Add_5_output_0, %/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_6_output_0, %onnx::Conv_1253, %onnx::Conv_1254)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_1256, %onnx::Conv_1257)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_1259, %onnx::Conv_1260)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_1262, %onnx::Conv_1263)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_1265, %onnx::Conv_1266)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/Add_2_output_0, %/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_1268, %onnx::Conv_1269)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_1271, %onnx::Conv_1272)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_1274, %onnx::Conv_1275)
%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_5_output_0 = Add(%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_4_output_0)
%/layers.10/Add_6_output_0 = Add(%/layers.10/Add_5_output_0, %/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_6_output_0, %onnx::Conv_1277, %onnx::Conv_1278)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1280, %onnx::Conv_1281)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1283, %onnx::Conv_1284)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_1286, %onnx::Conv_1287)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1289, %onnx::Conv_1290)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/Add_2_output_0, %/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_1292, %onnx::Conv_1293)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_4_output_0, %onnx::Conv_1295, %onnx::Conv_1296)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1298, %onnx::Conv_1299)
%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_5_output_0 = Add(%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_4_output_0)
%/layers.11/Add_6_output_0 = Add(%/layers.11/Add_5_output_0, %/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_6_output_0, %onnx::Conv_1301, %onnx::Conv_1302)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%1083 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %1083
}
|
val_accuracy
| 92.838544
| 2,076,977,152
| 6,928,010
|
{'zcp_epe_nas': 88.95476724671964, 'zcp_fisher': 6.9454731941223145, 'zcp_flops': 33231634432.0, 'zcp_grad_norm': 63.99729919433594, 'zcp_grasp': -1.90155029296875, 'zcp_jacov': -16.052011490539424, 'zcp_l2_norm': 1386.12060546875, 'zcp_nwot': 228.94947097464984, 'zcp_params': 6928010.0, 'zcp_plain': 0.047811817377805, 'zcp_snip': 430.2031555175781, 'zcp_synflow': 139.91230227630516, 'zcp_zen': 124.39988708496094, 'zcp_val_accuracy': 0.912760436534881}
| |
NASBench101_157920
|
NASBench101
|
157920
|
5f9a646af14cf9d5b085c127fb04b64d
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_716[FLOAT, 128x3x3x3]
%onnx::Conv_717[FLOAT, 128]
%onnx::Conv_719[FLOAT, 43x128x1x1]
%onnx::Conv_720[FLOAT, 43]
%onnx::Conv_722[FLOAT, 43x43x1x1]
%onnx::Conv_725[FLOAT, 42x42x3x3]
%onnx::Conv_726[FLOAT, 42]
%onnx::Conv_728[FLOAT, 42x42x1x1]
%onnx::Conv_731[FLOAT, 43x128x1x1]
%onnx::Conv_734[FLOAT, 43x43x1x1]
%onnx::Conv_737[FLOAT, 42x42x3x3]
%onnx::Conv_740[FLOAT, 42x42x1x1]
%onnx::Conv_743[FLOAT, 43x128x1x1]
%onnx::Conv_746[FLOAT, 43x43x1x1]
%onnx::Conv_749[FLOAT, 42x42x3x3]
%onnx::Conv_752[FLOAT, 42x42x1x1]
%onnx::Conv_755[FLOAT, 86x128x1x1]
%onnx::Conv_756[FLOAT, 86]
%onnx::Conv_758[FLOAT, 86x86x1x1]
%onnx::Conv_761[FLOAT, 85x85x3x3]
%onnx::Conv_762[FLOAT, 85]
%onnx::Conv_764[FLOAT, 85x85x1x1]
%onnx::Conv_767[FLOAT, 86x256x1x1]
%onnx::Conv_770[FLOAT, 86x86x1x1]
%onnx::Conv_773[FLOAT, 85x85x3x3]
%onnx::Conv_776[FLOAT, 85x85x1x1]
%onnx::Conv_779[FLOAT, 86x256x1x1]
%onnx::Conv_782[FLOAT, 86x86x1x1]
%onnx::Conv_785[FLOAT, 85x85x3x3]
%onnx::Conv_788[FLOAT, 85x85x1x1]
%onnx::Conv_791[FLOAT, 171x256x1x1]
%onnx::Conv_792[FLOAT, 171]
%onnx::Conv_794[FLOAT, 171x171x1x1]
%onnx::Conv_797[FLOAT, 170x170x3x3]
%onnx::Conv_798[FLOAT, 170]
%onnx::Conv_800[FLOAT, 170x170x1x1]
%onnx::Conv_803[FLOAT, 171x512x1x1]
%onnx::Conv_806[FLOAT, 171x171x1x1]
%onnx::Conv_809[FLOAT, 170x170x3x3]
%onnx::Conv_812[FLOAT, 170x170x1x1]
%onnx::Conv_815[FLOAT, 171x512x1x1]
%onnx::Conv_818[FLOAT, 171x171x1x1]
%onnx::Conv_821[FLOAT, 170x170x3x3]
%onnx::Conv_824[FLOAT, 170x170x1x1]
) {
%onnx::Conv_825 = Identity(%onnx::Conv_798)
%onnx::Conv_822 = Identity(%onnx::Conv_798)
%onnx::Conv_819 = Identity(%onnx::Conv_792)
%onnx::Conv_816 = Identity(%onnx::Conv_792)
%onnx::Conv_813 = Identity(%onnx::Conv_798)
%onnx::Conv_810 = Identity(%onnx::Conv_798)
%onnx::Conv_807 = Identity(%onnx::Conv_792)
%onnx::Conv_804 = Identity(%onnx::Conv_792)
%onnx::Conv_801 = Identity(%onnx::Conv_798)
%onnx::Conv_795 = Identity(%onnx::Conv_792)
%onnx::Conv_789 = Identity(%onnx::Conv_762)
%onnx::Conv_786 = Identity(%onnx::Conv_762)
%onnx::Conv_783 = Identity(%onnx::Conv_756)
%onnx::Conv_780 = Identity(%onnx::Conv_756)
%onnx::Conv_777 = Identity(%onnx::Conv_762)
%onnx::Conv_774 = Identity(%onnx::Conv_762)
%onnx::Conv_771 = Identity(%onnx::Conv_756)
%onnx::Conv_768 = Identity(%onnx::Conv_756)
%onnx::Conv_765 = Identity(%onnx::Conv_762)
%onnx::Conv_759 = Identity(%onnx::Conv_756)
%onnx::Conv_753 = Identity(%onnx::Conv_726)
%onnx::Conv_750 = Identity(%onnx::Conv_726)
%onnx::Conv_747 = Identity(%onnx::Conv_720)
%onnx::Conv_744 = Identity(%onnx::Conv_720)
%onnx::Conv_741 = Identity(%onnx::Conv_726)
%onnx::Conv_738 = Identity(%onnx::Conv_726)
%onnx::Conv_735 = Identity(%onnx::Conv_720)
%onnx::Conv_732 = Identity(%onnx::Conv_720)
%onnx::Conv_729 = Identity(%onnx::Conv_726)
%onnx::Conv_723 = Identity(%onnx::Conv_720)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_716, %onnx::Conv_717)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_719, %onnx::Conv_720)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_722, %onnx::Conv_723)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.1/Slice_output_0 = Slice(%/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %/layers.1/Constant_3_output_0, %/layers.1/Constant_4_output_0, %/layers.1/Constant_2_output_0, %/layers.1/Constant_5_output_0)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Slice_output_0, %onnx::Conv_725, %onnx::Conv_726)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.1/Slice_1_output_0 = Slice(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_7_output_0, %/layers.1/Constant_8_output_0, %/layers.1/Constant_6_output_0, %/layers.1/Constant_9_output_0)
%/layers.1/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/Slice_1_output_0, %/layers.1/Constant_10_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/Add_2_output_0, %/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_728, %onnx::Conv_729)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_731, %onnx::Conv_732)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_734, %onnx::Conv_735)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.2/Slice_output_0 = Slice(%/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %/layers.2/Constant_3_output_0, %/layers.2/Constant_4_output_0, %/layers.2/Constant_2_output_0, %/layers.2/Constant_5_output_0)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Slice_output_0, %onnx::Conv_737, %onnx::Conv_738)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.2/Slice_1_output_0 = Slice(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_7_output_0, %/layers.2/Constant_8_output_0, %/layers.2/Constant_6_output_0, %/layers.2/Constant_9_output_0)
%/layers.2/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/Slice_1_output_0, %/layers.2/Constant_10_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/Add_2_output_0, %/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_740, %onnx::Conv_741)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_743, %onnx::Conv_744)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_746, %onnx::Conv_747)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.3/Slice_output_0 = Slice(%/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %/layers.3/Constant_3_output_0, %/layers.3/Constant_4_output_0, %/layers.3/Constant_2_output_0, %/layers.3/Constant_5_output_0)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Slice_output_0, %onnx::Conv_749, %onnx::Conv_750)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.3/Slice_1_output_0 = Slice(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_7_output_0, %/layers.3/Constant_8_output_0, %/layers.3/Constant_6_output_0, %/layers.3/Constant_9_output_0)
%/layers.3/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/Slice_1_output_0, %/layers.3/Constant_10_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/Add_2_output_0, %/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_752, %onnx::Conv_753)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_755, %onnx::Conv_756)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_758, %onnx::Conv_759)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.5/Slice_output_0 = Slice(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0, %/layers.5/Constant_3_output_0, %/layers.5/Constant_1_output_0, %/layers.5/Constant_4_output_0)
%/layers.5/Constant_5_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/Slice_output_0, %/layers.5/Constant_5_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_761, %onnx::Conv_762)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.5/Slice_1_output_0 = Slice(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_7_output_0, %/layers.5/Constant_8_output_0, %/layers.5/Constant_6_output_0, %/layers.5/Constant_9_output_0)
%/layers.5/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/Slice_1_output_0, %/layers.5/Constant_10_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/Add_2_output_0, %/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_764, %onnx::Conv_765)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_767, %onnx::Conv_768)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_770, %onnx::Conv_771)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.6/Slice_output_0 = Slice(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0, %/layers.6/Constant_3_output_0, %/layers.6/Constant_1_output_0, %/layers.6/Constant_4_output_0)
%/layers.6/Constant_5_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/Slice_output_0, %/layers.6/Constant_5_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_773, %onnx::Conv_774)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.6/Slice_1_output_0 = Slice(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_7_output_0, %/layers.6/Constant_8_output_0, %/layers.6/Constant_6_output_0, %/layers.6/Constant_9_output_0)
%/layers.6/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/Slice_1_output_0, %/layers.6/Constant_10_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/Add_2_output_0, %/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_776, %onnx::Conv_777)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_779, %onnx::Conv_780)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_782, %onnx::Conv_783)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.7/Slice_output_0 = Slice(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0, %/layers.7/Constant_3_output_0, %/layers.7/Constant_1_output_0, %/layers.7/Constant_4_output_0)
%/layers.7/Constant_5_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/Slice_output_0, %/layers.7/Constant_5_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_785, %onnx::Conv_786)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.7/Slice_1_output_0 = Slice(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_7_output_0, %/layers.7/Constant_8_output_0, %/layers.7/Constant_6_output_0, %/layers.7/Constant_9_output_0)
%/layers.7/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/Slice_1_output_0, %/layers.7/Constant_10_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/Add_2_output_0, %/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_788, %onnx::Conv_789)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.9/Slice_output_0 = Slice(%/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %/layers.9/Constant_3_output_0, %/layers.9/Constant_4_output_0, %/layers.9/Constant_2_output_0, %/layers.9/Constant_5_output_0)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Slice_output_0, %onnx::Conv_797, %onnx::Conv_798)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.9/Slice_1_output_0 = Slice(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_7_output_0, %/layers.9/Constant_8_output_0, %/layers.9/Constant_6_output_0, %/layers.9/Constant_9_output_0)
%/layers.9/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/Slice_1_output_0, %/layers.9/Constant_10_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/Add_2_output_0, %/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_800, %onnx::Conv_801)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_803, %onnx::Conv_804)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_806, %onnx::Conv_807)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.10/Slice_output_0 = Slice(%/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %/layers.10/Constant_3_output_0, %/layers.10/Constant_4_output_0, %/layers.10/Constant_2_output_0, %/layers.10/Constant_5_output_0)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Slice_output_0, %onnx::Conv_809, %onnx::Conv_810)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.10/Slice_1_output_0 = Slice(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_7_output_0, %/layers.10/Constant_8_output_0, %/layers.10/Constant_6_output_0, %/layers.10/Constant_9_output_0)
%/layers.10/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/Slice_1_output_0, %/layers.10/Constant_10_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/Add_2_output_0, %/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_812, %onnx::Conv_813)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_815, %onnx::Conv_816)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_818, %onnx::Conv_819)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.11/Slice_output_0 = Slice(%/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %/layers.11/Constant_3_output_0, %/layers.11/Constant_4_output_0, %/layers.11/Constant_2_output_0, %/layers.11/Constant_5_output_0)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Slice_output_0, %onnx::Conv_821, %onnx::Conv_822)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.11/Slice_1_output_0 = Slice(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_7_output_0, %/layers.11/Constant_8_output_0, %/layers.11/Constant_6_output_0, %/layers.11/Constant_9_output_0)
%/layers.11/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/Slice_1_output_0, %/layers.11/Constant_10_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/Add_2_output_0, %/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_824, %onnx::Conv_825)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%714 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %714
}
|
val_accuracy
| 91.145831
| 469,758,336
| 1,558,566
|
{'zcp_epe_nas': 153.16627116064276, 'zcp_fisher': 11.853450775146484, 'zcp_flops': 7516133376.0, 'zcp_grad_norm': 73.60101318359375, 'zcp_grasp': -4.955322265625, 'zcp_jacov': -16.04925249374512, 'zcp_l2_norm': 564.9120483398438, 'zcp_nwot': 212.5675222776597, 'zcp_params': 1558566.0, 'zcp_plain': 0.027749836444854, 'zcp_snip': 309.24700927734375, 'zcp_synflow': 100.14922505328846, 'zcp_zen': 60.230648040771484, 'zcp_val_accuracy': 0.9354968070983881}
| |
NASBench101_358719
|
NASBench101
|
358719
|
d8d3194de6632b78970afed982ec1821
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_716[FLOAT, 128x3x3x3]
%onnx::Conv_717[FLOAT, 128]
%onnx::Conv_719[FLOAT, 128x128x1x1]
%onnx::Conv_722[FLOAT, 128x128x1x1]
%onnx::Conv_725[FLOAT, 128x128x1x1]
%onnx::Conv_728[FLOAT, 128x128x1x1]
%onnx::Conv_731[FLOAT, 128x128x3x3]
%onnx::Conv_734[FLOAT, 128x128x1x1]
%onnx::Conv_737[FLOAT, 128x128x1x1]
%onnx::Conv_740[FLOAT, 128x128x1x1]
%onnx::Conv_743[FLOAT, 128x128x1x1]
%onnx::Conv_746[FLOAT, 128x128x3x3]
%onnx::Conv_749[FLOAT, 128x128x1x1]
%onnx::Conv_752[FLOAT, 128x128x1x1]
%onnx::Conv_755[FLOAT, 128x128x1x1]
%onnx::Conv_758[FLOAT, 128x128x1x1]
%onnx::Conv_761[FLOAT, 128x128x3x3]
%onnx::Conv_764[FLOAT, 256x128x1x1]
%onnx::Conv_765[FLOAT, 256]
%onnx::Conv_767[FLOAT, 256x128x1x1]
%onnx::Conv_770[FLOAT, 256x256x1x1]
%onnx::Conv_773[FLOAT, 256x128x1x1]
%onnx::Conv_776[FLOAT, 256x256x3x3]
%onnx::Conv_779[FLOAT, 256x256x1x1]
%onnx::Conv_782[FLOAT, 256x256x1x1]
%onnx::Conv_785[FLOAT, 256x256x1x1]
%onnx::Conv_788[FLOAT, 256x256x1x1]
%onnx::Conv_791[FLOAT, 256x256x3x3]
%onnx::Conv_794[FLOAT, 256x256x1x1]
%onnx::Conv_797[FLOAT, 256x256x1x1]
%onnx::Conv_800[FLOAT, 256x256x1x1]
%onnx::Conv_803[FLOAT, 256x256x1x1]
%onnx::Conv_806[FLOAT, 256x256x3x3]
%onnx::Conv_809[FLOAT, 512x256x1x1]
%onnx::Conv_810[FLOAT, 512]
%onnx::Conv_812[FLOAT, 512x256x1x1]
%onnx::Conv_815[FLOAT, 512x512x1x1]
%onnx::Conv_818[FLOAT, 512x256x1x1]
%onnx::Conv_821[FLOAT, 512x512x3x3]
%onnx::Conv_824[FLOAT, 512x512x1x1]
%onnx::Conv_827[FLOAT, 512x512x1x1]
%onnx::Conv_830[FLOAT, 512x512x1x1]
%onnx::Conv_833[FLOAT, 512x512x1x1]
%onnx::Conv_836[FLOAT, 512x512x3x3]
%onnx::Conv_839[FLOAT, 512x512x1x1]
%onnx::Conv_842[FLOAT, 512x512x1x1]
%onnx::Conv_845[FLOAT, 512x512x1x1]
%onnx::Conv_848[FLOAT, 512x512x1x1]
%onnx::Conv_851[FLOAT, 512x512x3x3]
) {
%onnx::Conv_852 = Identity(%onnx::Conv_810)
%onnx::Conv_849 = Identity(%onnx::Conv_810)
%onnx::Conv_846 = Identity(%onnx::Conv_810)
%onnx::Conv_843 = Identity(%onnx::Conv_810)
%onnx::Conv_840 = Identity(%onnx::Conv_810)
%onnx::Conv_837 = Identity(%onnx::Conv_810)
%onnx::Conv_834 = Identity(%onnx::Conv_810)
%onnx::Conv_831 = Identity(%onnx::Conv_810)
%onnx::Conv_828 = Identity(%onnx::Conv_810)
%onnx::Conv_825 = Identity(%onnx::Conv_810)
%onnx::Conv_822 = Identity(%onnx::Conv_810)
%onnx::Conv_819 = Identity(%onnx::Conv_810)
%onnx::Conv_816 = Identity(%onnx::Conv_810)
%onnx::Conv_813 = Identity(%onnx::Conv_810)
%onnx::Conv_807 = Identity(%onnx::Conv_765)
%onnx::Conv_804 = Identity(%onnx::Conv_765)
%onnx::Conv_801 = Identity(%onnx::Conv_765)
%onnx::Conv_798 = Identity(%onnx::Conv_765)
%onnx::Conv_795 = Identity(%onnx::Conv_765)
%onnx::Conv_792 = Identity(%onnx::Conv_765)
%onnx::Conv_789 = Identity(%onnx::Conv_765)
%onnx::Conv_786 = Identity(%onnx::Conv_765)
%onnx::Conv_783 = Identity(%onnx::Conv_765)
%onnx::Conv_780 = Identity(%onnx::Conv_765)
%onnx::Conv_777 = Identity(%onnx::Conv_765)
%onnx::Conv_774 = Identity(%onnx::Conv_765)
%onnx::Conv_771 = Identity(%onnx::Conv_765)
%onnx::Conv_768 = Identity(%onnx::Conv_765)
%onnx::Conv_762 = Identity(%onnx::Conv_717)
%onnx::Conv_759 = Identity(%onnx::Conv_717)
%onnx::Conv_756 = Identity(%onnx::Conv_717)
%onnx::Conv_753 = Identity(%onnx::Conv_717)
%onnx::Conv_750 = Identity(%onnx::Conv_717)
%onnx::Conv_747 = Identity(%onnx::Conv_717)
%onnx::Conv_744 = Identity(%onnx::Conv_717)
%onnx::Conv_741 = Identity(%onnx::Conv_717)
%onnx::Conv_738 = Identity(%onnx::Conv_717)
%onnx::Conv_735 = Identity(%onnx::Conv_717)
%onnx::Conv_732 = Identity(%onnx::Conv_717)
%onnx::Conv_729 = Identity(%onnx::Conv_717)
%onnx::Conv_726 = Identity(%onnx::Conv_717)
%onnx::Conv_723 = Identity(%onnx::Conv_717)
%onnx::Conv_720 = Identity(%onnx::Conv_717)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_716, %onnx::Conv_717)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_719, %onnx::Conv_720)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_722, %onnx::Conv_723)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_725, %onnx::Conv_726)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_728, %onnx::Conv_729)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/Add_2_output_0, %/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_731, %onnx::Conv_732)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_4_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_734, %onnx::Conv_735)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_737, %onnx::Conv_738)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_740, %onnx::Conv_741)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_743, %onnx::Conv_744)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/Add_2_output_0, %/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_746, %onnx::Conv_747)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_4_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_749, %onnx::Conv_750)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_752, %onnx::Conv_753)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_755, %onnx::Conv_756)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_758, %onnx::Conv_759)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/Add_2_output_0, %/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_761, %onnx::Conv_762)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_4_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_764, %onnx::Conv_765)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_767, %onnx::Conv_768)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_770, %onnx::Conv_771)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_773, %onnx::Conv_774)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/Add_2_output_0, %/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_776, %onnx::Conv_777)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_4_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_779, %onnx::Conv_780)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_782, %onnx::Conv_783)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_785, %onnx::Conv_786)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_788, %onnx::Conv_789)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/Add_2_output_0, %/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_4_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_797, %onnx::Conv_798)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_800, %onnx::Conv_801)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_803, %onnx::Conv_804)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/Add_2_output_0, %/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_806, %onnx::Conv_807)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_4_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_809, %onnx::Conv_810)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_812, %onnx::Conv_813)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_815, %onnx::Conv_816)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_818, %onnx::Conv_819)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/Add_2_output_0, %/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_821, %onnx::Conv_822)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_4_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_824, %onnx::Conv_825)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_827, %onnx::Conv_828)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_830, %onnx::Conv_831)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_833, %onnx::Conv_834)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/Add_2_output_0, %/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_836, %onnx::Conv_837)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_4_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_839, %onnx::Conv_840)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_842, %onnx::Conv_843)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_845, %onnx::Conv_846)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_848, %onnx::Conv_849)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/Add_2_output_0, %/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_851, %onnx::Conv_852)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_4_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/vertex_op.4/maxpool/MaxPool_output_0)
%714 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %714
}
|
val_accuracy
| 91.266024
| 3,860,867,072
| 12,962,698
|
{'zcp_epe_nas': 113.88763874455539, 'zcp_fisher': 23.803123474121094, 'zcp_flops': 61773873152.0, 'zcp_grad_norm': 85.22228240966797, 'zcp_grasp': -15.0302734375, 'zcp_jacov': -16.049400092289655, 'zcp_l2_norm': 1015.0928955078125, 'zcp_nwot': 231.95676340258643, 'zcp_params': 12962698.0, 'zcp_plain': 0.11340443044900801, 'zcp_snip': 718.7100219726562, 'zcp_synflow': 100.3330497582655, 'zcp_zen': 101.64884948730469, 'zcp_val_accuracy': 0.917668282985687}
| |
NASBench101_304095
|
NASBench101
|
304095
|
b7fb87d799c4398a738c7b320b16b68a
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_857[FLOAT, 128x3x3x3]
%onnx::Conv_858[FLOAT, 128]
%onnx::Conv_860[FLOAT, 43x128x1x1]
%onnx::Conv_861[FLOAT, 43]
%onnx::Conv_863[FLOAT, 43x43x1x1]
%onnx::Conv_866[FLOAT, 43x43x3x3]
%onnx::Conv_869[FLOAT, 42x42x1x1]
%onnx::Conv_870[FLOAT, 42]
%onnx::Conv_872[FLOAT, 42x128x1x1]
%onnx::Conv_875[FLOAT, 43x128x1x1]
%onnx::Conv_878[FLOAT, 43x43x1x1]
%onnx::Conv_881[FLOAT, 43x43x3x3]
%onnx::Conv_884[FLOAT, 42x42x1x1]
%onnx::Conv_887[FLOAT, 42x128x1x1]
%onnx::Conv_890[FLOAT, 43x128x1x1]
%onnx::Conv_893[FLOAT, 43x43x1x1]
%onnx::Conv_896[FLOAT, 43x43x3x3]
%onnx::Conv_899[FLOAT, 42x42x1x1]
%onnx::Conv_902[FLOAT, 42x128x1x1]
%onnx::Conv_905[FLOAT, 86x128x1x1]
%onnx::Conv_906[FLOAT, 86]
%onnx::Conv_908[FLOAT, 86x86x1x1]
%onnx::Conv_911[FLOAT, 85x85x3x3]
%onnx::Conv_912[FLOAT, 85]
%onnx::Conv_914[FLOAT, 85x85x1x1]
%onnx::Conv_917[FLOAT, 85x128x1x1]
%onnx::Conv_920[FLOAT, 86x256x1x1]
%onnx::Conv_923[FLOAT, 86x86x1x1]
%onnx::Conv_926[FLOAT, 85x85x3x3]
%onnx::Conv_929[FLOAT, 85x85x1x1]
%onnx::Conv_932[FLOAT, 85x256x1x1]
%onnx::Conv_935[FLOAT, 86x256x1x1]
%onnx::Conv_938[FLOAT, 86x86x1x1]
%onnx::Conv_941[FLOAT, 85x85x3x3]
%onnx::Conv_944[FLOAT, 85x85x1x1]
%onnx::Conv_947[FLOAT, 85x256x1x1]
%onnx::Conv_950[FLOAT, 171x256x1x1]
%onnx::Conv_951[FLOAT, 171]
%onnx::Conv_953[FLOAT, 171x171x1x1]
%onnx::Conv_956[FLOAT, 171x171x3x3]
%onnx::Conv_959[FLOAT, 170x170x1x1]
%onnx::Conv_960[FLOAT, 170]
%onnx::Conv_962[FLOAT, 170x256x1x1]
%onnx::Conv_965[FLOAT, 171x512x1x1]
%onnx::Conv_968[FLOAT, 171x171x1x1]
%onnx::Conv_971[FLOAT, 171x171x3x3]
%onnx::Conv_974[FLOAT, 170x170x1x1]
%onnx::Conv_977[FLOAT, 170x512x1x1]
%onnx::Conv_980[FLOAT, 171x512x1x1]
%onnx::Conv_983[FLOAT, 171x171x1x1]
%onnx::Conv_986[FLOAT, 171x171x3x3]
%onnx::Conv_989[FLOAT, 170x170x1x1]
%onnx::Conv_992[FLOAT, 170x512x1x1]
) {
%onnx::Conv_993 = Identity(%onnx::Conv_960)
%onnx::Conv_990 = Identity(%onnx::Conv_960)
%onnx::Conv_987 = Identity(%onnx::Conv_951)
%onnx::Conv_984 = Identity(%onnx::Conv_951)
%onnx::Conv_981 = Identity(%onnx::Conv_951)
%onnx::Conv_978 = Identity(%onnx::Conv_960)
%onnx::Conv_975 = Identity(%onnx::Conv_960)
%onnx::Conv_972 = Identity(%onnx::Conv_951)
%onnx::Conv_969 = Identity(%onnx::Conv_951)
%onnx::Conv_966 = Identity(%onnx::Conv_951)
%onnx::Conv_963 = Identity(%onnx::Conv_960)
%onnx::Conv_957 = Identity(%onnx::Conv_951)
%onnx::Conv_954 = Identity(%onnx::Conv_951)
%onnx::Conv_948 = Identity(%onnx::Conv_912)
%onnx::Conv_945 = Identity(%onnx::Conv_912)
%onnx::Conv_942 = Identity(%onnx::Conv_912)
%onnx::Conv_939 = Identity(%onnx::Conv_906)
%onnx::Conv_936 = Identity(%onnx::Conv_906)
%onnx::Conv_933 = Identity(%onnx::Conv_912)
%onnx::Conv_930 = Identity(%onnx::Conv_912)
%onnx::Conv_927 = Identity(%onnx::Conv_912)
%onnx::Conv_924 = Identity(%onnx::Conv_906)
%onnx::Conv_921 = Identity(%onnx::Conv_906)
%onnx::Conv_918 = Identity(%onnx::Conv_912)
%onnx::Conv_915 = Identity(%onnx::Conv_912)
%onnx::Conv_909 = Identity(%onnx::Conv_906)
%onnx::Conv_903 = Identity(%onnx::Conv_870)
%onnx::Conv_900 = Identity(%onnx::Conv_870)
%onnx::Conv_897 = Identity(%onnx::Conv_861)
%onnx::Conv_894 = Identity(%onnx::Conv_861)
%onnx::Conv_891 = Identity(%onnx::Conv_861)
%onnx::Conv_888 = Identity(%onnx::Conv_870)
%onnx::Conv_885 = Identity(%onnx::Conv_870)
%onnx::Conv_882 = Identity(%onnx::Conv_861)
%onnx::Conv_879 = Identity(%onnx::Conv_861)
%onnx::Conv_876 = Identity(%onnx::Conv_861)
%onnx::Conv_873 = Identity(%onnx::Conv_870)
%onnx::Conv_867 = Identity(%onnx::Conv_861)
%onnx::Conv_864 = Identity(%onnx::Conv_861)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_857, %onnx::Conv_858)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.1/Slice_output_0 = Slice(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0, %/layers.1/Constant_4_output_0, %/layers.1/Constant_2_output_0, %/layers.1/Constant_5_output_0)
%/layers.1/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/Slice_output_0, %/layers.1/Constant_6_output_0)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_10_output_0 = Constant[value = <Tensor>]()
%/layers.1/Slice_1_output_0 = Slice(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_8_output_0, %/layers.1/Constant_9_output_0, %/layers.1/Constant_7_output_0, %/layers.1/Constant_10_output_0)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_11_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/Slice_1_output_0, %/layers.1/Constant_11_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/Add_3_output_0, %/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/Add_5_output_0 = Add(%/layers.1/Add_4_output_0, %/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_5_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.2/Slice_output_0 = Slice(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0, %/layers.2/Constant_4_output_0, %/layers.2/Constant_2_output_0, %/layers.2/Constant_5_output_0)
%/layers.2/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/Slice_output_0, %/layers.2/Constant_6_output_0)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_10_output_0 = Constant[value = <Tensor>]()
%/layers.2/Slice_1_output_0 = Slice(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_8_output_0, %/layers.2/Constant_9_output_0, %/layers.2/Constant_7_output_0, %/layers.2/Constant_10_output_0)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_11_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/Slice_1_output_0, %/layers.2/Constant_11_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/Add_3_output_0, %/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/Add_5_output_0 = Add(%/layers.2/Add_4_output_0, %/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_5_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.3/Slice_output_0 = Slice(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0, %/layers.3/Constant_4_output_0, %/layers.3/Constant_2_output_0, %/layers.3/Constant_5_output_0)
%/layers.3/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/Slice_output_0, %/layers.3/Constant_6_output_0)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_10_output_0 = Constant[value = <Tensor>]()
%/layers.3/Slice_1_output_0 = Slice(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_8_output_0, %/layers.3/Constant_9_output_0, %/layers.3/Constant_7_output_0, %/layers.3/Constant_10_output_0)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_11_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/Slice_1_output_0, %/layers.3/Constant_11_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/Add_3_output_0, %/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/Add_5_output_0 = Add(%/layers.3/Add_4_output_0, %/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_5_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_908, %onnx::Conv_909)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.5/Slice_output_0 = Slice(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0, %/layers.5/Constant_3_output_0, %/layers.5/Constant_1_output_0, %/layers.5/Constant_4_output_0)
%/layers.5/Constant_5_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/Slice_output_0, %/layers.5/Constant_5_output_0)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_911, %onnx::Conv_912)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.5/Slice_1_output_0 = Slice(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_7_output_0, %/layers.5/Constant_8_output_0, %/layers.5/Constant_6_output_0, %/layers.5/Constant_9_output_0)
%/layers.5/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/Slice_1_output_0, %/layers.5/Constant_10_output_0)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_914, %onnx::Conv_915)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_11_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_12_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_13_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_14_output_0 = Constant[value = <Tensor>]()
%/layers.5/Slice_2_output_0 = Slice(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_12_output_0, %/layers.5/Constant_13_output_0, %/layers.5/Constant_11_output_0, %/layers.5/Constant_14_output_0)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_917, %onnx::Conv_918)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_15_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/Slice_2_output_0, %/layers.5/Constant_15_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/Add_3_output_0, %/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/Add_5_output_0 = Add(%/layers.5/Add_4_output_0, %/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_5_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_920, %onnx::Conv_921)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_923, %onnx::Conv_924)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.6/Slice_output_0 = Slice(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0, %/layers.6/Constant_3_output_0, %/layers.6/Constant_1_output_0, %/layers.6/Constant_4_output_0)
%/layers.6/Constant_5_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/Slice_output_0, %/layers.6/Constant_5_output_0)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_926, %onnx::Conv_927)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.6/Slice_1_output_0 = Slice(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_7_output_0, %/layers.6/Constant_8_output_0, %/layers.6/Constant_6_output_0, %/layers.6/Constant_9_output_0)
%/layers.6/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/Slice_1_output_0, %/layers.6/Constant_10_output_0)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_929, %onnx::Conv_930)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_11_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_12_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_13_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_14_output_0 = Constant[value = <Tensor>]()
%/layers.6/Slice_2_output_0 = Slice(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_12_output_0, %/layers.6/Constant_13_output_0, %/layers.6/Constant_11_output_0, %/layers.6/Constant_14_output_0)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_932, %onnx::Conv_933)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_15_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/Slice_2_output_0, %/layers.6/Constant_15_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/Add_3_output_0, %/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/Add_5_output_0 = Add(%/layers.6/Add_4_output_0, %/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_5_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_935, %onnx::Conv_936)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_938, %onnx::Conv_939)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.7/Slice_output_0 = Slice(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0, %/layers.7/Constant_3_output_0, %/layers.7/Constant_1_output_0, %/layers.7/Constant_4_output_0)
%/layers.7/Constant_5_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/Slice_output_0, %/layers.7/Constant_5_output_0)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_941, %onnx::Conv_942)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.7/Slice_1_output_0 = Slice(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_7_output_0, %/layers.7/Constant_8_output_0, %/layers.7/Constant_6_output_0, %/layers.7/Constant_9_output_0)
%/layers.7/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/Slice_1_output_0, %/layers.7/Constant_10_output_0)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_944, %onnx::Conv_945)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_11_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_12_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_13_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_14_output_0 = Constant[value = <Tensor>]()
%/layers.7/Slice_2_output_0 = Slice(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_12_output_0, %/layers.7/Constant_13_output_0, %/layers.7/Constant_11_output_0, %/layers.7/Constant_14_output_0)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_947, %onnx::Conv_948)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_15_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/Slice_2_output_0, %/layers.7/Constant_15_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/Add_3_output_0, %/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/Add_5_output_0 = Add(%/layers.7/Add_4_output_0, %/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_5_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_950, %onnx::Conv_951)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_953, %onnx::Conv_954)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_956, %onnx::Conv_957)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.9/Slice_output_0 = Slice(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0, %/layers.9/Constant_4_output_0, %/layers.9/Constant_2_output_0, %/layers.9/Constant_5_output_0)
%/layers.9/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/Slice_output_0, %/layers.9/Constant_6_output_0)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_959, %onnx::Conv_960)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_10_output_0 = Constant[value = <Tensor>]()
%/layers.9/Slice_1_output_0 = Slice(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_8_output_0, %/layers.9/Constant_9_output_0, %/layers.9/Constant_7_output_0, %/layers.9/Constant_10_output_0)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_962, %onnx::Conv_963)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_11_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/Slice_1_output_0, %/layers.9/Constant_11_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/Add_3_output_0, %/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/Add_5_output_0 = Add(%/layers.9/Add_4_output_0, %/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_5_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_965, %onnx::Conv_966)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_968, %onnx::Conv_969)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.10/Slice_output_0 = Slice(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0, %/layers.10/Constant_4_output_0, %/layers.10/Constant_2_output_0, %/layers.10/Constant_5_output_0)
%/layers.10/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/Slice_output_0, %/layers.10/Constant_6_output_0)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_974, %onnx::Conv_975)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_10_output_0 = Constant[value = <Tensor>]()
%/layers.10/Slice_1_output_0 = Slice(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_8_output_0, %/layers.10/Constant_9_output_0, %/layers.10/Constant_7_output_0, %/layers.10/Constant_10_output_0)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_977, %onnx::Conv_978)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_11_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/Slice_1_output_0, %/layers.10/Constant_11_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/Add_3_output_0, %/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/Add_5_output_0 = Add(%/layers.10/Add_4_output_0, %/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_5_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.11/Slice_output_0 = Slice(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0, %/layers.11/Constant_4_output_0, %/layers.11/Constant_2_output_0, %/layers.11/Constant_5_output_0)
%/layers.11/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/Slice_output_0, %/layers.11/Constant_6_output_0)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_10_output_0 = Constant[value = <Tensor>]()
%/layers.11/Slice_1_output_0 = Slice(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_8_output_0, %/layers.11/Constant_9_output_0, %/layers.11/Constant_7_output_0, %/layers.11/Constant_10_output_0)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_11_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/Slice_1_output_0, %/layers.11/Constant_11_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/Add_3_output_0, %/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/Add_5_output_0 = Add(%/layers.11/Add_4_output_0, %/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_5_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.4/maxpool/MaxPool_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%855 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %855
}
|
val_accuracy
| 90.404648
| 566,214,400
| 1,859,990
|
{'zcp_epe_nas': 85.11412651909829, 'zcp_fisher': 10.688121795654297, 'zcp_flops': 9059430400.0, 'zcp_grad_norm': 71.77729034423828, 'zcp_grasp': -17.35430908203125, 'zcp_jacov': -16.042863804071494, 'zcp_l2_norm': 762.4550170898438, 'zcp_nwot': 215.78754486332224, 'zcp_params': 1859990.0, 'zcp_plain': -0.004487576428800001, 'zcp_snip': 334.3464050292969, 'zcp_synflow': 81.53368696753006, 'zcp_zen': 71.47158813476562, 'zcp_val_accuracy': 0.887319684028625}
| |
NASBench101_210438
|
NASBench101
|
210438
|
7f6db4854c616ebd85d8ae01a8383e34
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_617[FLOAT, 128x3x3x3]
%onnx::Conv_618[FLOAT, 128]
%onnx::Conv_620[FLOAT, 128x128x1x1]
%onnx::Conv_623[FLOAT, 128x128x1x1]
%onnx::Conv_626[FLOAT, 128x128x1x1]
%onnx::Conv_629[FLOAT, 128x128x1x1]
%onnx::Conv_632[FLOAT, 128x128x1x1]
%onnx::Conv_635[FLOAT, 128x128x1x1]
%onnx::Conv_638[FLOAT, 128x128x1x1]
%onnx::Conv_641[FLOAT, 128x128x1x1]
%onnx::Conv_644[FLOAT, 128x128x1x1]
%onnx::Conv_647[FLOAT, 128x128x1x1]
%onnx::Conv_650[FLOAT, 128x128x1x1]
%onnx::Conv_653[FLOAT, 128x128x1x1]
%onnx::Conv_656[FLOAT, 256x128x1x1]
%onnx::Conv_657[FLOAT, 256]
%onnx::Conv_659[FLOAT, 256x128x1x1]
%onnx::Conv_662[FLOAT, 256x128x1x1]
%onnx::Conv_665[FLOAT, 256x256x1x1]
%onnx::Conv_668[FLOAT, 256x256x1x1]
%onnx::Conv_671[FLOAT, 256x256x1x1]
%onnx::Conv_674[FLOAT, 256x256x1x1]
%onnx::Conv_677[FLOAT, 256x256x1x1]
%onnx::Conv_680[FLOAT, 256x256x1x1]
%onnx::Conv_683[FLOAT, 256x256x1x1]
%onnx::Conv_686[FLOAT, 256x256x1x1]
%onnx::Conv_689[FLOAT, 256x256x1x1]
%onnx::Conv_692[FLOAT, 512x256x1x1]
%onnx::Conv_693[FLOAT, 512]
%onnx::Conv_695[FLOAT, 512x256x1x1]
%onnx::Conv_698[FLOAT, 512x256x1x1]
%onnx::Conv_701[FLOAT, 512x512x1x1]
%onnx::Conv_704[FLOAT, 512x512x1x1]
%onnx::Conv_707[FLOAT, 512x512x1x1]
%onnx::Conv_710[FLOAT, 512x512x1x1]
%onnx::Conv_713[FLOAT, 512x512x1x1]
%onnx::Conv_716[FLOAT, 512x512x1x1]
%onnx::Conv_719[FLOAT, 512x512x1x1]
%onnx::Conv_722[FLOAT, 512x512x1x1]
%onnx::Conv_725[FLOAT, 512x512x1x1]
) {
%onnx::Conv_726 = Identity(%onnx::Conv_693)
%onnx::Conv_723 = Identity(%onnx::Conv_693)
%onnx::Conv_720 = Identity(%onnx::Conv_693)
%onnx::Conv_717 = Identity(%onnx::Conv_693)
%onnx::Conv_714 = Identity(%onnx::Conv_693)
%onnx::Conv_711 = Identity(%onnx::Conv_693)
%onnx::Conv_708 = Identity(%onnx::Conv_693)
%onnx::Conv_705 = Identity(%onnx::Conv_693)
%onnx::Conv_702 = Identity(%onnx::Conv_693)
%onnx::Conv_699 = Identity(%onnx::Conv_693)
%onnx::Conv_696 = Identity(%onnx::Conv_693)
%onnx::Conv_690 = Identity(%onnx::Conv_657)
%onnx::Conv_687 = Identity(%onnx::Conv_657)
%onnx::Conv_684 = Identity(%onnx::Conv_657)
%onnx::Conv_681 = Identity(%onnx::Conv_657)
%onnx::Conv_678 = Identity(%onnx::Conv_657)
%onnx::Conv_675 = Identity(%onnx::Conv_657)
%onnx::Conv_672 = Identity(%onnx::Conv_657)
%onnx::Conv_669 = Identity(%onnx::Conv_657)
%onnx::Conv_666 = Identity(%onnx::Conv_657)
%onnx::Conv_663 = Identity(%onnx::Conv_657)
%onnx::Conv_660 = Identity(%onnx::Conv_657)
%onnx::Conv_654 = Identity(%onnx::Conv_618)
%onnx::Conv_651 = Identity(%onnx::Conv_618)
%onnx::Conv_648 = Identity(%onnx::Conv_618)
%onnx::Conv_645 = Identity(%onnx::Conv_618)
%onnx::Conv_642 = Identity(%onnx::Conv_618)
%onnx::Conv_639 = Identity(%onnx::Conv_618)
%onnx::Conv_636 = Identity(%onnx::Conv_618)
%onnx::Conv_633 = Identity(%onnx::Conv_618)
%onnx::Conv_630 = Identity(%onnx::Conv_618)
%onnx::Conv_627 = Identity(%onnx::Conv_618)
%onnx::Conv_624 = Identity(%onnx::Conv_618)
%onnx::Conv_621 = Identity(%onnx::Conv_618)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_617, %onnx::Conv_618)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_620, %onnx::Conv_621)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_623, %onnx::Conv_624)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_626, %onnx::Conv_627)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/Add_2_output_0, %/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_629, %onnx::Conv_630)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_632, %onnx::Conv_633)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_635, %onnx::Conv_636)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_638, %onnx::Conv_639)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/Add_2_output_0, %/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_641, %onnx::Conv_642)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_644, %onnx::Conv_645)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_647, %onnx::Conv_648)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_650, %onnx::Conv_651)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/Add_2_output_0, %/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_653, %onnx::Conv_654)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_656, %onnx::Conv_657)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_659, %onnx::Conv_660)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_662, %onnx::Conv_663)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/Add_2_output_0, %/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_665, %onnx::Conv_666)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_668, %onnx::Conv_669)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_671, %onnx::Conv_672)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_674, %onnx::Conv_675)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/Add_2_output_0, %/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_677, %onnx::Conv_678)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_680, %onnx::Conv_681)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_683, %onnx::Conv_684)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_686, %onnx::Conv_687)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/Add_2_output_0, %/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_689, %onnx::Conv_690)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_692, %onnx::Conv_693)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_695, %onnx::Conv_696)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_698, %onnx::Conv_699)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/Add_2_output_0, %/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_701, %onnx::Conv_702)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_704, %onnx::Conv_705)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_707, %onnx::Conv_708)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_710, %onnx::Conv_711)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/Add_2_output_0, %/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_713, %onnx::Conv_714)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_716, %onnx::Conv_717)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_719, %onnx::Conv_720)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_722, %onnx::Conv_723)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/Add_2_output_0, %/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_725, %onnx::Conv_726)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%615 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %615
}
|
val_accuracy
| 88.731968
| 1,137,453,056
| 3,667,594
|
{'zcp_epe_nas': 80.515078945473, 'zcp_fisher': 9.79069995880127, 'zcp_flops': 18199248896.0, 'zcp_grad_norm': 59.501739501953125, 'zcp_grasp': -0.79486083984375, 'zcp_jacov': -16.067015819968923, 'zcp_l2_norm': 803.0823974609375, 'zcp_nwot': 228.91580327235374, 'zcp_params': 3667594.0, 'zcp_plain': -0.009259011596441, 'zcp_snip': 447.9893798828125, 'zcp_synflow': 69.77570669425428, 'zcp_zen': 66.5781478881836, 'zcp_val_accuracy': 0.9376001358032221}
| |
NASBench101_218853
|
NASBench101
|
218853
|
8496fbcd4da754f82192fb1586214b6d
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_536[FLOAT, 128x3x3x3]
%onnx::Conv_537[FLOAT, 128]
%onnx::Conv_539[FLOAT, 64x128x1x1]
%onnx::Conv_540[FLOAT, 64]
%onnx::Conv_542[FLOAT, 64x128x1x1]
%onnx::Conv_545[FLOAT, 64x64x1x1]
%onnx::Conv_548[FLOAT, 64x128x1x1]
%onnx::Conv_551[FLOAT, 64x128x1x1]
%onnx::Conv_554[FLOAT, 64x64x1x1]
%onnx::Conv_557[FLOAT, 64x128x1x1]
%onnx::Conv_560[FLOAT, 64x128x1x1]
%onnx::Conv_563[FLOAT, 64x64x1x1]
%onnx::Conv_566[FLOAT, 128x128x1x1]
%onnx::Conv_569[FLOAT, 128x128x1x1]
%onnx::Conv_572[FLOAT, 128x128x1x1]
%onnx::Conv_575[FLOAT, 128x256x1x1]
%onnx::Conv_578[FLOAT, 128x256x1x1]
%onnx::Conv_581[FLOAT, 128x128x1x1]
%onnx::Conv_584[FLOAT, 128x256x1x1]
%onnx::Conv_587[FLOAT, 128x256x1x1]
%onnx::Conv_590[FLOAT, 128x128x1x1]
%onnx::Conv_593[FLOAT, 256x256x1x1]
%onnx::Conv_594[FLOAT, 256]
%onnx::Conv_596[FLOAT, 256x256x1x1]
%onnx::Conv_599[FLOAT, 256x256x1x1]
%onnx::Conv_602[FLOAT, 256x512x1x1]
%onnx::Conv_605[FLOAT, 256x512x1x1]
%onnx::Conv_608[FLOAT, 256x256x1x1]
%onnx::Conv_611[FLOAT, 256x512x1x1]
%onnx::Conv_614[FLOAT, 256x512x1x1]
%onnx::Conv_617[FLOAT, 256x256x1x1]
) {
%onnx::Conv_618 = Identity(%onnx::Conv_594)
%onnx::Conv_615 = Identity(%onnx::Conv_594)
%onnx::Conv_612 = Identity(%onnx::Conv_594)
%onnx::Conv_609 = Identity(%onnx::Conv_594)
%onnx::Conv_606 = Identity(%onnx::Conv_594)
%onnx::Conv_603 = Identity(%onnx::Conv_594)
%onnx::Conv_600 = Identity(%onnx::Conv_594)
%onnx::Conv_597 = Identity(%onnx::Conv_594)
%onnx::Conv_591 = Identity(%onnx::Conv_537)
%onnx::Conv_588 = Identity(%onnx::Conv_537)
%onnx::Conv_585 = Identity(%onnx::Conv_537)
%onnx::Conv_582 = Identity(%onnx::Conv_537)
%onnx::Conv_579 = Identity(%onnx::Conv_537)
%onnx::Conv_576 = Identity(%onnx::Conv_537)
%onnx::Conv_573 = Identity(%onnx::Conv_537)
%onnx::Conv_570 = Identity(%onnx::Conv_537)
%onnx::Conv_567 = Identity(%onnx::Conv_537)
%onnx::Conv_564 = Identity(%onnx::Conv_540)
%onnx::Conv_561 = Identity(%onnx::Conv_540)
%onnx::Conv_558 = Identity(%onnx::Conv_540)
%onnx::Conv_555 = Identity(%onnx::Conv_540)
%onnx::Conv_552 = Identity(%onnx::Conv_540)
%onnx::Conv_549 = Identity(%onnx::Conv_540)
%onnx::Conv_546 = Identity(%onnx::Conv_540)
%onnx::Conv_543 = Identity(%onnx::Conv_540)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_536, %onnx::Conv_537)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_539, %onnx::Conv_540)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_542, %onnx::Conv_543)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_545, %onnx::Conv_546)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_548, %onnx::Conv_549)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_551, %onnx::Conv_552)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_554, %onnx::Conv_555)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_557, %onnx::Conv_558)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_560, %onnx::Conv_561)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_563, %onnx::Conv_564)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_566, %onnx::Conv_567)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_569, %onnx::Conv_570)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_572, %onnx::Conv_573)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_575, %onnx::Conv_576)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_578, %onnx::Conv_579)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_581, %onnx::Conv_582)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_584, %onnx::Conv_585)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_587, %onnx::Conv_588)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_590, %onnx::Conv_591)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_593, %onnx::Conv_594)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_596, %onnx::Conv_597)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_599, %onnx::Conv_600)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_602, %onnx::Conv_603)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_605, %onnx::Conv_606)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_608, %onnx::Conv_609)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_611, %onnx::Conv_612)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_614, %onnx::Conv_615)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_617, %onnx::Conv_618)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/vertex_op.4/maxpool/MaxPool_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%534 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %534
}
|
val_accuracy
| 89.292866
| 360,327,168
| 1,143,306
|
{'zcp_epe_nas': 92.9477651950066, 'zcp_fisher': 2.257873773574829, 'zcp_flops': 5765234688.0, 'zcp_grad_norm': 28.419084548950195, 'zcp_grasp': 1.008651733398437, 'zcp_jacov': -16.065472272422916, 'zcp_l2_norm': 545.1327514648438, 'zcp_nwot': 214.54012136169524, 'zcp_params': 1143306.0, 'zcp_plain': 0.000659571203868, 'zcp_snip': 167.5330352783203, 'zcp_synflow': 51.71495558792884, 'zcp_zen': 46.359458923339844, 'zcp_val_accuracy': 0.8762019276618951}
| |
NASBench101_363049
|
NASBench101
|
363049
|
db70ef69687912d12cd943a38a1353df
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_761[FLOAT, 128x3x3x3]
%onnx::Conv_762[FLOAT, 128]
%onnx::Conv_764[FLOAT, 64x128x1x1]
%onnx::Conv_765[FLOAT, 64]
%onnx::Conv_767[FLOAT, 64x64x1x1]
%onnx::Conv_770[FLOAT, 64x128x1x1]
%onnx::Conv_773[FLOAT, 64x64x3x3]
%onnx::Conv_776[FLOAT, 64x64x3x3]
%onnx::Conv_779[FLOAT, 64x128x1x1]
%onnx::Conv_782[FLOAT, 64x64x1x1]
%onnx::Conv_785[FLOAT, 64x128x1x1]
%onnx::Conv_788[FLOAT, 64x64x3x3]
%onnx::Conv_791[FLOAT, 64x64x3x3]
%onnx::Conv_794[FLOAT, 64x128x1x1]
%onnx::Conv_797[FLOAT, 64x64x1x1]
%onnx::Conv_800[FLOAT, 64x128x1x1]
%onnx::Conv_803[FLOAT, 64x64x3x3]
%onnx::Conv_806[FLOAT, 64x64x3x3]
%onnx::Conv_809[FLOAT, 128x128x1x1]
%onnx::Conv_812[FLOAT, 128x128x1x1]
%onnx::Conv_815[FLOAT, 128x128x1x1]
%onnx::Conv_818[FLOAT, 128x128x3x3]
%onnx::Conv_821[FLOAT, 128x128x3x3]
%onnx::Conv_824[FLOAT, 128x256x1x1]
%onnx::Conv_827[FLOAT, 128x128x1x1]
%onnx::Conv_830[FLOAT, 128x256x1x1]
%onnx::Conv_833[FLOAT, 128x128x3x3]
%onnx::Conv_836[FLOAT, 128x128x3x3]
%onnx::Conv_839[FLOAT, 128x256x1x1]
%onnx::Conv_842[FLOAT, 128x128x1x1]
%onnx::Conv_845[FLOAT, 128x256x1x1]
%onnx::Conv_848[FLOAT, 128x128x3x3]
%onnx::Conv_851[FLOAT, 128x128x3x3]
%onnx::Conv_854[FLOAT, 256x256x1x1]
%onnx::Conv_855[FLOAT, 256]
%onnx::Conv_857[FLOAT, 256x256x1x1]
%onnx::Conv_860[FLOAT, 256x256x1x1]
%onnx::Conv_863[FLOAT, 256x256x3x3]
%onnx::Conv_866[FLOAT, 256x256x3x3]
%onnx::Conv_869[FLOAT, 256x512x1x1]
%onnx::Conv_872[FLOAT, 256x256x1x1]
%onnx::Conv_875[FLOAT, 256x512x1x1]
%onnx::Conv_878[FLOAT, 256x256x3x3]
%onnx::Conv_881[FLOAT, 256x256x3x3]
%onnx::Conv_884[FLOAT, 256x512x1x1]
%onnx::Conv_887[FLOAT, 256x256x1x1]
%onnx::Conv_890[FLOAT, 256x512x1x1]
%onnx::Conv_893[FLOAT, 256x256x3x3]
%onnx::Conv_896[FLOAT, 256x256x3x3]
) {
%onnx::Conv_897 = Identity(%onnx::Conv_855)
%onnx::Conv_894 = Identity(%onnx::Conv_855)
%onnx::Conv_891 = Identity(%onnx::Conv_855)
%onnx::Conv_888 = Identity(%onnx::Conv_855)
%onnx::Conv_885 = Identity(%onnx::Conv_855)
%onnx::Conv_882 = Identity(%onnx::Conv_855)
%onnx::Conv_879 = Identity(%onnx::Conv_855)
%onnx::Conv_876 = Identity(%onnx::Conv_855)
%onnx::Conv_873 = Identity(%onnx::Conv_855)
%onnx::Conv_870 = Identity(%onnx::Conv_855)
%onnx::Conv_867 = Identity(%onnx::Conv_855)
%onnx::Conv_864 = Identity(%onnx::Conv_855)
%onnx::Conv_861 = Identity(%onnx::Conv_855)
%onnx::Conv_858 = Identity(%onnx::Conv_855)
%onnx::Conv_852 = Identity(%onnx::Conv_762)
%onnx::Conv_849 = Identity(%onnx::Conv_762)
%onnx::Conv_846 = Identity(%onnx::Conv_762)
%onnx::Conv_843 = Identity(%onnx::Conv_762)
%onnx::Conv_840 = Identity(%onnx::Conv_762)
%onnx::Conv_837 = Identity(%onnx::Conv_762)
%onnx::Conv_834 = Identity(%onnx::Conv_762)
%onnx::Conv_831 = Identity(%onnx::Conv_762)
%onnx::Conv_828 = Identity(%onnx::Conv_762)
%onnx::Conv_825 = Identity(%onnx::Conv_762)
%onnx::Conv_822 = Identity(%onnx::Conv_762)
%onnx::Conv_819 = Identity(%onnx::Conv_762)
%onnx::Conv_816 = Identity(%onnx::Conv_762)
%onnx::Conv_813 = Identity(%onnx::Conv_762)
%onnx::Conv_810 = Identity(%onnx::Conv_762)
%onnx::Conv_807 = Identity(%onnx::Conv_765)
%onnx::Conv_804 = Identity(%onnx::Conv_765)
%onnx::Conv_801 = Identity(%onnx::Conv_765)
%onnx::Conv_798 = Identity(%onnx::Conv_765)
%onnx::Conv_795 = Identity(%onnx::Conv_765)
%onnx::Conv_792 = Identity(%onnx::Conv_765)
%onnx::Conv_789 = Identity(%onnx::Conv_765)
%onnx::Conv_786 = Identity(%onnx::Conv_765)
%onnx::Conv_783 = Identity(%onnx::Conv_765)
%onnx::Conv_780 = Identity(%onnx::Conv_765)
%onnx::Conv_777 = Identity(%onnx::Conv_765)
%onnx::Conv_774 = Identity(%onnx::Conv_765)
%onnx::Conv_771 = Identity(%onnx::Conv_765)
%onnx::Conv_768 = Identity(%onnx::Conv_765)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_761, %onnx::Conv_762)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_764, %onnx::Conv_765)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_767, %onnx::Conv_768)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_770, %onnx::Conv_771)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_773, %onnx::Conv_774)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/Add_3_output_0, %/layers.1/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_776, %onnx::Conv_777)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_779, %onnx::Conv_780)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_782, %onnx::Conv_783)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_785, %onnx::Conv_786)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_788, %onnx::Conv_789)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/Add_3_output_0, %/layers.2/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_797, %onnx::Conv_798)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_800, %onnx::Conv_801)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_803, %onnx::Conv_804)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/Add_3_output_0, %/layers.3/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_4_output_0, %onnx::Conv_806, %onnx::Conv_807)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_809, %onnx::Conv_810)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_812, %onnx::Conv_813)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_815, %onnx::Conv_816)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_818, %onnx::Conv_819)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/Add_3_output_0, %/layers.5/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_821, %onnx::Conv_822)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_824, %onnx::Conv_825)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_827, %onnx::Conv_828)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_830, %onnx::Conv_831)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_833, %onnx::Conv_834)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/Add_3_output_0, %/layers.6/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_836, %onnx::Conv_837)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_839, %onnx::Conv_840)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_842, %onnx::Conv_843)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_845, %onnx::Conv_846)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_848, %onnx::Conv_849)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/Add_3_output_0, %/layers.7/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_4_output_0, %onnx::Conv_851, %onnx::Conv_852)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_854, %onnx::Conv_855)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_857, %onnx::Conv_858)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/Add_3_output_0, %/layers.9/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/Add_3_output_0, %/layers.10/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/Add_3_output_0, %/layers.11/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_4_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%759 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %759
}
|
val_accuracy
| 92.217547
| 1,724,786,688
| 5,793,546
|
{'zcp_epe_nas': 216.46414824746932, 'zcp_fisher': 9.452518463134766, 'zcp_flops': 27596587008.0, 'zcp_grad_norm': 49.81298828125, 'zcp_grasp': -1.748046875, 'zcp_jacov': -16.060193718882815, 'zcp_l2_norm': 844.1419067382812, 'zcp_nwot': 221.4971218911647, 'zcp_params': 5793546.0, 'zcp_plain': 0.019581984728574, 'zcp_snip': 325.9831237792969, 'zcp_synflow': 106.29829676856198, 'zcp_zen': 93.6736831665039, 'zcp_val_accuracy': 0.889322936534881}
| |
NASBench101_270334
|
NASBench101
|
270334
|
a3c1a8b6c5670491758bdf9bc042cec0
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_986[FLOAT, 128x3x3x3]
%onnx::Conv_987[FLOAT, 128]
%onnx::Conv_989[FLOAT, 64x128x1x1]
%onnx::Conv_990[FLOAT, 64]
%onnx::Conv_992[FLOAT, 64x64x1x1]
%onnx::Conv_995[FLOAT, 64x64x1x1]
%onnx::Conv_998[FLOAT, 64x128x1x1]
%onnx::Conv_1001[FLOAT, 64x64x1x1]
%onnx::Conv_1004[FLOAT, 64x64x1x1]
%onnx::Conv_1007[FLOAT, 128x128x1x1]
%onnx::Conv_1010[FLOAT, 64x128x1x1]
%onnx::Conv_1013[FLOAT, 64x64x1x1]
%onnx::Conv_1016[FLOAT, 64x64x1x1]
%onnx::Conv_1019[FLOAT, 64x128x1x1]
%onnx::Conv_1022[FLOAT, 64x64x1x1]
%onnx::Conv_1025[FLOAT, 64x64x1x1]
%onnx::Conv_1028[FLOAT, 128x128x1x1]
%onnx::Conv_1031[FLOAT, 64x128x1x1]
%onnx::Conv_1034[FLOAT, 64x64x1x1]
%onnx::Conv_1037[FLOAT, 64x64x1x1]
%onnx::Conv_1040[FLOAT, 64x128x1x1]
%onnx::Conv_1043[FLOAT, 64x64x1x1]
%onnx::Conv_1046[FLOAT, 64x64x1x1]
%onnx::Conv_1049[FLOAT, 128x128x1x1]
%onnx::Conv_1052[FLOAT, 128x128x1x1]
%onnx::Conv_1055[FLOAT, 128x128x1x1]
%onnx::Conv_1058[FLOAT, 128x128x1x1]
%onnx::Conv_1061[FLOAT, 128x128x1x1]
%onnx::Conv_1064[FLOAT, 128x128x1x1]
%onnx::Conv_1067[FLOAT, 128x128x1x1]
%onnx::Conv_1070[FLOAT, 256x128x1x1]
%onnx::Conv_1071[FLOAT, 256]
%onnx::Conv_1073[FLOAT, 128x256x1x1]
%onnx::Conv_1076[FLOAT, 128x128x1x1]
%onnx::Conv_1079[FLOAT, 128x128x1x1]
%onnx::Conv_1082[FLOAT, 128x256x1x1]
%onnx::Conv_1085[FLOAT, 128x128x1x1]
%onnx::Conv_1088[FLOAT, 128x128x1x1]
%onnx::Conv_1091[FLOAT, 256x256x1x1]
%onnx::Conv_1094[FLOAT, 128x256x1x1]
%onnx::Conv_1097[FLOAT, 128x128x1x1]
%onnx::Conv_1100[FLOAT, 128x128x1x1]
%onnx::Conv_1103[FLOAT, 128x256x1x1]
%onnx::Conv_1106[FLOAT, 128x128x1x1]
%onnx::Conv_1109[FLOAT, 128x128x1x1]
%onnx::Conv_1112[FLOAT, 256x256x1x1]
%onnx::Conv_1115[FLOAT, 256x256x1x1]
%onnx::Conv_1118[FLOAT, 256x256x1x1]
%onnx::Conv_1121[FLOAT, 256x256x1x1]
%onnx::Conv_1124[FLOAT, 256x256x1x1]
%onnx::Conv_1127[FLOAT, 256x256x1x1]
%onnx::Conv_1130[FLOAT, 256x256x1x1]
%onnx::Conv_1133[FLOAT, 512x256x1x1]
%onnx::Conv_1134[FLOAT, 512]
%onnx::Conv_1136[FLOAT, 256x512x1x1]
%onnx::Conv_1139[FLOAT, 256x256x1x1]
%onnx::Conv_1142[FLOAT, 256x256x1x1]
%onnx::Conv_1145[FLOAT, 256x512x1x1]
%onnx::Conv_1148[FLOAT, 256x256x1x1]
%onnx::Conv_1151[FLOAT, 256x256x1x1]
%onnx::Conv_1154[FLOAT, 512x512x1x1]
%onnx::Conv_1157[FLOAT, 256x512x1x1]
%onnx::Conv_1160[FLOAT, 256x256x1x1]
%onnx::Conv_1163[FLOAT, 256x256x1x1]
%onnx::Conv_1166[FLOAT, 256x512x1x1]
%onnx::Conv_1169[FLOAT, 256x256x1x1]
%onnx::Conv_1172[FLOAT, 256x256x1x1]
%onnx::Conv_1175[FLOAT, 512x512x1x1]
) {
%onnx::Conv_1176 = Identity(%onnx::Conv_1134)
%onnx::Conv_1173 = Identity(%onnx::Conv_1071)
%onnx::Conv_1170 = Identity(%onnx::Conv_1071)
%onnx::Conv_1167 = Identity(%onnx::Conv_1071)
%onnx::Conv_1164 = Identity(%onnx::Conv_1071)
%onnx::Conv_1161 = Identity(%onnx::Conv_1071)
%onnx::Conv_1158 = Identity(%onnx::Conv_1071)
%onnx::Conv_1155 = Identity(%onnx::Conv_1134)
%onnx::Conv_1152 = Identity(%onnx::Conv_1071)
%onnx::Conv_1149 = Identity(%onnx::Conv_1071)
%onnx::Conv_1146 = Identity(%onnx::Conv_1071)
%onnx::Conv_1143 = Identity(%onnx::Conv_1071)
%onnx::Conv_1140 = Identity(%onnx::Conv_1071)
%onnx::Conv_1137 = Identity(%onnx::Conv_1071)
%onnx::Conv_1131 = Identity(%onnx::Conv_1071)
%onnx::Conv_1128 = Identity(%onnx::Conv_1071)
%onnx::Conv_1125 = Identity(%onnx::Conv_1071)
%onnx::Conv_1122 = Identity(%onnx::Conv_1071)
%onnx::Conv_1119 = Identity(%onnx::Conv_1071)
%onnx::Conv_1116 = Identity(%onnx::Conv_1071)
%onnx::Conv_1113 = Identity(%onnx::Conv_1071)
%onnx::Conv_1110 = Identity(%onnx::Conv_987)
%onnx::Conv_1107 = Identity(%onnx::Conv_987)
%onnx::Conv_1104 = Identity(%onnx::Conv_987)
%onnx::Conv_1101 = Identity(%onnx::Conv_987)
%onnx::Conv_1098 = Identity(%onnx::Conv_987)
%onnx::Conv_1095 = Identity(%onnx::Conv_987)
%onnx::Conv_1092 = Identity(%onnx::Conv_1071)
%onnx::Conv_1089 = Identity(%onnx::Conv_987)
%onnx::Conv_1086 = Identity(%onnx::Conv_987)
%onnx::Conv_1083 = Identity(%onnx::Conv_987)
%onnx::Conv_1080 = Identity(%onnx::Conv_987)
%onnx::Conv_1077 = Identity(%onnx::Conv_987)
%onnx::Conv_1074 = Identity(%onnx::Conv_987)
%onnx::Conv_1068 = Identity(%onnx::Conv_987)
%onnx::Conv_1065 = Identity(%onnx::Conv_987)
%onnx::Conv_1062 = Identity(%onnx::Conv_987)
%onnx::Conv_1059 = Identity(%onnx::Conv_987)
%onnx::Conv_1056 = Identity(%onnx::Conv_987)
%onnx::Conv_1053 = Identity(%onnx::Conv_987)
%onnx::Conv_1050 = Identity(%onnx::Conv_987)
%onnx::Conv_1047 = Identity(%onnx::Conv_990)
%onnx::Conv_1044 = Identity(%onnx::Conv_990)
%onnx::Conv_1041 = Identity(%onnx::Conv_990)
%onnx::Conv_1038 = Identity(%onnx::Conv_990)
%onnx::Conv_1035 = Identity(%onnx::Conv_990)
%onnx::Conv_1032 = Identity(%onnx::Conv_990)
%onnx::Conv_1029 = Identity(%onnx::Conv_987)
%onnx::Conv_1026 = Identity(%onnx::Conv_990)
%onnx::Conv_1023 = Identity(%onnx::Conv_990)
%onnx::Conv_1020 = Identity(%onnx::Conv_990)
%onnx::Conv_1017 = Identity(%onnx::Conv_990)
%onnx::Conv_1014 = Identity(%onnx::Conv_990)
%onnx::Conv_1011 = Identity(%onnx::Conv_990)
%onnx::Conv_1008 = Identity(%onnx::Conv_987)
%onnx::Conv_1005 = Identity(%onnx::Conv_990)
%onnx::Conv_1002 = Identity(%onnx::Conv_990)
%onnx::Conv_999 = Identity(%onnx::Conv_990)
%onnx::Conv_996 = Identity(%onnx::Conv_990)
%onnx::Conv_993 = Identity(%onnx::Conv_990)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_986, %onnx::Conv_987)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_4_output_0)
%/layers.1/Add_5_output_0 = Add(%/layers.1/Add_4_output_0, %/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_5_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_6_output_0 = Add(%/layers.1/Concat_output_0, %/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_6_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_6_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_4_output_0)
%/layers.2/Add_5_output_0 = Add(%/layers.2/Add_4_output_0, %/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_5_output_0, %onnx::Conv_1025, %onnx::Conv_1026)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_6_output_0, %onnx::Conv_1028, %onnx::Conv_1029)
%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_6_output_0 = Add(%/layers.2/Concat_output_0, %/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_6_output_0, %onnx::Conv_1031, %onnx::Conv_1032)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_1034, %onnx::Conv_1035)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_1037, %onnx::Conv_1038)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_6_output_0, %onnx::Conv_1040, %onnx::Conv_1041)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_1043, %onnx::Conv_1044)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_4_output_0)
%/layers.3/Add_5_output_0 = Add(%/layers.3/Add_4_output_0, %/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_5_output_0, %onnx::Conv_1046, %onnx::Conv_1047)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_6_output_0, %onnx::Conv_1049, %onnx::Conv_1050)
%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_6_output_0 = Add(%/layers.3/Concat_output_0, %/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Add_6_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1052, %onnx::Conv_1053)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_1055, %onnx::Conv_1056)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_1058, %onnx::Conv_1059)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1061, %onnx::Conv_1062)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_1064, %onnx::Conv_1065)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_4_output_0)
%/layers.5/Add_5_output_0 = Add(%/layers.5/Add_4_output_0, %/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_5_output_0, %onnx::Conv_1067, %onnx::Conv_1068)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1070, %onnx::Conv_1071)
%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_6_output_0 = Add(%/layers.5/Concat_output_0, %/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_6_output_0, %onnx::Conv_1073, %onnx::Conv_1074)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_1076, %onnx::Conv_1077)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_1079, %onnx::Conv_1080)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_6_output_0, %onnx::Conv_1082, %onnx::Conv_1083)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_1085, %onnx::Conv_1086)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_4_output_0)
%/layers.6/Add_5_output_0 = Add(%/layers.6/Add_4_output_0, %/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_5_output_0, %onnx::Conv_1088, %onnx::Conv_1089)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_6_output_0, %onnx::Conv_1091, %onnx::Conv_1092)
%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_6_output_0 = Add(%/layers.6/Concat_output_0, %/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_6_output_0, %onnx::Conv_1094, %onnx::Conv_1095)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_1097, %onnx::Conv_1098)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_1100, %onnx::Conv_1101)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_6_output_0, %onnx::Conv_1103, %onnx::Conv_1104)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_1106, %onnx::Conv_1107)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_4_output_0)
%/layers.7/Add_5_output_0 = Add(%/layers.7/Add_4_output_0, %/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_5_output_0, %onnx::Conv_1109, %onnx::Conv_1110)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_6_output_0, %onnx::Conv_1112, %onnx::Conv_1113)
%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_6_output_0 = Add(%/layers.7/Concat_output_0, %/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Add_6_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1115, %onnx::Conv_1116)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_1118, %onnx::Conv_1119)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_1121, %onnx::Conv_1122)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1124, %onnx::Conv_1125)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_1127, %onnx::Conv_1128)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_4_output_0)
%/layers.9/Add_5_output_0 = Add(%/layers.9/Add_4_output_0, %/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_5_output_0, %onnx::Conv_1130, %onnx::Conv_1131)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1133, %onnx::Conv_1134)
%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_6_output_0 = Add(%/layers.9/Concat_output_0, %/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_6_output_0, %onnx::Conv_1136, %onnx::Conv_1137)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_1139, %onnx::Conv_1140)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_1142, %onnx::Conv_1143)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_6_output_0, %onnx::Conv_1145, %onnx::Conv_1146)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_1148, %onnx::Conv_1149)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_4_output_0)
%/layers.10/Add_5_output_0 = Add(%/layers.10/Add_4_output_0, %/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_5_output_0, %onnx::Conv_1151, %onnx::Conv_1152)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_6_output_0, %onnx::Conv_1154, %onnx::Conv_1155)
%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_6_output_0 = Add(%/layers.10/Concat_output_0, %/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_6_output_0, %onnx::Conv_1157, %onnx::Conv_1158)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_1160, %onnx::Conv_1161)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_1163, %onnx::Conv_1164)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_6_output_0, %onnx::Conv_1166, %onnx::Conv_1167)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_1169, %onnx::Conv_1170)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_4_output_0)
%/layers.11/Add_5_output_0 = Add(%/layers.11/Add_4_output_0, %/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_5_output_0, %onnx::Conv_1172, %onnx::Conv_1173)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_6_output_0, %onnx::Conv_1175, %onnx::Conv_1176)
%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_6_output_0 = Add(%/layers.11/Concat_output_0, %/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Add_6_output_0)
%984 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %984
}
|
val_accuracy
| 89.443111
| 869,017,600
| 2,799,242
|
{'zcp_epe_nas': 94.86415505620636, 'zcp_fisher': 3.222362995147705, 'zcp_flops': 13904281600.0, 'zcp_grad_norm': 51.63423156738281, 'zcp_grasp': -8.24700927734375, 'zcp_jacov': -16.060275681026088, 'zcp_l2_norm': 1189.6202392578125, 'zcp_nwot': 229.04579286889032, 'zcp_params': 2799242.0, 'zcp_plain': 0.046233419328927, 'zcp_snip': 312.1652526855469, 'zcp_synflow': 98.27025858902039, 'zcp_zen': 96.77135467529297, 'zcp_val_accuracy': 0.923677861690521}
| |
NASBench101_143791
|
NASBench101
|
143791
|
56fb30e42fa56937b03f81fda6e16b8d
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_755[FLOAT, 128x3x3x3]
%onnx::Conv_756[FLOAT, 128]
%onnx::Conv_758[FLOAT, 43x128x1x1]
%onnx::Conv_759[FLOAT, 43]
%onnx::Conv_761[FLOAT, 43x43x3x3]
%onnx::Conv_764[FLOAT, 43x128x1x1]
%onnx::Conv_767[FLOAT, 43x43x3x3]
%onnx::Conv_770[FLOAT, 42x42x1x1]
%onnx::Conv_771[FLOAT, 42]
%onnx::Conv_773[FLOAT, 43x128x1x1]
%onnx::Conv_776[FLOAT, 43x43x3x3]
%onnx::Conv_779[FLOAT, 43x128x1x1]
%onnx::Conv_782[FLOAT, 43x43x3x3]
%onnx::Conv_785[FLOAT, 42x42x1x1]
%onnx::Conv_788[FLOAT, 43x128x1x1]
%onnx::Conv_791[FLOAT, 43x43x3x3]
%onnx::Conv_794[FLOAT, 43x128x1x1]
%onnx::Conv_797[FLOAT, 43x43x3x3]
%onnx::Conv_800[FLOAT, 42x42x1x1]
%onnx::Conv_803[FLOAT, 86x128x1x1]
%onnx::Conv_804[FLOAT, 86]
%onnx::Conv_806[FLOAT, 86x86x3x3]
%onnx::Conv_809[FLOAT, 85x128x1x1]
%onnx::Conv_810[FLOAT, 85]
%onnx::Conv_812[FLOAT, 85x85x3x3]
%onnx::Conv_815[FLOAT, 85x85x1x1]
%onnx::Conv_818[FLOAT, 86x256x1x1]
%onnx::Conv_821[FLOAT, 86x86x3x3]
%onnx::Conv_824[FLOAT, 85x256x1x1]
%onnx::Conv_827[FLOAT, 85x85x3x3]
%onnx::Conv_830[FLOAT, 85x85x1x1]
%onnx::Conv_833[FLOAT, 86x256x1x1]
%onnx::Conv_836[FLOAT, 86x86x3x3]
%onnx::Conv_839[FLOAT, 85x256x1x1]
%onnx::Conv_842[FLOAT, 85x85x3x3]
%onnx::Conv_845[FLOAT, 85x85x1x1]
%onnx::Conv_848[FLOAT, 171x256x1x1]
%onnx::Conv_849[FLOAT, 171]
%onnx::Conv_851[FLOAT, 171x171x3x3]
%onnx::Conv_854[FLOAT, 171x256x1x1]
%onnx::Conv_857[FLOAT, 171x171x3x3]
%onnx::Conv_860[FLOAT, 170x170x1x1]
%onnx::Conv_861[FLOAT, 170]
%onnx::Conv_863[FLOAT, 171x512x1x1]
%onnx::Conv_866[FLOAT, 171x171x3x3]
%onnx::Conv_869[FLOAT, 171x512x1x1]
%onnx::Conv_872[FLOAT, 171x171x3x3]
%onnx::Conv_875[FLOAT, 170x170x1x1]
%onnx::Conv_878[FLOAT, 171x512x1x1]
%onnx::Conv_881[FLOAT, 171x171x3x3]
%onnx::Conv_884[FLOAT, 171x512x1x1]
%onnx::Conv_887[FLOAT, 171x171x3x3]
%onnx::Conv_890[FLOAT, 170x170x1x1]
) {
%onnx::Conv_891 = Identity(%onnx::Conv_861)
%onnx::Conv_888 = Identity(%onnx::Conv_849)
%onnx::Conv_885 = Identity(%onnx::Conv_849)
%onnx::Conv_882 = Identity(%onnx::Conv_849)
%onnx::Conv_879 = Identity(%onnx::Conv_849)
%onnx::Conv_876 = Identity(%onnx::Conv_861)
%onnx::Conv_873 = Identity(%onnx::Conv_849)
%onnx::Conv_870 = Identity(%onnx::Conv_849)
%onnx::Conv_867 = Identity(%onnx::Conv_849)
%onnx::Conv_864 = Identity(%onnx::Conv_849)
%onnx::Conv_858 = Identity(%onnx::Conv_849)
%onnx::Conv_855 = Identity(%onnx::Conv_849)
%onnx::Conv_852 = Identity(%onnx::Conv_849)
%onnx::Conv_846 = Identity(%onnx::Conv_810)
%onnx::Conv_843 = Identity(%onnx::Conv_810)
%onnx::Conv_840 = Identity(%onnx::Conv_810)
%onnx::Conv_837 = Identity(%onnx::Conv_804)
%onnx::Conv_834 = Identity(%onnx::Conv_804)
%onnx::Conv_831 = Identity(%onnx::Conv_810)
%onnx::Conv_828 = Identity(%onnx::Conv_810)
%onnx::Conv_825 = Identity(%onnx::Conv_810)
%onnx::Conv_822 = Identity(%onnx::Conv_804)
%onnx::Conv_819 = Identity(%onnx::Conv_804)
%onnx::Conv_816 = Identity(%onnx::Conv_810)
%onnx::Conv_813 = Identity(%onnx::Conv_810)
%onnx::Conv_807 = Identity(%onnx::Conv_804)
%onnx::Conv_801 = Identity(%onnx::Conv_771)
%onnx::Conv_798 = Identity(%onnx::Conv_759)
%onnx::Conv_795 = Identity(%onnx::Conv_759)
%onnx::Conv_792 = Identity(%onnx::Conv_759)
%onnx::Conv_789 = Identity(%onnx::Conv_759)
%onnx::Conv_786 = Identity(%onnx::Conv_771)
%onnx::Conv_783 = Identity(%onnx::Conv_759)
%onnx::Conv_780 = Identity(%onnx::Conv_759)
%onnx::Conv_777 = Identity(%onnx::Conv_759)
%onnx::Conv_774 = Identity(%onnx::Conv_759)
%onnx::Conv_768 = Identity(%onnx::Conv_759)
%onnx::Conv_765 = Identity(%onnx::Conv_759)
%onnx::Conv_762 = Identity(%onnx::Conv_759)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_755, %onnx::Conv_756)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_758, %onnx::Conv_759)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_761, %onnx::Conv_762)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_764, %onnx::Conv_765)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_767, %onnx::Conv_768)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.1/Slice_output_0 = Slice(%/layers.1/vertex_op.3/maxpool/MaxPool_output_0, %/layers.1/Constant_3_output_0, %/layers.1/Constant_4_output_0, %/layers.1/Constant_2_output_0, %/layers.1/Constant_5_output_0)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Slice_output_0, %onnx::Conv_770, %onnx::Conv_771)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_773, %onnx::Conv_774)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_776, %onnx::Conv_777)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_779, %onnx::Conv_780)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_782, %onnx::Conv_783)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.2/Slice_output_0 = Slice(%/layers.2/vertex_op.3/maxpool/MaxPool_output_0, %/layers.2/Constant_3_output_0, %/layers.2/Constant_4_output_0, %/layers.2/Constant_2_output_0, %/layers.2/Constant_5_output_0)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Slice_output_0, %onnx::Conv_785, %onnx::Conv_786)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_788, %onnx::Conv_789)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_797, %onnx::Conv_798)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.3/Slice_output_0 = Slice(%/layers.3/vertex_op.3/maxpool/MaxPool_output_0, %/layers.3/Constant_3_output_0, %/layers.3/Constant_4_output_0, %/layers.3/Constant_2_output_0, %/layers.3/Constant_5_output_0)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Slice_output_0, %onnx::Conv_800, %onnx::Conv_801)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_803, %onnx::Conv_804)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_806, %onnx::Conv_807)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_809, %onnx::Conv_810)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_812, %onnx::Conv_813)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_815, %onnx::Conv_816)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_818, %onnx::Conv_819)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_821, %onnx::Conv_822)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_824, %onnx::Conv_825)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_827, %onnx::Conv_828)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_830, %onnx::Conv_831)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_833, %onnx::Conv_834)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_836, %onnx::Conv_837)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_839, %onnx::Conv_840)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_842, %onnx::Conv_843)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_845, %onnx::Conv_846)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_848, %onnx::Conv_849)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_851, %onnx::Conv_852)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_854, %onnx::Conv_855)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_857, %onnx::Conv_858)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.9/Slice_output_0 = Slice(%/layers.9/vertex_op.3/maxpool/MaxPool_output_0, %/layers.9/Constant_3_output_0, %/layers.9/Constant_4_output_0, %/layers.9/Constant_2_output_0, %/layers.9/Constant_5_output_0)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Slice_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.10/Slice_output_0 = Slice(%/layers.10/vertex_op.3/maxpool/MaxPool_output_0, %/layers.10/Constant_3_output_0, %/layers.10/Constant_4_output_0, %/layers.10/Constant_2_output_0, %/layers.10/Constant_5_output_0)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Slice_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.11/Slice_output_0 = Slice(%/layers.11/vertex_op.3/maxpool/MaxPool_output_0, %/layers.11/Constant_3_output_0, %/layers.11/Constant_4_output_0, %/layers.11/Constant_2_output_0, %/layers.11/Constant_5_output_0)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Slice_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%753 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %753
}
|
val_accuracy
| 92.097354
| 838,783,232
| 2,785,330
|
{'zcp_epe_nas': 118.73912481590503, 'zcp_fisher': 5.945605278015137, 'zcp_flops': 13420531712.0, 'zcp_grad_norm': 46.91943359375, 'zcp_grasp': 2.499114990234375, 'zcp_jacov': -16.047149735410567, 'zcp_l2_norm': 761.4710083007812, 'zcp_nwot': 215.44921985312658, 'zcp_params': 2785330.0, 'zcp_plain': -0.043729860335588004, 'zcp_snip': 259.2041015625, 'zcp_synflow': 64.70853818901641, 'zcp_zen': 79.63982391357422, 'zcp_val_accuracy': 0.8891226053237911}
| |
NASBench101_165604
|
NASBench101
|
165604
|
6446b12312a600a1e8d4ff25e4a06b23
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_437[FLOAT, 128x3x3x3]
%onnx::Conv_438[FLOAT, 128]
%onnx::Conv_440[FLOAT, 128x128x1x1]
%onnx::Conv_443[FLOAT, 128x128x3x3]
%onnx::Conv_446[FLOAT, 128x128x1x1]
%onnx::Conv_449[FLOAT, 128x128x3x3]
%onnx::Conv_452[FLOAT, 128x128x1x1]
%onnx::Conv_455[FLOAT, 128x128x3x3]
%onnx::Conv_458[FLOAT, 256x128x1x1]
%onnx::Conv_459[FLOAT, 256]
%onnx::Conv_461[FLOAT, 256x256x3x3]
%onnx::Conv_464[FLOAT, 256x256x1x1]
%onnx::Conv_467[FLOAT, 256x256x3x3]
%onnx::Conv_470[FLOAT, 256x256x1x1]
%onnx::Conv_473[FLOAT, 256x256x3x3]
%onnx::Conv_476[FLOAT, 512x256x1x1]
%onnx::Conv_477[FLOAT, 512]
%onnx::Conv_479[FLOAT, 512x512x3x3]
%onnx::Conv_482[FLOAT, 512x512x1x1]
%onnx::Conv_485[FLOAT, 512x512x3x3]
%onnx::Conv_488[FLOAT, 512x512x1x1]
%onnx::Conv_491[FLOAT, 512x512x3x3]
) {
%onnx::Conv_492 = Identity(%onnx::Conv_477)
%onnx::Conv_489 = Identity(%onnx::Conv_477)
%onnx::Conv_486 = Identity(%onnx::Conv_477)
%onnx::Conv_483 = Identity(%onnx::Conv_477)
%onnx::Conv_480 = Identity(%onnx::Conv_477)
%onnx::Conv_474 = Identity(%onnx::Conv_459)
%onnx::Conv_471 = Identity(%onnx::Conv_459)
%onnx::Conv_468 = Identity(%onnx::Conv_459)
%onnx::Conv_465 = Identity(%onnx::Conv_459)
%onnx::Conv_462 = Identity(%onnx::Conv_459)
%onnx::Conv_456 = Identity(%onnx::Conv_438)
%onnx::Conv_453 = Identity(%onnx::Conv_438)
%onnx::Conv_450 = Identity(%onnx::Conv_438)
%onnx::Conv_447 = Identity(%onnx::Conv_438)
%onnx::Conv_444 = Identity(%onnx::Conv_438)
%onnx::Conv_441 = Identity(%onnx::Conv_438)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_437, %onnx::Conv_438)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_440, %onnx::Conv_441)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_443, %onnx::Conv_444)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %/layers.1/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.1/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_446, %onnx::Conv_447)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_449, %onnx::Conv_450)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %/layers.2/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.2/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_452, %onnx::Conv_453)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_455, %onnx::Conv_456)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %/layers.3/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.3/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_458, %onnx::Conv_459)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_461, %onnx::Conv_462)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %/layers.5/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.5/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_464, %onnx::Conv_465)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_467, %onnx::Conv_468)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %/layers.6/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.6/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_470, %onnx::Conv_471)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_473, %onnx::Conv_474)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %/layers.7/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.7/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_476, %onnx::Conv_477)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_479, %onnx::Conv_480)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %/layers.9/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.9/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_482, %onnx::Conv_483)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_485, %onnx::Conv_486)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %/layers.10/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.10/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_488, %onnx::Conv_489)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_491, %onnx::Conv_492)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %/layers.11/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.11/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/vertex_op.5/maxpool/MaxPool_output_0)
%435 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %435
}
|
val_accuracy
| 9.545273
| 3,005,491,200
| 10,177,674
|
{'zcp_epe_nas': 109.31739542896067, 'zcp_fisher': 33.76662826538086, 'zcp_flops': 48087859200.0, 'zcp_grad_norm': 89.0451889038086, 'zcp_grasp': 16.34759521484375, 'zcp_jacov': -16.052433952376912, 'zcp_l2_norm': 411.0027160644531, 'zcp_nwot': 218.25206114599584, 'zcp_params': 10177674.0, 'zcp_plain': -0.044774055480957003, 'zcp_snip': 711.772705078125, 'zcp_synflow': 67.5216263614659, 'zcp_zen': 51.83863830566406, 'zcp_val_accuracy': 0.8780047893524171}
| |
NASBench101_423274
|
NASBench101
|
423274
|
ffc65ed90508517669104ca56a173fb4
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_968[FLOAT, 128x3x3x3]
%onnx::Conv_969[FLOAT, 128]
%onnx::Conv_971[FLOAT, 128x128x1x1]
%onnx::Conv_974[FLOAT, 128x128x1x1]
%onnx::Conv_977[FLOAT, 128x128x1x1]
%onnx::Conv_980[FLOAT, 128x128x1x1]
%onnx::Conv_983[FLOAT, 128x128x1x1]
%onnx::Conv_986[FLOAT, 128x128x3x3]
%onnx::Conv_989[FLOAT, 128x128x1x1]
%onnx::Conv_992[FLOAT, 128x128x1x1]
%onnx::Conv_995[FLOAT, 128x128x1x1]
%onnx::Conv_998[FLOAT, 128x128x1x1]
%onnx::Conv_1001[FLOAT, 128x128x1x1]
%onnx::Conv_1004[FLOAT, 128x128x1x1]
%onnx::Conv_1007[FLOAT, 128x128x3x3]
%onnx::Conv_1010[FLOAT, 128x128x1x1]
%onnx::Conv_1013[FLOAT, 128x128x1x1]
%onnx::Conv_1016[FLOAT, 128x128x1x1]
%onnx::Conv_1019[FLOAT, 128x128x1x1]
%onnx::Conv_1022[FLOAT, 128x128x1x1]
%onnx::Conv_1025[FLOAT, 128x128x1x1]
%onnx::Conv_1028[FLOAT, 128x128x3x3]
%onnx::Conv_1031[FLOAT, 128x128x1x1]
%onnx::Conv_1034[FLOAT, 256x128x1x1]
%onnx::Conv_1035[FLOAT, 256]
%onnx::Conv_1037[FLOAT, 256x256x1x1]
%onnx::Conv_1040[FLOAT, 256x128x1x1]
%onnx::Conv_1043[FLOAT, 256x256x1x1]
%onnx::Conv_1046[FLOAT, 256x128x1x1]
%onnx::Conv_1049[FLOAT, 256x256x3x3]
%onnx::Conv_1052[FLOAT, 256x256x1x1]
%onnx::Conv_1055[FLOAT, 256x256x1x1]
%onnx::Conv_1058[FLOAT, 256x256x1x1]
%onnx::Conv_1061[FLOAT, 256x256x1x1]
%onnx::Conv_1064[FLOAT, 256x256x1x1]
%onnx::Conv_1067[FLOAT, 256x256x1x1]
%onnx::Conv_1070[FLOAT, 256x256x3x3]
%onnx::Conv_1073[FLOAT, 256x256x1x1]
%onnx::Conv_1076[FLOAT, 256x256x1x1]
%onnx::Conv_1079[FLOAT, 256x256x1x1]
%onnx::Conv_1082[FLOAT, 256x256x1x1]
%onnx::Conv_1085[FLOAT, 256x256x1x1]
%onnx::Conv_1088[FLOAT, 256x256x1x1]
%onnx::Conv_1091[FLOAT, 256x256x3x3]
%onnx::Conv_1094[FLOAT, 256x256x1x1]
%onnx::Conv_1097[FLOAT, 512x256x1x1]
%onnx::Conv_1098[FLOAT, 512]
%onnx::Conv_1100[FLOAT, 512x512x1x1]
%onnx::Conv_1103[FLOAT, 512x256x1x1]
%onnx::Conv_1106[FLOAT, 512x512x1x1]
%onnx::Conv_1109[FLOAT, 512x256x1x1]
%onnx::Conv_1112[FLOAT, 512x512x3x3]
%onnx::Conv_1115[FLOAT, 512x512x1x1]
%onnx::Conv_1118[FLOAT, 512x512x1x1]
%onnx::Conv_1121[FLOAT, 512x512x1x1]
%onnx::Conv_1124[FLOAT, 512x512x1x1]
%onnx::Conv_1127[FLOAT, 512x512x1x1]
%onnx::Conv_1130[FLOAT, 512x512x1x1]
%onnx::Conv_1133[FLOAT, 512x512x3x3]
%onnx::Conv_1136[FLOAT, 512x512x1x1]
%onnx::Conv_1139[FLOAT, 512x512x1x1]
%onnx::Conv_1142[FLOAT, 512x512x1x1]
%onnx::Conv_1145[FLOAT, 512x512x1x1]
%onnx::Conv_1148[FLOAT, 512x512x1x1]
%onnx::Conv_1151[FLOAT, 512x512x1x1]
%onnx::Conv_1154[FLOAT, 512x512x3x3]
%onnx::Conv_1157[FLOAT, 512x512x1x1]
) {
%onnx::Conv_1158 = Identity(%onnx::Conv_1098)
%onnx::Conv_1155 = Identity(%onnx::Conv_1098)
%onnx::Conv_1152 = Identity(%onnx::Conv_1098)
%onnx::Conv_1149 = Identity(%onnx::Conv_1098)
%onnx::Conv_1146 = Identity(%onnx::Conv_1098)
%onnx::Conv_1143 = Identity(%onnx::Conv_1098)
%onnx::Conv_1140 = Identity(%onnx::Conv_1098)
%onnx::Conv_1137 = Identity(%onnx::Conv_1098)
%onnx::Conv_1134 = Identity(%onnx::Conv_1098)
%onnx::Conv_1131 = Identity(%onnx::Conv_1098)
%onnx::Conv_1128 = Identity(%onnx::Conv_1098)
%onnx::Conv_1125 = Identity(%onnx::Conv_1098)
%onnx::Conv_1122 = Identity(%onnx::Conv_1098)
%onnx::Conv_1119 = Identity(%onnx::Conv_1098)
%onnx::Conv_1116 = Identity(%onnx::Conv_1098)
%onnx::Conv_1113 = Identity(%onnx::Conv_1098)
%onnx::Conv_1110 = Identity(%onnx::Conv_1098)
%onnx::Conv_1107 = Identity(%onnx::Conv_1098)
%onnx::Conv_1104 = Identity(%onnx::Conv_1098)
%onnx::Conv_1101 = Identity(%onnx::Conv_1098)
%onnx::Conv_1095 = Identity(%onnx::Conv_1035)
%onnx::Conv_1092 = Identity(%onnx::Conv_1035)
%onnx::Conv_1089 = Identity(%onnx::Conv_1035)
%onnx::Conv_1086 = Identity(%onnx::Conv_1035)
%onnx::Conv_1083 = Identity(%onnx::Conv_1035)
%onnx::Conv_1080 = Identity(%onnx::Conv_1035)
%onnx::Conv_1077 = Identity(%onnx::Conv_1035)
%onnx::Conv_1074 = Identity(%onnx::Conv_1035)
%onnx::Conv_1071 = Identity(%onnx::Conv_1035)
%onnx::Conv_1068 = Identity(%onnx::Conv_1035)
%onnx::Conv_1065 = Identity(%onnx::Conv_1035)
%onnx::Conv_1062 = Identity(%onnx::Conv_1035)
%onnx::Conv_1059 = Identity(%onnx::Conv_1035)
%onnx::Conv_1056 = Identity(%onnx::Conv_1035)
%onnx::Conv_1053 = Identity(%onnx::Conv_1035)
%onnx::Conv_1050 = Identity(%onnx::Conv_1035)
%onnx::Conv_1047 = Identity(%onnx::Conv_1035)
%onnx::Conv_1044 = Identity(%onnx::Conv_1035)
%onnx::Conv_1041 = Identity(%onnx::Conv_1035)
%onnx::Conv_1038 = Identity(%onnx::Conv_1035)
%onnx::Conv_1032 = Identity(%onnx::Conv_969)
%onnx::Conv_1029 = Identity(%onnx::Conv_969)
%onnx::Conv_1026 = Identity(%onnx::Conv_969)
%onnx::Conv_1023 = Identity(%onnx::Conv_969)
%onnx::Conv_1020 = Identity(%onnx::Conv_969)
%onnx::Conv_1017 = Identity(%onnx::Conv_969)
%onnx::Conv_1014 = Identity(%onnx::Conv_969)
%onnx::Conv_1011 = Identity(%onnx::Conv_969)
%onnx::Conv_1008 = Identity(%onnx::Conv_969)
%onnx::Conv_1005 = Identity(%onnx::Conv_969)
%onnx::Conv_1002 = Identity(%onnx::Conv_969)
%onnx::Conv_999 = Identity(%onnx::Conv_969)
%onnx::Conv_996 = Identity(%onnx::Conv_969)
%onnx::Conv_993 = Identity(%onnx::Conv_969)
%onnx::Conv_990 = Identity(%onnx::Conv_969)
%onnx::Conv_987 = Identity(%onnx::Conv_969)
%onnx::Conv_984 = Identity(%onnx::Conv_969)
%onnx::Conv_981 = Identity(%onnx::Conv_969)
%onnx::Conv_978 = Identity(%onnx::Conv_969)
%onnx::Conv_975 = Identity(%onnx::Conv_969)
%onnx::Conv_972 = Identity(%onnx::Conv_969)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_968, %onnx::Conv_969)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_974, %onnx::Conv_975)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_977, %onnx::Conv_978)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/Add_2_output_0, %/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/Add_5_output_0 = Add(%/layers.1/Add_4_output_0, %/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/Add_6_output_0 = Add(%/layers.1/Add_5_output_0, %/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_6_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/Add_2_output_0, %/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/Add_5_output_0 = Add(%/layers.2/Add_4_output_0, %/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/Add_6_output_0 = Add(%/layers.2/Add_5_output_0, %/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_6_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1025, %onnx::Conv_1026)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/Add_2_output_0, %/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_1028, %onnx::Conv_1029)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/Add_5_output_0 = Add(%/layers.3/Add_4_output_0, %/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/Add_6_output_0 = Add(%/layers.3/Add_5_output_0, %/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_6_output_0, %onnx::Conv_1031, %onnx::Conv_1032)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1034, %onnx::Conv_1035)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_1037, %onnx::Conv_1038)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1040, %onnx::Conv_1041)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_1043, %onnx::Conv_1044)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1046, %onnx::Conv_1047)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/Add_2_output_0, %/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_1049, %onnx::Conv_1050)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/Add_5_output_0 = Add(%/layers.5/Add_4_output_0, %/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/Add_6_output_0 = Add(%/layers.5/Add_5_output_0, %/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_6_output_0, %onnx::Conv_1052, %onnx::Conv_1053)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1055, %onnx::Conv_1056)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_1058, %onnx::Conv_1059)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1061, %onnx::Conv_1062)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_1064, %onnx::Conv_1065)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1067, %onnx::Conv_1068)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/Add_2_output_0, %/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_1070, %onnx::Conv_1071)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/Add_5_output_0 = Add(%/layers.6/Add_4_output_0, %/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/Add_6_output_0 = Add(%/layers.6/Add_5_output_0, %/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_6_output_0, %onnx::Conv_1073, %onnx::Conv_1074)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1076, %onnx::Conv_1077)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_1079, %onnx::Conv_1080)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1082, %onnx::Conv_1083)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_1085, %onnx::Conv_1086)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1088, %onnx::Conv_1089)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/Add_2_output_0, %/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_1091, %onnx::Conv_1092)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/Add_5_output_0 = Add(%/layers.7/Add_4_output_0, %/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/Add_6_output_0 = Add(%/layers.7/Add_5_output_0, %/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_6_output_0, %onnx::Conv_1094, %onnx::Conv_1095)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1097, %onnx::Conv_1098)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_1100, %onnx::Conv_1101)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1103, %onnx::Conv_1104)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_1106, %onnx::Conv_1107)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1109, %onnx::Conv_1110)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/Add_2_output_0, %/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_1112, %onnx::Conv_1113)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/Add_5_output_0 = Add(%/layers.9/Add_4_output_0, %/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/Add_6_output_0 = Add(%/layers.9/Add_5_output_0, %/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_6_output_0, %onnx::Conv_1115, %onnx::Conv_1116)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1118, %onnx::Conv_1119)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_1121, %onnx::Conv_1122)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1124, %onnx::Conv_1125)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_1127, %onnx::Conv_1128)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1130, %onnx::Conv_1131)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/Add_2_output_0, %/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_1133, %onnx::Conv_1134)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/Add_5_output_0 = Add(%/layers.10/Add_4_output_0, %/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/Add_6_output_0 = Add(%/layers.10/Add_5_output_0, %/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_6_output_0, %onnx::Conv_1136, %onnx::Conv_1137)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1139, %onnx::Conv_1140)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_1142, %onnx::Conv_1143)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1145, %onnx::Conv_1146)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_1148, %onnx::Conv_1149)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1151, %onnx::Conv_1152)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/Add_2_output_0, %/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_1154, %onnx::Conv_1155)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/Add_5_output_0 = Add(%/layers.11/Add_4_output_0, %/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/Add_6_output_0 = Add(%/layers.11/Add_5_output_0, %/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_6_output_0, %onnx::Conv_1157, %onnx::Conv_1158)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%966 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %966
}
|
val_accuracy
| 91.796875
| 4,475,856,896
| 15,037,834
|
{'zcp_epe_nas': 126.90808381865932, 'zcp_fisher': 21.437166213989258, 'zcp_flops': 71613710336.0, 'zcp_grad_norm': 101.77654266357422, 'zcp_grasp': -17.9140625, 'zcp_jacov': -16.047418861525742, 'zcp_l2_norm': 1437.9437255859375, 'zcp_nwot': 237.29551404902904, 'zcp_params': 15037834.0, 'zcp_plain': -0.018805447965860003, 'zcp_snip': 751.9995727539062, 'zcp_synflow': 120.54919220916653, 'zcp_zen': 116.28592681884766, 'zcp_val_accuracy': 0.9172676205635071}
| |
NASBench101_275325
|
NASBench101
|
275325
|
a6b4beee6516b6e044e6d317130c2f17
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_686[FLOAT, 128x3x3x3]
%onnx::Conv_687[FLOAT, 128]
%onnx::Conv_689[FLOAT, 43x128x1x1]
%onnx::Conv_690[FLOAT, 43]
%onnx::Conv_692[FLOAT, 43x128x1x1]
%onnx::Conv_695[FLOAT, 43x43x3x3]
%onnx::Conv_698[FLOAT, 42x42x1x1]
%onnx::Conv_699[FLOAT, 42]
%onnx::Conv_701[FLOAT, 43x128x1x1]
%onnx::Conv_704[FLOAT, 43x128x1x1]
%onnx::Conv_707[FLOAT, 43x43x3x3]
%onnx::Conv_710[FLOAT, 42x42x1x1]
%onnx::Conv_713[FLOAT, 43x128x1x1]
%onnx::Conv_716[FLOAT, 43x128x1x1]
%onnx::Conv_719[FLOAT, 43x43x3x3]
%onnx::Conv_722[FLOAT, 42x42x1x1]
%onnx::Conv_725[FLOAT, 86x128x1x1]
%onnx::Conv_726[FLOAT, 86]
%onnx::Conv_728[FLOAT, 86x128x1x1]
%onnx::Conv_731[FLOAT, 85x85x3x3]
%onnx::Conv_732[FLOAT, 85]
%onnx::Conv_734[FLOAT, 85x85x1x1]
%onnx::Conv_737[FLOAT, 86x256x1x1]
%onnx::Conv_740[FLOAT, 86x256x1x1]
%onnx::Conv_743[FLOAT, 85x85x3x3]
%onnx::Conv_746[FLOAT, 85x85x1x1]
%onnx::Conv_749[FLOAT, 86x256x1x1]
%onnx::Conv_752[FLOAT, 86x256x1x1]
%onnx::Conv_755[FLOAT, 85x85x3x3]
%onnx::Conv_758[FLOAT, 85x85x1x1]
%onnx::Conv_761[FLOAT, 171x256x1x1]
%onnx::Conv_762[FLOAT, 171]
%onnx::Conv_764[FLOAT, 171x256x1x1]
%onnx::Conv_767[FLOAT, 171x171x3x3]
%onnx::Conv_770[FLOAT, 170x170x1x1]
%onnx::Conv_771[FLOAT, 170]
%onnx::Conv_773[FLOAT, 171x512x1x1]
%onnx::Conv_776[FLOAT, 171x512x1x1]
%onnx::Conv_779[FLOAT, 171x171x3x3]
%onnx::Conv_782[FLOAT, 170x170x1x1]
%onnx::Conv_785[FLOAT, 171x512x1x1]
%onnx::Conv_788[FLOAT, 171x512x1x1]
%onnx::Conv_791[FLOAT, 171x171x3x3]
%onnx::Conv_794[FLOAT, 170x170x1x1]
) {
%onnx::Conv_795 = Identity(%onnx::Conv_771)
%onnx::Conv_792 = Identity(%onnx::Conv_762)
%onnx::Conv_789 = Identity(%onnx::Conv_762)
%onnx::Conv_786 = Identity(%onnx::Conv_762)
%onnx::Conv_783 = Identity(%onnx::Conv_771)
%onnx::Conv_780 = Identity(%onnx::Conv_762)
%onnx::Conv_777 = Identity(%onnx::Conv_762)
%onnx::Conv_774 = Identity(%onnx::Conv_762)
%onnx::Conv_768 = Identity(%onnx::Conv_762)
%onnx::Conv_765 = Identity(%onnx::Conv_762)
%onnx::Conv_759 = Identity(%onnx::Conv_732)
%onnx::Conv_756 = Identity(%onnx::Conv_732)
%onnx::Conv_753 = Identity(%onnx::Conv_726)
%onnx::Conv_750 = Identity(%onnx::Conv_726)
%onnx::Conv_747 = Identity(%onnx::Conv_732)
%onnx::Conv_744 = Identity(%onnx::Conv_732)
%onnx::Conv_741 = Identity(%onnx::Conv_726)
%onnx::Conv_738 = Identity(%onnx::Conv_726)
%onnx::Conv_735 = Identity(%onnx::Conv_732)
%onnx::Conv_729 = Identity(%onnx::Conv_726)
%onnx::Conv_723 = Identity(%onnx::Conv_699)
%onnx::Conv_720 = Identity(%onnx::Conv_690)
%onnx::Conv_717 = Identity(%onnx::Conv_690)
%onnx::Conv_714 = Identity(%onnx::Conv_690)
%onnx::Conv_711 = Identity(%onnx::Conv_699)
%onnx::Conv_708 = Identity(%onnx::Conv_690)
%onnx::Conv_705 = Identity(%onnx::Conv_690)
%onnx::Conv_702 = Identity(%onnx::Conv_690)
%onnx::Conv_696 = Identity(%onnx::Conv_690)
%onnx::Conv_693 = Identity(%onnx::Conv_690)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_686, %onnx::Conv_687)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_689, %onnx::Conv_690)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_692, %onnx::Conv_693)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_695, %onnx::Conv_696)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.1/Slice_output_0 = Slice(%/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %/layers.1/Constant_2_output_0, %/layers.1/Constant_3_output_0, %/layers.1/Constant_1_output_0, %/layers.1/Constant_4_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Slice_output_0)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_698, %onnx::Conv_699)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_701, %onnx::Conv_702)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_704, %onnx::Conv_705)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_707, %onnx::Conv_708)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.2/Slice_output_0 = Slice(%/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %/layers.2/Constant_2_output_0, %/layers.2/Constant_3_output_0, %/layers.2/Constant_1_output_0, %/layers.2/Constant_4_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Slice_output_0)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_710, %onnx::Conv_711)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_713, %onnx::Conv_714)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_716, %onnx::Conv_717)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_719, %onnx::Conv_720)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.3/Slice_output_0 = Slice(%/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %/layers.3/Constant_2_output_0, %/layers.3/Constant_3_output_0, %/layers.3/Constant_1_output_0, %/layers.3/Constant_4_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Slice_output_0)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_722, %onnx::Conv_723)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_725, %onnx::Conv_726)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_728, %onnx::Conv_729)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.5/Slice_output_0 = Slice(%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/Constant_2_output_0, %/layers.5/Constant_3_output_0, %/layers.5/Constant_1_output_0, %/layers.5/Constant_4_output_0)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Slice_output_0, %onnx::Conv_731, %onnx::Conv_732)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.5/Slice_1_output_0 = Slice(%/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %/layers.5/Constant_6_output_0, %/layers.5/Constant_7_output_0, %/layers.5/Constant_5_output_0, %/layers.5/Constant_8_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Slice_1_output_0)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_734, %onnx::Conv_735)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_737, %onnx::Conv_738)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_740, %onnx::Conv_741)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.6/Slice_output_0 = Slice(%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/Constant_2_output_0, %/layers.6/Constant_3_output_0, %/layers.6/Constant_1_output_0, %/layers.6/Constant_4_output_0)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Slice_output_0, %onnx::Conv_743, %onnx::Conv_744)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.6/Slice_1_output_0 = Slice(%/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %/layers.6/Constant_6_output_0, %/layers.6/Constant_7_output_0, %/layers.6/Constant_5_output_0, %/layers.6/Constant_8_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Slice_1_output_0)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_746, %onnx::Conv_747)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_749, %onnx::Conv_750)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_752, %onnx::Conv_753)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.7/Slice_output_0 = Slice(%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/Constant_2_output_0, %/layers.7/Constant_3_output_0, %/layers.7/Constant_1_output_0, %/layers.7/Constant_4_output_0)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Slice_output_0, %onnx::Conv_755, %onnx::Conv_756)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.7/Slice_1_output_0 = Slice(%/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %/layers.7/Constant_6_output_0, %/layers.7/Constant_7_output_0, %/layers.7/Constant_5_output_0, %/layers.7/Constant_8_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Slice_1_output_0)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_758, %onnx::Conv_759)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_761, %onnx::Conv_762)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_764, %onnx::Conv_765)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_767, %onnx::Conv_768)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.9/Slice_output_0 = Slice(%/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %/layers.9/Constant_2_output_0, %/layers.9/Constant_3_output_0, %/layers.9/Constant_1_output_0, %/layers.9/Constant_4_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Slice_output_0)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_770, %onnx::Conv_771)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_773, %onnx::Conv_774)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_776, %onnx::Conv_777)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_779, %onnx::Conv_780)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.10/Slice_output_0 = Slice(%/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %/layers.10/Constant_2_output_0, %/layers.10/Constant_3_output_0, %/layers.10/Constant_1_output_0, %/layers.10/Constant_4_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Slice_output_0)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_782, %onnx::Conv_783)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_785, %onnx::Conv_786)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_788, %onnx::Conv_789)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.11/Slice_output_0 = Slice(%/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %/layers.11/Constant_2_output_0, %/layers.11/Constant_3_output_0, %/layers.11/Constant_1_output_0, %/layers.11/Constant_4_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Slice_output_0)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%684 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %684
}
|
val_accuracy
| 89.623398
| 531,727,744
| 1,745,054
|
{'zcp_epe_nas': 65.36512137818917, 'zcp_fisher': 3.479568004608154, 'zcp_flops': 8507643904.0, 'zcp_grad_norm': 34.38115310668945, 'zcp_grasp': 0.5734405517578121, 'zcp_jacov': -16.046241717933192, 'zcp_l2_norm': 638.6775512695312, 'zcp_nwot': 212.4881015075008, 'zcp_params': 1745054.0, 'zcp_plain': -0.00024168047821100002, 'zcp_snip': 190.37240600585938, 'zcp_synflow': 62.05549789215189, 'zcp_zen': 67.20059204101562, 'zcp_val_accuracy': 0.9377003312110901}
| |
NASBench101_66131
|
NASBench101
|
66131
|
2825397244355f35d7eec551533f04cc
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_635[FLOAT, 128x3x3x3]
%onnx::Conv_636[FLOAT, 128]
%onnx::Conv_638[FLOAT, 64x128x1x1]
%onnx::Conv_639[FLOAT, 64]
%onnx::Conv_641[FLOAT, 64x128x1x1]
%onnx::Conv_644[FLOAT, 64x64x3x3]
%onnx::Conv_647[FLOAT, 64x64x3x3]
%onnx::Conv_650[FLOAT, 64x128x1x1]
%onnx::Conv_653[FLOAT, 64x128x1x1]
%onnx::Conv_656[FLOAT, 64x64x3x3]
%onnx::Conv_659[FLOAT, 64x64x3x3]
%onnx::Conv_662[FLOAT, 64x128x1x1]
%onnx::Conv_665[FLOAT, 64x128x1x1]
%onnx::Conv_668[FLOAT, 64x64x3x3]
%onnx::Conv_671[FLOAT, 64x64x3x3]
%onnx::Conv_674[FLOAT, 128x128x1x1]
%onnx::Conv_677[FLOAT, 128x128x1x1]
%onnx::Conv_680[FLOAT, 128x128x3x3]
%onnx::Conv_683[FLOAT, 128x128x3x3]
%onnx::Conv_686[FLOAT, 128x256x1x1]
%onnx::Conv_689[FLOAT, 128x256x1x1]
%onnx::Conv_692[FLOAT, 128x128x3x3]
%onnx::Conv_695[FLOAT, 128x128x3x3]
%onnx::Conv_698[FLOAT, 128x256x1x1]
%onnx::Conv_701[FLOAT, 128x256x1x1]
%onnx::Conv_704[FLOAT, 128x128x3x3]
%onnx::Conv_707[FLOAT, 128x128x3x3]
%onnx::Conv_710[FLOAT, 256x256x1x1]
%onnx::Conv_711[FLOAT, 256]
%onnx::Conv_713[FLOAT, 256x256x1x1]
%onnx::Conv_716[FLOAT, 256x256x3x3]
%onnx::Conv_719[FLOAT, 256x256x3x3]
%onnx::Conv_722[FLOAT, 256x512x1x1]
%onnx::Conv_725[FLOAT, 256x512x1x1]
%onnx::Conv_728[FLOAT, 256x256x3x3]
%onnx::Conv_731[FLOAT, 256x256x3x3]
%onnx::Conv_734[FLOAT, 256x512x1x1]
%onnx::Conv_737[FLOAT, 256x512x1x1]
%onnx::Conv_740[FLOAT, 256x256x3x3]
%onnx::Conv_743[FLOAT, 256x256x3x3]
) {
%onnx::Conv_744 = Identity(%onnx::Conv_711)
%onnx::Conv_741 = Identity(%onnx::Conv_711)
%onnx::Conv_738 = Identity(%onnx::Conv_711)
%onnx::Conv_735 = Identity(%onnx::Conv_711)
%onnx::Conv_732 = Identity(%onnx::Conv_711)
%onnx::Conv_729 = Identity(%onnx::Conv_711)
%onnx::Conv_726 = Identity(%onnx::Conv_711)
%onnx::Conv_723 = Identity(%onnx::Conv_711)
%onnx::Conv_720 = Identity(%onnx::Conv_711)
%onnx::Conv_717 = Identity(%onnx::Conv_711)
%onnx::Conv_714 = Identity(%onnx::Conv_711)
%onnx::Conv_708 = Identity(%onnx::Conv_636)
%onnx::Conv_705 = Identity(%onnx::Conv_636)
%onnx::Conv_702 = Identity(%onnx::Conv_636)
%onnx::Conv_699 = Identity(%onnx::Conv_636)
%onnx::Conv_696 = Identity(%onnx::Conv_636)
%onnx::Conv_693 = Identity(%onnx::Conv_636)
%onnx::Conv_690 = Identity(%onnx::Conv_636)
%onnx::Conv_687 = Identity(%onnx::Conv_636)
%onnx::Conv_684 = Identity(%onnx::Conv_636)
%onnx::Conv_681 = Identity(%onnx::Conv_636)
%onnx::Conv_678 = Identity(%onnx::Conv_636)
%onnx::Conv_675 = Identity(%onnx::Conv_636)
%onnx::Conv_672 = Identity(%onnx::Conv_639)
%onnx::Conv_669 = Identity(%onnx::Conv_639)
%onnx::Conv_666 = Identity(%onnx::Conv_639)
%onnx::Conv_663 = Identity(%onnx::Conv_639)
%onnx::Conv_660 = Identity(%onnx::Conv_639)
%onnx::Conv_657 = Identity(%onnx::Conv_639)
%onnx::Conv_654 = Identity(%onnx::Conv_639)
%onnx::Conv_651 = Identity(%onnx::Conv_639)
%onnx::Conv_648 = Identity(%onnx::Conv_639)
%onnx::Conv_645 = Identity(%onnx::Conv_639)
%onnx::Conv_642 = Identity(%onnx::Conv_639)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_635, %onnx::Conv_636)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_638, %onnx::Conv_639)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_641, %onnx::Conv_642)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_644, %onnx::Conv_645)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_647, %onnx::Conv_648)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_650, %onnx::Conv_651)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_653, %onnx::Conv_654)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_656, %onnx::Conv_657)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_659, %onnx::Conv_660)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_662, %onnx::Conv_663)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_665, %onnx::Conv_666)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_668, %onnx::Conv_669)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_671, %onnx::Conv_672)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_674, %onnx::Conv_675)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_677, %onnx::Conv_678)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_680, %onnx::Conv_681)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_683, %onnx::Conv_684)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_686, %onnx::Conv_687)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_689, %onnx::Conv_690)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_692, %onnx::Conv_693)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_695, %onnx::Conv_696)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_698, %onnx::Conv_699)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_701, %onnx::Conv_702)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_704, %onnx::Conv_705)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_707, %onnx::Conv_708)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_710, %onnx::Conv_711)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_713, %onnx::Conv_714)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_716, %onnx::Conv_717)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_719, %onnx::Conv_720)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_722, %onnx::Conv_723)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_725, %onnx::Conv_726)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_728, %onnx::Conv_729)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_731, %onnx::Conv_732)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_734, %onnx::Conv_735)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_737, %onnx::Conv_738)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_740, %onnx::Conv_741)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_743, %onnx::Conv_744)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%633 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %633
}
|
val_accuracy
| 89.853764
| 1,646,536,704
| 5,532,810
|
{'zcp_epe_nas': 143.46497345205987, 'zcp_fisher': 4.732057094573975, 'zcp_flops': 26344587264.0, 'zcp_grad_norm': 42.29754638671875, 'zcp_grasp': 1.095245361328125, 'zcp_jacov': -16.066192598755485, 'zcp_l2_norm': 694.6091918945312, 'zcp_nwot': 217.62764406809663, 'zcp_params': 5532810.0, 'zcp_plain': -0.006610546726733, 'zcp_snip': 259.4734802246094, 'zcp_synflow': 100.15632767248111, 'zcp_zen': 82.48707580566406, 'zcp_val_accuracy': 0.9278846383094781}
| |
NASBench101_402109
|
NASBench101
|
402109
|
f31d36ea5dd0450c1099f3cbe2fe238d
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_860[FLOAT, 128x3x3x3]
%onnx::Conv_861[FLOAT, 128]
%onnx::Conv_863[FLOAT, 64x128x1x1]
%onnx::Conv_864[FLOAT, 64]
%onnx::Conv_866[FLOAT, 64x64x3x3]
%onnx::Conv_869[FLOAT, 64x128x1x1]
%onnx::Conv_872[FLOAT, 64x128x1x1]
%onnx::Conv_875[FLOAT, 64x64x3x3]
%onnx::Conv_878[FLOAT, 64x64x3x3]
%onnx::Conv_881[FLOAT, 64x128x1x1]
%onnx::Conv_884[FLOAT, 64x64x3x3]
%onnx::Conv_887[FLOAT, 64x128x1x1]
%onnx::Conv_890[FLOAT, 64x128x1x1]
%onnx::Conv_893[FLOAT, 64x64x3x3]
%onnx::Conv_896[FLOAT, 64x64x3x3]
%onnx::Conv_899[FLOAT, 64x128x1x1]
%onnx::Conv_902[FLOAT, 64x64x3x3]
%onnx::Conv_905[FLOAT, 64x128x1x1]
%onnx::Conv_908[FLOAT, 64x128x1x1]
%onnx::Conv_911[FLOAT, 64x64x3x3]
%onnx::Conv_914[FLOAT, 64x64x3x3]
%onnx::Conv_917[FLOAT, 128x128x1x1]
%onnx::Conv_920[FLOAT, 128x128x3x3]
%onnx::Conv_923[FLOAT, 128x128x1x1]
%onnx::Conv_926[FLOAT, 128x128x1x1]
%onnx::Conv_929[FLOAT, 128x128x3x3]
%onnx::Conv_932[FLOAT, 128x128x3x3]
%onnx::Conv_935[FLOAT, 128x256x1x1]
%onnx::Conv_938[FLOAT, 128x128x3x3]
%onnx::Conv_941[FLOAT, 128x256x1x1]
%onnx::Conv_944[FLOAT, 128x256x1x1]
%onnx::Conv_947[FLOAT, 128x128x3x3]
%onnx::Conv_950[FLOAT, 128x128x3x3]
%onnx::Conv_953[FLOAT, 128x256x1x1]
%onnx::Conv_956[FLOAT, 128x128x3x3]
%onnx::Conv_959[FLOAT, 128x256x1x1]
%onnx::Conv_962[FLOAT, 128x256x1x1]
%onnx::Conv_965[FLOAT, 128x128x3x3]
%onnx::Conv_968[FLOAT, 128x128x3x3]
%onnx::Conv_971[FLOAT, 256x256x1x1]
%onnx::Conv_972[FLOAT, 256]
%onnx::Conv_974[FLOAT, 256x256x3x3]
%onnx::Conv_977[FLOAT, 256x256x1x1]
%onnx::Conv_980[FLOAT, 256x256x1x1]
%onnx::Conv_983[FLOAT, 256x256x3x3]
%onnx::Conv_986[FLOAT, 256x256x3x3]
%onnx::Conv_989[FLOAT, 256x512x1x1]
%onnx::Conv_992[FLOAT, 256x256x3x3]
%onnx::Conv_995[FLOAT, 256x512x1x1]
%onnx::Conv_998[FLOAT, 256x512x1x1]
%onnx::Conv_1001[FLOAT, 256x256x3x3]
%onnx::Conv_1004[FLOAT, 256x256x3x3]
%onnx::Conv_1007[FLOAT, 256x512x1x1]
%onnx::Conv_1010[FLOAT, 256x256x3x3]
%onnx::Conv_1013[FLOAT, 256x512x1x1]
%onnx::Conv_1016[FLOAT, 256x512x1x1]
%onnx::Conv_1019[FLOAT, 256x256x3x3]
%onnx::Conv_1022[FLOAT, 256x256x3x3]
) {
%onnx::Conv_1023 = Identity(%onnx::Conv_972)
%onnx::Conv_1020 = Identity(%onnx::Conv_972)
%onnx::Conv_1017 = Identity(%onnx::Conv_972)
%onnx::Conv_1014 = Identity(%onnx::Conv_972)
%onnx::Conv_1011 = Identity(%onnx::Conv_972)
%onnx::Conv_1008 = Identity(%onnx::Conv_972)
%onnx::Conv_1005 = Identity(%onnx::Conv_972)
%onnx::Conv_1002 = Identity(%onnx::Conv_972)
%onnx::Conv_999 = Identity(%onnx::Conv_972)
%onnx::Conv_996 = Identity(%onnx::Conv_972)
%onnx::Conv_993 = Identity(%onnx::Conv_972)
%onnx::Conv_990 = Identity(%onnx::Conv_972)
%onnx::Conv_987 = Identity(%onnx::Conv_972)
%onnx::Conv_984 = Identity(%onnx::Conv_972)
%onnx::Conv_981 = Identity(%onnx::Conv_972)
%onnx::Conv_978 = Identity(%onnx::Conv_972)
%onnx::Conv_975 = Identity(%onnx::Conv_972)
%onnx::Conv_969 = Identity(%onnx::Conv_861)
%onnx::Conv_966 = Identity(%onnx::Conv_861)
%onnx::Conv_963 = Identity(%onnx::Conv_861)
%onnx::Conv_960 = Identity(%onnx::Conv_861)
%onnx::Conv_957 = Identity(%onnx::Conv_861)
%onnx::Conv_954 = Identity(%onnx::Conv_861)
%onnx::Conv_951 = Identity(%onnx::Conv_861)
%onnx::Conv_948 = Identity(%onnx::Conv_861)
%onnx::Conv_945 = Identity(%onnx::Conv_861)
%onnx::Conv_942 = Identity(%onnx::Conv_861)
%onnx::Conv_939 = Identity(%onnx::Conv_861)
%onnx::Conv_936 = Identity(%onnx::Conv_861)
%onnx::Conv_933 = Identity(%onnx::Conv_861)
%onnx::Conv_930 = Identity(%onnx::Conv_861)
%onnx::Conv_927 = Identity(%onnx::Conv_861)
%onnx::Conv_924 = Identity(%onnx::Conv_861)
%onnx::Conv_921 = Identity(%onnx::Conv_861)
%onnx::Conv_918 = Identity(%onnx::Conv_861)
%onnx::Conv_915 = Identity(%onnx::Conv_864)
%onnx::Conv_912 = Identity(%onnx::Conv_864)
%onnx::Conv_909 = Identity(%onnx::Conv_864)
%onnx::Conv_906 = Identity(%onnx::Conv_864)
%onnx::Conv_903 = Identity(%onnx::Conv_864)
%onnx::Conv_900 = Identity(%onnx::Conv_864)
%onnx::Conv_897 = Identity(%onnx::Conv_864)
%onnx::Conv_894 = Identity(%onnx::Conv_864)
%onnx::Conv_891 = Identity(%onnx::Conv_864)
%onnx::Conv_888 = Identity(%onnx::Conv_864)
%onnx::Conv_885 = Identity(%onnx::Conv_864)
%onnx::Conv_882 = Identity(%onnx::Conv_864)
%onnx::Conv_879 = Identity(%onnx::Conv_864)
%onnx::Conv_876 = Identity(%onnx::Conv_864)
%onnx::Conv_873 = Identity(%onnx::Conv_864)
%onnx::Conv_870 = Identity(%onnx::Conv_864)
%onnx::Conv_867 = Identity(%onnx::Conv_864)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_860, %onnx::Conv_861)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.3/maxpool/MaxPool_output_0, %/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/Add_3_output_0, %/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.3/maxpool/MaxPool_output_0, %/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/Add_3_output_0, %/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_908, %onnx::Conv_909)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.3/maxpool/MaxPool_output_0, %/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_911, %onnx::Conv_912)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/Add_3_output_0, %/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_4_output_0, %onnx::Conv_914, %onnx::Conv_915)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_917, %onnx::Conv_918)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_920, %onnx::Conv_921)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_923, %onnx::Conv_924)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_926, %onnx::Conv_927)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.3/maxpool/MaxPool_output_0, %/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_929, %onnx::Conv_930)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/Add_3_output_0, %/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_932, %onnx::Conv_933)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_935, %onnx::Conv_936)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_938, %onnx::Conv_939)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_941, %onnx::Conv_942)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_944, %onnx::Conv_945)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.3/maxpool/MaxPool_output_0, %/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_947, %onnx::Conv_948)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/Add_3_output_0, %/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_950, %onnx::Conv_951)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_953, %onnx::Conv_954)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_956, %onnx::Conv_957)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_959, %onnx::Conv_960)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_962, %onnx::Conv_963)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.3/maxpool/MaxPool_output_0, %/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_965, %onnx::Conv_966)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/Add_3_output_0, %/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_4_output_0, %onnx::Conv_968, %onnx::Conv_969)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_974, %onnx::Conv_975)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_977, %onnx::Conv_978)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.3/maxpool/MaxPool_output_0, %/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/Add_3_output_0, %/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.3/maxpool/MaxPool_output_0, %/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/Add_3_output_0, %/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.3/maxpool/MaxPool_output_0, %/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/Add_3_output_0, %/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_4_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%858 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %858
}
|
val_accuracy
| 93.80008
| 2,465,736,704
| 8,294,794
|
{'zcp_epe_nas': 78.98818307947563, 'zcp_fisher': 8.270933151245117, 'zcp_flops': 39451787264.0, 'zcp_grad_norm': 54.180259704589844, 'zcp_grasp': -1.254287719726562, 'zcp_jacov': -16.055767262089077, 'zcp_l2_norm': 1040.3314208984375, 'zcp_nwot': 224.1306047724425, 'zcp_params': 8294794.0, 'zcp_plain': -0.017939947545528002, 'zcp_snip': 365.9729309082031, 'zcp_synflow': 103.50428610314485, 'zcp_zen': 110.12263488769531, 'zcp_val_accuracy': 0.9035456776618951}
| |
NASBench101_157046
|
NASBench101
|
157046
|
5f0f40b68179414705b5cad6d99313f3
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_878[FLOAT, 128x3x3x3]
%onnx::Conv_879[FLOAT, 128]
%onnx::Conv_881[FLOAT, 43x128x1x1]
%onnx::Conv_882[FLOAT, 43]
%onnx::Conv_884[FLOAT, 43x43x1x1]
%onnx::Conv_887[FLOAT, 43x128x1x1]
%onnx::Conv_890[FLOAT, 43x43x1x1]
%onnx::Conv_893[FLOAT, 43x43x1x1]
%onnx::Conv_896[FLOAT, 42x42x3x3]
%onnx::Conv_897[FLOAT, 42]
%onnx::Conv_899[FLOAT, 43x128x1x1]
%onnx::Conv_902[FLOAT, 43x43x1x1]
%onnx::Conv_905[FLOAT, 43x128x1x1]
%onnx::Conv_908[FLOAT, 43x43x1x1]
%onnx::Conv_911[FLOAT, 43x43x1x1]
%onnx::Conv_914[FLOAT, 42x42x3x3]
%onnx::Conv_917[FLOAT, 43x128x1x1]
%onnx::Conv_920[FLOAT, 43x43x1x1]
%onnx::Conv_923[FLOAT, 43x128x1x1]
%onnx::Conv_926[FLOAT, 43x43x1x1]
%onnx::Conv_929[FLOAT, 43x43x1x1]
%onnx::Conv_932[FLOAT, 42x42x3x3]
%onnx::Conv_935[FLOAT, 86x128x1x1]
%onnx::Conv_936[FLOAT, 86]
%onnx::Conv_938[FLOAT, 86x86x1x1]
%onnx::Conv_941[FLOAT, 86x128x1x1]
%onnx::Conv_944[FLOAT, 86x86x1x1]
%onnx::Conv_947[FLOAT, 85x85x1x1]
%onnx::Conv_948[FLOAT, 85]
%onnx::Conv_950[FLOAT, 85x85x3x3]
%onnx::Conv_953[FLOAT, 86x256x1x1]
%onnx::Conv_956[FLOAT, 86x86x1x1]
%onnx::Conv_959[FLOAT, 86x256x1x1]
%onnx::Conv_962[FLOAT, 86x86x1x1]
%onnx::Conv_965[FLOAT, 85x85x1x1]
%onnx::Conv_968[FLOAT, 85x85x3x3]
%onnx::Conv_971[FLOAT, 86x256x1x1]
%onnx::Conv_974[FLOAT, 86x86x1x1]
%onnx::Conv_977[FLOAT, 86x256x1x1]
%onnx::Conv_980[FLOAT, 86x86x1x1]
%onnx::Conv_983[FLOAT, 85x85x1x1]
%onnx::Conv_986[FLOAT, 85x85x3x3]
%onnx::Conv_989[FLOAT, 171x256x1x1]
%onnx::Conv_990[FLOAT, 171]
%onnx::Conv_992[FLOAT, 171x171x1x1]
%onnx::Conv_995[FLOAT, 171x256x1x1]
%onnx::Conv_998[FLOAT, 171x171x1x1]
%onnx::Conv_1001[FLOAT, 171x171x1x1]
%onnx::Conv_1004[FLOAT, 170x170x3x3]
%onnx::Conv_1005[FLOAT, 170]
%onnx::Conv_1007[FLOAT, 171x512x1x1]
%onnx::Conv_1010[FLOAT, 171x171x1x1]
%onnx::Conv_1013[FLOAT, 171x512x1x1]
%onnx::Conv_1016[FLOAT, 171x171x1x1]
%onnx::Conv_1019[FLOAT, 171x171x1x1]
%onnx::Conv_1022[FLOAT, 170x170x3x3]
%onnx::Conv_1025[FLOAT, 171x512x1x1]
%onnx::Conv_1028[FLOAT, 171x171x1x1]
%onnx::Conv_1031[FLOAT, 171x512x1x1]
%onnx::Conv_1034[FLOAT, 171x171x1x1]
%onnx::Conv_1037[FLOAT, 171x171x1x1]
%onnx::Conv_1040[FLOAT, 170x170x3x3]
) {
%onnx::Conv_1041 = Identity(%onnx::Conv_1005)
%onnx::Conv_1038 = Identity(%onnx::Conv_990)
%onnx::Conv_1035 = Identity(%onnx::Conv_990)
%onnx::Conv_1032 = Identity(%onnx::Conv_990)
%onnx::Conv_1029 = Identity(%onnx::Conv_990)
%onnx::Conv_1026 = Identity(%onnx::Conv_990)
%onnx::Conv_1023 = Identity(%onnx::Conv_1005)
%onnx::Conv_1020 = Identity(%onnx::Conv_990)
%onnx::Conv_1017 = Identity(%onnx::Conv_990)
%onnx::Conv_1014 = Identity(%onnx::Conv_990)
%onnx::Conv_1011 = Identity(%onnx::Conv_990)
%onnx::Conv_1008 = Identity(%onnx::Conv_990)
%onnx::Conv_1002 = Identity(%onnx::Conv_990)
%onnx::Conv_999 = Identity(%onnx::Conv_990)
%onnx::Conv_996 = Identity(%onnx::Conv_990)
%onnx::Conv_993 = Identity(%onnx::Conv_990)
%onnx::Conv_987 = Identity(%onnx::Conv_948)
%onnx::Conv_984 = Identity(%onnx::Conv_948)
%onnx::Conv_981 = Identity(%onnx::Conv_936)
%onnx::Conv_978 = Identity(%onnx::Conv_936)
%onnx::Conv_975 = Identity(%onnx::Conv_936)
%onnx::Conv_972 = Identity(%onnx::Conv_936)
%onnx::Conv_969 = Identity(%onnx::Conv_948)
%onnx::Conv_966 = Identity(%onnx::Conv_948)
%onnx::Conv_963 = Identity(%onnx::Conv_936)
%onnx::Conv_960 = Identity(%onnx::Conv_936)
%onnx::Conv_957 = Identity(%onnx::Conv_936)
%onnx::Conv_954 = Identity(%onnx::Conv_936)
%onnx::Conv_951 = Identity(%onnx::Conv_948)
%onnx::Conv_945 = Identity(%onnx::Conv_936)
%onnx::Conv_942 = Identity(%onnx::Conv_936)
%onnx::Conv_939 = Identity(%onnx::Conv_936)
%onnx::Conv_933 = Identity(%onnx::Conv_897)
%onnx::Conv_930 = Identity(%onnx::Conv_882)
%onnx::Conv_927 = Identity(%onnx::Conv_882)
%onnx::Conv_924 = Identity(%onnx::Conv_882)
%onnx::Conv_921 = Identity(%onnx::Conv_882)
%onnx::Conv_918 = Identity(%onnx::Conv_882)
%onnx::Conv_915 = Identity(%onnx::Conv_897)
%onnx::Conv_912 = Identity(%onnx::Conv_882)
%onnx::Conv_909 = Identity(%onnx::Conv_882)
%onnx::Conv_906 = Identity(%onnx::Conv_882)
%onnx::Conv_903 = Identity(%onnx::Conv_882)
%onnx::Conv_900 = Identity(%onnx::Conv_882)
%onnx::Conv_894 = Identity(%onnx::Conv_882)
%onnx::Conv_891 = Identity(%onnx::Conv_882)
%onnx::Conv_888 = Identity(%onnx::Conv_882)
%onnx::Conv_885 = Identity(%onnx::Conv_882)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_878, %onnx::Conv_879)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/Add_1_output_0, %/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.1/Slice_output_0 = Slice(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_4_output_0, %/layers.1/Constant_5_output_0, %/layers.1/Constant_3_output_0, %/layers.1/Constant_6_output_0)
%/layers.1/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_4_output_0 = Add(%/layers.1/Slice_output_0, %/layers.1/Constant_7_output_0)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/Add_1_output_0, %/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_908, %onnx::Conv_909)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_911, %onnx::Conv_912)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.2/Slice_output_0 = Slice(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_4_output_0, %/layers.2/Constant_5_output_0, %/layers.2/Constant_3_output_0, %/layers.2/Constant_6_output_0)
%/layers.2/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_4_output_0 = Add(%/layers.2/Slice_output_0, %/layers.2/Constant_7_output_0)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_914, %onnx::Conv_915)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_917, %onnx::Conv_918)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_920, %onnx::Conv_921)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_923, %onnx::Conv_924)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/Add_1_output_0, %/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_926, %onnx::Conv_927)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_929, %onnx::Conv_930)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.3/Slice_output_0 = Slice(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_4_output_0, %/layers.3/Constant_5_output_0, %/layers.3/Constant_3_output_0, %/layers.3/Constant_6_output_0)
%/layers.3/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_4_output_0 = Add(%/layers.3/Slice_output_0, %/layers.3/Constant_7_output_0)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_4_output_0, %onnx::Conv_932, %onnx::Conv_933)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_935, %onnx::Conv_936)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_938, %onnx::Conv_939)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_941, %onnx::Conv_942)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/Add_1_output_0, %/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_944, %onnx::Conv_945)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.5/Slice_output_0 = Slice(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0, %/layers.5/Constant_4_output_0, %/layers.5/Constant_2_output_0, %/layers.5/Constant_5_output_0)
%/layers.5/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/Slice_output_0, %/layers.5/Constant_6_output_0)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_947, %onnx::Conv_948)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_7_output_0)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_950, %onnx::Conv_951)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_953, %onnx::Conv_954)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_956, %onnx::Conv_957)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_959, %onnx::Conv_960)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/Add_1_output_0, %/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_962, %onnx::Conv_963)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.6/Slice_output_0 = Slice(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0, %/layers.6/Constant_4_output_0, %/layers.6/Constant_2_output_0, %/layers.6/Constant_5_output_0)
%/layers.6/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/Slice_output_0, %/layers.6/Constant_6_output_0)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_965, %onnx::Conv_966)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_7_output_0)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_968, %onnx::Conv_969)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_974, %onnx::Conv_975)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_977, %onnx::Conv_978)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/Add_1_output_0, %/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.7/Slice_output_0 = Slice(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0, %/layers.7/Constant_4_output_0, %/layers.7/Constant_2_output_0, %/layers.7/Constant_5_output_0)
%/layers.7/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/Slice_output_0, %/layers.7/Constant_6_output_0)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_7_output_0)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_4_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/Add_1_output_0, %/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.9/Slice_output_0 = Slice(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_4_output_0, %/layers.9/Constant_5_output_0, %/layers.9/Constant_3_output_0, %/layers.9/Constant_6_output_0)
%/layers.9/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_4_output_0 = Add(%/layers.9/Slice_output_0, %/layers.9/Constant_7_output_0)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/Add_1_output_0, %/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.10/Slice_output_0 = Slice(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_4_output_0, %/layers.10/Constant_5_output_0, %/layers.10/Constant_3_output_0, %/layers.10/Constant_6_output_0)
%/layers.10/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_4_output_0 = Add(%/layers.10/Slice_output_0, %/layers.10/Constant_7_output_0)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1025, %onnx::Conv_1026)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_1028, %onnx::Conv_1029)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1031, %onnx::Conv_1032)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/Add_1_output_0, %/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_1034, %onnx::Conv_1035)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_1037, %onnx::Conv_1038)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.11/Slice_output_0 = Slice(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_4_output_0, %/layers.11/Constant_5_output_0, %/layers.11/Constant_3_output_0, %/layers.11/Constant_6_output_0)
%/layers.11/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_4_output_0 = Add(%/layers.11/Slice_output_0, %/layers.11/Constant_7_output_0)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_4_output_0, %onnx::Conv_1040, %onnx::Conv_1041)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%876 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %876
}
|
val_accuracy
| 91.316104
| 598,096,000
| 1,969,346
|
{'zcp_epe_nas': 74.57080475442092, 'zcp_fisher': 31.943143844604492, 'zcp_flops': 9569536000.0, 'zcp_grad_norm': 114.36219024658203, 'zcp_grasp': 32.879638671875, 'zcp_jacov': -16.070695194192623, 'zcp_l2_norm': 886.1952514648438, 'zcp_nwot': 218.97406794087934, 'zcp_params': 1969346.0, 'zcp_plain': -0.013767983764410002, 'zcp_snip': 536.1673583984375, 'zcp_synflow': 122.11645533848115, 'zcp_zen': 78.72198486328125, 'zcp_val_accuracy': 0.9183694124221801}
| |
NASBench101_87069
|
NASBench101
|
87069
|
34bf49f3e041514c227cb6e929401688
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_905[FLOAT, 128x3x3x3]
%onnx::Conv_906[FLOAT, 128]
%onnx::Conv_908[FLOAT, 43x128x1x1]
%onnx::Conv_909[FLOAT, 43]
%onnx::Conv_911[FLOAT, 43x43x3x3]
%onnx::Conv_914[FLOAT, 43x43x3x3]
%onnx::Conv_917[FLOAT, 43x43x3x3]
%onnx::Conv_920[FLOAT, 42x128x1x1]
%onnx::Conv_921[FLOAT, 42]
%onnx::Conv_923[FLOAT, 42x42x1x1]
%onnx::Conv_926[FLOAT, 43x128x1x1]
%onnx::Conv_929[FLOAT, 43x43x3x3]
%onnx::Conv_932[FLOAT, 43x43x3x3]
%onnx::Conv_935[FLOAT, 43x43x3x3]
%onnx::Conv_938[FLOAT, 42x128x1x1]
%onnx::Conv_941[FLOAT, 42x42x1x1]
%onnx::Conv_944[FLOAT, 43x128x1x1]
%onnx::Conv_947[FLOAT, 43x43x3x3]
%onnx::Conv_950[FLOAT, 43x43x3x3]
%onnx::Conv_953[FLOAT, 43x43x3x3]
%onnx::Conv_956[FLOAT, 42x128x1x1]
%onnx::Conv_959[FLOAT, 42x42x1x1]
%onnx::Conv_962[FLOAT, 86x128x1x1]
%onnx::Conv_963[FLOAT, 86]
%onnx::Conv_965[FLOAT, 86x86x3x3]
%onnx::Conv_968[FLOAT, 86x86x3x3]
%onnx::Conv_971[FLOAT, 85x85x3x3]
%onnx::Conv_972[FLOAT, 85]
%onnx::Conv_974[FLOAT, 85x128x1x1]
%onnx::Conv_977[FLOAT, 85x85x1x1]
%onnx::Conv_980[FLOAT, 86x256x1x1]
%onnx::Conv_983[FLOAT, 86x86x3x3]
%onnx::Conv_986[FLOAT, 86x86x3x3]
%onnx::Conv_989[FLOAT, 85x85x3x3]
%onnx::Conv_992[FLOAT, 85x256x1x1]
%onnx::Conv_995[FLOAT, 85x85x1x1]
%onnx::Conv_998[FLOAT, 86x256x1x1]
%onnx::Conv_1001[FLOAT, 86x86x3x3]
%onnx::Conv_1004[FLOAT, 86x86x3x3]
%onnx::Conv_1007[FLOAT, 85x85x3x3]
%onnx::Conv_1010[FLOAT, 85x256x1x1]
%onnx::Conv_1013[FLOAT, 85x85x1x1]
%onnx::Conv_1016[FLOAT, 171x256x1x1]
%onnx::Conv_1017[FLOAT, 171]
%onnx::Conv_1019[FLOAT, 171x171x3x3]
%onnx::Conv_1022[FLOAT, 171x171x3x3]
%onnx::Conv_1025[FLOAT, 171x171x3x3]
%onnx::Conv_1028[FLOAT, 170x256x1x1]
%onnx::Conv_1029[FLOAT, 170]
%onnx::Conv_1031[FLOAT, 170x170x1x1]
%onnx::Conv_1034[FLOAT, 171x512x1x1]
%onnx::Conv_1037[FLOAT, 171x171x3x3]
%onnx::Conv_1040[FLOAT, 171x171x3x3]
%onnx::Conv_1043[FLOAT, 171x171x3x3]
%onnx::Conv_1046[FLOAT, 170x512x1x1]
%onnx::Conv_1049[FLOAT, 170x170x1x1]
%onnx::Conv_1052[FLOAT, 171x512x1x1]
%onnx::Conv_1055[FLOAT, 171x171x3x3]
%onnx::Conv_1058[FLOAT, 171x171x3x3]
%onnx::Conv_1061[FLOAT, 171x171x3x3]
%onnx::Conv_1064[FLOAT, 170x512x1x1]
%onnx::Conv_1067[FLOAT, 170x170x1x1]
) {
%onnx::Conv_1068 = Identity(%onnx::Conv_1029)
%onnx::Conv_1065 = Identity(%onnx::Conv_1029)
%onnx::Conv_1062 = Identity(%onnx::Conv_1017)
%onnx::Conv_1059 = Identity(%onnx::Conv_1017)
%onnx::Conv_1056 = Identity(%onnx::Conv_1017)
%onnx::Conv_1053 = Identity(%onnx::Conv_1017)
%onnx::Conv_1050 = Identity(%onnx::Conv_1029)
%onnx::Conv_1047 = Identity(%onnx::Conv_1029)
%onnx::Conv_1044 = Identity(%onnx::Conv_1017)
%onnx::Conv_1041 = Identity(%onnx::Conv_1017)
%onnx::Conv_1038 = Identity(%onnx::Conv_1017)
%onnx::Conv_1035 = Identity(%onnx::Conv_1017)
%onnx::Conv_1032 = Identity(%onnx::Conv_1029)
%onnx::Conv_1026 = Identity(%onnx::Conv_1017)
%onnx::Conv_1023 = Identity(%onnx::Conv_1017)
%onnx::Conv_1020 = Identity(%onnx::Conv_1017)
%onnx::Conv_1014 = Identity(%onnx::Conv_972)
%onnx::Conv_1011 = Identity(%onnx::Conv_972)
%onnx::Conv_1008 = Identity(%onnx::Conv_972)
%onnx::Conv_1005 = Identity(%onnx::Conv_963)
%onnx::Conv_1002 = Identity(%onnx::Conv_963)
%onnx::Conv_999 = Identity(%onnx::Conv_963)
%onnx::Conv_996 = Identity(%onnx::Conv_972)
%onnx::Conv_993 = Identity(%onnx::Conv_972)
%onnx::Conv_990 = Identity(%onnx::Conv_972)
%onnx::Conv_987 = Identity(%onnx::Conv_963)
%onnx::Conv_984 = Identity(%onnx::Conv_963)
%onnx::Conv_981 = Identity(%onnx::Conv_963)
%onnx::Conv_978 = Identity(%onnx::Conv_972)
%onnx::Conv_975 = Identity(%onnx::Conv_972)
%onnx::Conv_969 = Identity(%onnx::Conv_963)
%onnx::Conv_966 = Identity(%onnx::Conv_963)
%onnx::Conv_960 = Identity(%onnx::Conv_921)
%onnx::Conv_957 = Identity(%onnx::Conv_921)
%onnx::Conv_954 = Identity(%onnx::Conv_909)
%onnx::Conv_951 = Identity(%onnx::Conv_909)
%onnx::Conv_948 = Identity(%onnx::Conv_909)
%onnx::Conv_945 = Identity(%onnx::Conv_909)
%onnx::Conv_942 = Identity(%onnx::Conv_921)
%onnx::Conv_939 = Identity(%onnx::Conv_921)
%onnx::Conv_936 = Identity(%onnx::Conv_909)
%onnx::Conv_933 = Identity(%onnx::Conv_909)
%onnx::Conv_930 = Identity(%onnx::Conv_909)
%onnx::Conv_927 = Identity(%onnx::Conv_909)
%onnx::Conv_924 = Identity(%onnx::Conv_921)
%onnx::Conv_918 = Identity(%onnx::Conv_909)
%onnx::Conv_915 = Identity(%onnx::Conv_909)
%onnx::Conv_912 = Identity(%onnx::Conv_909)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_905, %onnx::Conv_906)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_908, %onnx::Conv_909)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_911, %onnx::Conv_912)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_914, %onnx::Conv_915)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_917, %onnx::Conv_918)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.1/Slice_output_0 = Slice(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_4_output_0, %/layers.1/Constant_5_output_0, %/layers.1/Constant_3_output_0, %/layers.1/Constant_6_output_0)
%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_920, %onnx::Conv_921)
%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/Slice_output_0, %/layers.1/Constant_7_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/Add_3_output_0, %/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_923, %onnx::Conv_924)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.3/maxpool/MaxPool_output_0, %/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_926, %onnx::Conv_927)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_929, %onnx::Conv_930)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_932, %onnx::Conv_933)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_935, %onnx::Conv_936)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.2/Slice_output_0 = Slice(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_4_output_0, %/layers.2/Constant_5_output_0, %/layers.2/Constant_3_output_0, %/layers.2/Constant_6_output_0)
%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_938, %onnx::Conv_939)
%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/Slice_output_0, %/layers.2/Constant_7_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/Add_3_output_0, %/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_941, %onnx::Conv_942)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.3/maxpool/MaxPool_output_0, %/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_944, %onnx::Conv_945)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_947, %onnx::Conv_948)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_950, %onnx::Conv_951)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_953, %onnx::Conv_954)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.3/Slice_output_0 = Slice(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_4_output_0, %/layers.3/Constant_5_output_0, %/layers.3/Constant_3_output_0, %/layers.3/Constant_6_output_0)
%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_956, %onnx::Conv_957)
%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/Slice_output_0, %/layers.3/Constant_7_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/Add_3_output_0, %/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_4_output_0, %onnx::Conv_959, %onnx::Conv_960)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.3/maxpool/MaxPool_output_0, %/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_962, %onnx::Conv_963)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_965, %onnx::Conv_966)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_968, %onnx::Conv_969)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.5/Slice_output_0 = Slice(%/layers.5/vertex_op.3/maxpool/MaxPool_output_0, %/layers.5/Constant_4_output_0, %/layers.5/Constant_5_output_0, %/layers.5/Constant_3_output_0, %/layers.5/Constant_6_output_0)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Slice_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_974, %onnx::Conv_975)
%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_7_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/Add_3_output_0, %/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_977, %onnx::Conv_978)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.3/maxpool/MaxPool_output_0, %/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.6/Slice_output_0 = Slice(%/layers.6/vertex_op.3/maxpool/MaxPool_output_0, %/layers.6/Constant_4_output_0, %/layers.6/Constant_5_output_0, %/layers.6/Constant_3_output_0, %/layers.6/Constant_6_output_0)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Slice_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_7_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/Add_3_output_0, %/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.3/maxpool/MaxPool_output_0, %/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.7/Slice_output_0 = Slice(%/layers.7/vertex_op.3/maxpool/MaxPool_output_0, %/layers.7/Constant_4_output_0, %/layers.7/Constant_5_output_0, %/layers.7/Constant_3_output_0, %/layers.7/Constant_6_output_0)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Slice_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_7_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/Add_3_output_0, %/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_4_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.3/maxpool/MaxPool_output_0, %/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_1025, %onnx::Conv_1026)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.9/Slice_output_0 = Slice(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_4_output_0, %/layers.9/Constant_5_output_0, %/layers.9/Constant_3_output_0, %/layers.9/Constant_6_output_0)
%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1028, %onnx::Conv_1029)
%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/Slice_output_0, %/layers.9/Constant_7_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/Add_3_output_0, %/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_1031, %onnx::Conv_1032)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.3/maxpool/MaxPool_output_0, %/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_1034, %onnx::Conv_1035)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_1037, %onnx::Conv_1038)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_1040, %onnx::Conv_1041)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_1043, %onnx::Conv_1044)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.10/Slice_output_0 = Slice(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_4_output_0, %/layers.10/Constant_5_output_0, %/layers.10/Constant_3_output_0, %/layers.10/Constant_6_output_0)
%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_1046, %onnx::Conv_1047)
%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/Slice_output_0, %/layers.10/Constant_7_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/Add_3_output_0, %/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_1049, %onnx::Conv_1050)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.3/maxpool/MaxPool_output_0, %/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1052, %onnx::Conv_1053)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_1055, %onnx::Conv_1056)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_1058, %onnx::Conv_1059)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_1061, %onnx::Conv_1062)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.11/Slice_output_0 = Slice(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_4_output_0, %/layers.11/Constant_5_output_0, %/layers.11/Constant_3_output_0, %/layers.11/Constant_6_output_0)
%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1064, %onnx::Conv_1065)
%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_7_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/Slice_output_0, %/layers.11/Constant_7_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/Add_3_output_0, %/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_4_output_0, %onnx::Conv_1067, %onnx::Conv_1068)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.3/maxpool/MaxPool_output_0, %/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%903 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %903
}
|
val_accuracy
| 91.736782
| 1,145,196,160
| 3,824,576
|
{'zcp_epe_nas': 109.34919604260844, 'zcp_fisher': 844.2882690429688, 'zcp_flops': 18323138560.0, 'zcp_grad_norm': 470.1122131347656, 'zcp_grasp': 235.9375, 'zcp_jacov': -16.05010328647952, 'zcp_l2_norm': 884.7759399414062, 'zcp_nwot': 218.2252591742358, 'zcp_params': 3824576.0, 'zcp_plain': -0.009550760500133001, 'zcp_snip': 2295.290771484375, 'zcp_synflow': 141.70967992893168, 'zcp_zen': 100.83678436279297, 'zcp_val_accuracy': 0.9261819124221801}
| |
NASBench101_73801
|
NASBench101
|
73801
|
2cc3fda60daf8ba0e0765135f2c37554
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_977[FLOAT, 128x3x3x3]
%onnx::Conv_978[FLOAT, 128]
%onnx::Conv_980[FLOAT, 128x128x1x1]
%onnx::Conv_983[FLOAT, 128x128x1x1]
%onnx::Conv_986[FLOAT, 128x128x1x1]
%onnx::Conv_989[FLOAT, 128x128x3x3]
%onnx::Conv_992[FLOAT, 128x128x1x1]
%onnx::Conv_995[FLOAT, 128x128x1x1]
%onnx::Conv_998[FLOAT, 128x128x1x1]
%onnx::Conv_1001[FLOAT, 128x128x1x1]
%onnx::Conv_1004[FLOAT, 128x128x1x1]
%onnx::Conv_1007[FLOAT, 128x128x1x1]
%onnx::Conv_1010[FLOAT, 128x128x3x3]
%onnx::Conv_1013[FLOAT, 128x128x1x1]
%onnx::Conv_1016[FLOAT, 128x128x1x1]
%onnx::Conv_1019[FLOAT, 128x128x1x1]
%onnx::Conv_1022[FLOAT, 128x128x1x1]
%onnx::Conv_1025[FLOAT, 128x128x1x1]
%onnx::Conv_1028[FLOAT, 128x128x1x1]
%onnx::Conv_1031[FLOAT, 128x128x3x3]
%onnx::Conv_1034[FLOAT, 128x128x1x1]
%onnx::Conv_1037[FLOAT, 128x128x1x1]
%onnx::Conv_1040[FLOAT, 128x128x1x1]
%onnx::Conv_1043[FLOAT, 256x128x1x1]
%onnx::Conv_1044[FLOAT, 256]
%onnx::Conv_1046[FLOAT, 256x256x1x1]
%onnx::Conv_1049[FLOAT, 256x128x1x1]
%onnx::Conv_1052[FLOAT, 256x256x3x3]
%onnx::Conv_1055[FLOAT, 256x256x1x1]
%onnx::Conv_1058[FLOAT, 256x256x1x1]
%onnx::Conv_1061[FLOAT, 256x256x1x1]
%onnx::Conv_1064[FLOAT, 256x256x1x1]
%onnx::Conv_1067[FLOAT, 256x256x1x1]
%onnx::Conv_1070[FLOAT, 256x256x1x1]
%onnx::Conv_1073[FLOAT, 256x256x3x3]
%onnx::Conv_1076[FLOAT, 256x256x1x1]
%onnx::Conv_1079[FLOAT, 256x256x1x1]
%onnx::Conv_1082[FLOAT, 256x256x1x1]
%onnx::Conv_1085[FLOAT, 256x256x1x1]
%onnx::Conv_1088[FLOAT, 256x256x1x1]
%onnx::Conv_1091[FLOAT, 256x256x1x1]
%onnx::Conv_1094[FLOAT, 256x256x3x3]
%onnx::Conv_1097[FLOAT, 256x256x1x1]
%onnx::Conv_1100[FLOAT, 256x256x1x1]
%onnx::Conv_1103[FLOAT, 256x256x1x1]
%onnx::Conv_1106[FLOAT, 512x256x1x1]
%onnx::Conv_1107[FLOAT, 512]
%onnx::Conv_1109[FLOAT, 512x512x1x1]
%onnx::Conv_1112[FLOAT, 512x256x1x1]
%onnx::Conv_1115[FLOAT, 512x512x3x3]
%onnx::Conv_1118[FLOAT, 512x512x1x1]
%onnx::Conv_1121[FLOAT, 512x512x1x1]
%onnx::Conv_1124[FLOAT, 512x512x1x1]
%onnx::Conv_1127[FLOAT, 512x512x1x1]
%onnx::Conv_1130[FLOAT, 512x512x1x1]
%onnx::Conv_1133[FLOAT, 512x512x1x1]
%onnx::Conv_1136[FLOAT, 512x512x3x3]
%onnx::Conv_1139[FLOAT, 512x512x1x1]
%onnx::Conv_1142[FLOAT, 512x512x1x1]
%onnx::Conv_1145[FLOAT, 512x512x1x1]
%onnx::Conv_1148[FLOAT, 512x512x1x1]
%onnx::Conv_1151[FLOAT, 512x512x1x1]
%onnx::Conv_1154[FLOAT, 512x512x1x1]
%onnx::Conv_1157[FLOAT, 512x512x3x3]
%onnx::Conv_1160[FLOAT, 512x512x1x1]
%onnx::Conv_1163[FLOAT, 512x512x1x1]
%onnx::Conv_1166[FLOAT, 512x512x1x1]
) {
%onnx::Conv_1167 = Identity(%onnx::Conv_1107)
%onnx::Conv_1164 = Identity(%onnx::Conv_1107)
%onnx::Conv_1161 = Identity(%onnx::Conv_1107)
%onnx::Conv_1158 = Identity(%onnx::Conv_1107)
%onnx::Conv_1155 = Identity(%onnx::Conv_1107)
%onnx::Conv_1152 = Identity(%onnx::Conv_1107)
%onnx::Conv_1149 = Identity(%onnx::Conv_1107)
%onnx::Conv_1146 = Identity(%onnx::Conv_1107)
%onnx::Conv_1143 = Identity(%onnx::Conv_1107)
%onnx::Conv_1140 = Identity(%onnx::Conv_1107)
%onnx::Conv_1137 = Identity(%onnx::Conv_1107)
%onnx::Conv_1134 = Identity(%onnx::Conv_1107)
%onnx::Conv_1131 = Identity(%onnx::Conv_1107)
%onnx::Conv_1128 = Identity(%onnx::Conv_1107)
%onnx::Conv_1125 = Identity(%onnx::Conv_1107)
%onnx::Conv_1122 = Identity(%onnx::Conv_1107)
%onnx::Conv_1119 = Identity(%onnx::Conv_1107)
%onnx::Conv_1116 = Identity(%onnx::Conv_1107)
%onnx::Conv_1113 = Identity(%onnx::Conv_1107)
%onnx::Conv_1110 = Identity(%onnx::Conv_1107)
%onnx::Conv_1104 = Identity(%onnx::Conv_1044)
%onnx::Conv_1101 = Identity(%onnx::Conv_1044)
%onnx::Conv_1098 = Identity(%onnx::Conv_1044)
%onnx::Conv_1095 = Identity(%onnx::Conv_1044)
%onnx::Conv_1092 = Identity(%onnx::Conv_1044)
%onnx::Conv_1089 = Identity(%onnx::Conv_1044)
%onnx::Conv_1086 = Identity(%onnx::Conv_1044)
%onnx::Conv_1083 = Identity(%onnx::Conv_1044)
%onnx::Conv_1080 = Identity(%onnx::Conv_1044)
%onnx::Conv_1077 = Identity(%onnx::Conv_1044)
%onnx::Conv_1074 = Identity(%onnx::Conv_1044)
%onnx::Conv_1071 = Identity(%onnx::Conv_1044)
%onnx::Conv_1068 = Identity(%onnx::Conv_1044)
%onnx::Conv_1065 = Identity(%onnx::Conv_1044)
%onnx::Conv_1062 = Identity(%onnx::Conv_1044)
%onnx::Conv_1059 = Identity(%onnx::Conv_1044)
%onnx::Conv_1056 = Identity(%onnx::Conv_1044)
%onnx::Conv_1053 = Identity(%onnx::Conv_1044)
%onnx::Conv_1050 = Identity(%onnx::Conv_1044)
%onnx::Conv_1047 = Identity(%onnx::Conv_1044)
%onnx::Conv_1041 = Identity(%onnx::Conv_978)
%onnx::Conv_1038 = Identity(%onnx::Conv_978)
%onnx::Conv_1035 = Identity(%onnx::Conv_978)
%onnx::Conv_1032 = Identity(%onnx::Conv_978)
%onnx::Conv_1029 = Identity(%onnx::Conv_978)
%onnx::Conv_1026 = Identity(%onnx::Conv_978)
%onnx::Conv_1023 = Identity(%onnx::Conv_978)
%onnx::Conv_1020 = Identity(%onnx::Conv_978)
%onnx::Conv_1017 = Identity(%onnx::Conv_978)
%onnx::Conv_1014 = Identity(%onnx::Conv_978)
%onnx::Conv_1011 = Identity(%onnx::Conv_978)
%onnx::Conv_1008 = Identity(%onnx::Conv_978)
%onnx::Conv_1005 = Identity(%onnx::Conv_978)
%onnx::Conv_1002 = Identity(%onnx::Conv_978)
%onnx::Conv_999 = Identity(%onnx::Conv_978)
%onnx::Conv_996 = Identity(%onnx::Conv_978)
%onnx::Conv_993 = Identity(%onnx::Conv_978)
%onnx::Conv_990 = Identity(%onnx::Conv_978)
%onnx::Conv_987 = Identity(%onnx::Conv_978)
%onnx::Conv_984 = Identity(%onnx::Conv_978)
%onnx::Conv_981 = Identity(%onnx::Conv_978)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_977, %onnx::Conv_978)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/Add_1_output_0, %/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/Add_3_output_0, %/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_5_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/Add_6_output_0 = Add(%/layers.1/Add_5_output_0, %/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_6_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_7_output_0 = Add(%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_4_output_0)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_7_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/Add_1_output_0, %/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/Add_3_output_0, %/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_5_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/Add_6_output_0 = Add(%/layers.2/Add_5_output_0, %/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_6_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_7_output_0 = Add(%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_4_output_0)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_7_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_1025, %onnx::Conv_1026)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1028, %onnx::Conv_1029)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/Add_1_output_0, %/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_1031, %onnx::Conv_1032)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/Add_3_output_0, %/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_4_output_0, %onnx::Conv_1034, %onnx::Conv_1035)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_5_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/Add_6_output_0 = Add(%/layers.3/Add_5_output_0, %/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_6_output_0, %onnx::Conv_1037, %onnx::Conv_1038)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_7_output_0 = Add(%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_4_output_0)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_7_output_0, %onnx::Conv_1040, %onnx::Conv_1041)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1043, %onnx::Conv_1044)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_1046, %onnx::Conv_1047)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1049, %onnx::Conv_1050)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/Add_1_output_0, %/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_1052, %onnx::Conv_1053)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/Add_3_output_0, %/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_1055, %onnx::Conv_1056)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_5_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/Add_6_output_0 = Add(%/layers.5/Add_5_output_0, %/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_6_output_0, %onnx::Conv_1058, %onnx::Conv_1059)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_7_output_0 = Add(%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_4_output_0)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_7_output_0, %onnx::Conv_1061, %onnx::Conv_1062)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1064, %onnx::Conv_1065)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_1067, %onnx::Conv_1068)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1070, %onnx::Conv_1071)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/Add_1_output_0, %/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_1073, %onnx::Conv_1074)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/Add_3_output_0, %/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_1076, %onnx::Conv_1077)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_5_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/Add_6_output_0 = Add(%/layers.6/Add_5_output_0, %/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_6_output_0, %onnx::Conv_1079, %onnx::Conv_1080)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_7_output_0 = Add(%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_4_output_0)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_7_output_0, %onnx::Conv_1082, %onnx::Conv_1083)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1085, %onnx::Conv_1086)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_1088, %onnx::Conv_1089)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1091, %onnx::Conv_1092)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/Add_1_output_0, %/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_1094, %onnx::Conv_1095)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/Add_3_output_0, %/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_4_output_0, %onnx::Conv_1097, %onnx::Conv_1098)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_5_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/Add_6_output_0 = Add(%/layers.7/Add_5_output_0, %/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_6_output_0, %onnx::Conv_1100, %onnx::Conv_1101)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_7_output_0 = Add(%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_4_output_0)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_7_output_0, %onnx::Conv_1103, %onnx::Conv_1104)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1106, %onnx::Conv_1107)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_1109, %onnx::Conv_1110)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1112, %onnx::Conv_1113)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/Add_1_output_0, %/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_1115, %onnx::Conv_1116)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/Add_3_output_0, %/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_1118, %onnx::Conv_1119)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_5_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/Add_6_output_0 = Add(%/layers.9/Add_5_output_0, %/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_6_output_0, %onnx::Conv_1121, %onnx::Conv_1122)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_7_output_0 = Add(%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_4_output_0)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_7_output_0, %onnx::Conv_1124, %onnx::Conv_1125)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1127, %onnx::Conv_1128)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_1130, %onnx::Conv_1131)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1133, %onnx::Conv_1134)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/Add_1_output_0, %/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_1136, %onnx::Conv_1137)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/Add_3_output_0, %/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_1139, %onnx::Conv_1140)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_5_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/Add_6_output_0 = Add(%/layers.10/Add_5_output_0, %/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_6_output_0, %onnx::Conv_1142, %onnx::Conv_1143)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_7_output_0 = Add(%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_4_output_0)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_7_output_0, %onnx::Conv_1145, %onnx::Conv_1146)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1148, %onnx::Conv_1149)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_1151, %onnx::Conv_1152)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1154, %onnx::Conv_1155)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/Add_1_output_0, %/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_1157, %onnx::Conv_1158)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/Add_3_output_0, %/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_4_output_0, %onnx::Conv_1160, %onnx::Conv_1161)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_5_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/Add_6_output_0 = Add(%/layers.11/Add_5_output_0, %/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_6_output_0, %onnx::Conv_1163, %onnx::Conv_1164)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_7_output_0 = Add(%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_4_output_0)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_7_output_0, %onnx::Conv_1166, %onnx::Conv_1167)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%975 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %975
}
|
val_accuracy
| 89.222759
| 4,509,411,328
| 15,201,674
|
{'zcp_epe_nas': 121.64504548909063, 'zcp_fisher': 1031.284912109375, 'zcp_flops': 72150581248.0, 'zcp_grad_norm': 663.186767578125, 'zcp_grasp': 5116.5859375, 'zcp_jacov': -16.035986488651726, 'zcp_l2_norm': 1453.691650390625, 'zcp_nwot': 237.8393240986465, 'zcp_params': 15201674.0, 'zcp_plain': 0.005963878240436001, 'zcp_snip': 4489.52587890625, 'zcp_synflow': 172.52870402302665, 'zcp_zen': 118.60411834716797, 'zcp_val_accuracy': 0.931490361690521}
| |
NASBench101_99351
|
NASBench101
|
99351
|
3c2b1c86969a9177477953b2e09cdb02
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_770[FLOAT, 128x3x3x3]
%onnx::Conv_771[FLOAT, 128]
%onnx::Conv_773[FLOAT, 64x128x1x1]
%onnx::Conv_774[FLOAT, 64]
%onnx::Conv_776[FLOAT, 64x64x3x3]
%onnx::Conv_779[FLOAT, 64x128x1x1]
%onnx::Conv_782[FLOAT, 64x64x1x1]
%onnx::Conv_785[FLOAT, 64x64x1x1]
%onnx::Conv_788[FLOAT, 64x128x1x1]
%onnx::Conv_791[FLOAT, 64x64x3x3]
%onnx::Conv_794[FLOAT, 64x128x1x1]
%onnx::Conv_797[FLOAT, 64x64x1x1]
%onnx::Conv_800[FLOAT, 64x64x1x1]
%onnx::Conv_803[FLOAT, 64x128x1x1]
%onnx::Conv_806[FLOAT, 64x64x3x3]
%onnx::Conv_809[FLOAT, 64x128x1x1]
%onnx::Conv_812[FLOAT, 64x64x1x1]
%onnx::Conv_815[FLOAT, 64x64x1x1]
%onnx::Conv_818[FLOAT, 128x128x1x1]
%onnx::Conv_821[FLOAT, 128x128x3x3]
%onnx::Conv_824[FLOAT, 128x128x1x1]
%onnx::Conv_827[FLOAT, 128x128x1x1]
%onnx::Conv_830[FLOAT, 128x128x1x1]
%onnx::Conv_833[FLOAT, 128x256x1x1]
%onnx::Conv_836[FLOAT, 128x128x3x3]
%onnx::Conv_839[FLOAT, 128x256x1x1]
%onnx::Conv_842[FLOAT, 128x128x1x1]
%onnx::Conv_845[FLOAT, 128x128x1x1]
%onnx::Conv_848[FLOAT, 128x256x1x1]
%onnx::Conv_851[FLOAT, 128x128x3x3]
%onnx::Conv_854[FLOAT, 128x256x1x1]
%onnx::Conv_857[FLOAT, 128x128x1x1]
%onnx::Conv_860[FLOAT, 128x128x1x1]
%onnx::Conv_863[FLOAT, 256x256x1x1]
%onnx::Conv_864[FLOAT, 256]
%onnx::Conv_866[FLOAT, 256x256x3x3]
%onnx::Conv_869[FLOAT, 256x256x1x1]
%onnx::Conv_872[FLOAT, 256x256x1x1]
%onnx::Conv_875[FLOAT, 256x256x1x1]
%onnx::Conv_878[FLOAT, 256x512x1x1]
%onnx::Conv_881[FLOAT, 256x256x3x3]
%onnx::Conv_884[FLOAT, 256x512x1x1]
%onnx::Conv_887[FLOAT, 256x256x1x1]
%onnx::Conv_890[FLOAT, 256x256x1x1]
%onnx::Conv_893[FLOAT, 256x512x1x1]
%onnx::Conv_896[FLOAT, 256x256x3x3]
%onnx::Conv_899[FLOAT, 256x512x1x1]
%onnx::Conv_902[FLOAT, 256x256x1x1]
%onnx::Conv_905[FLOAT, 256x256x1x1]
) {
%onnx::Conv_906 = Identity(%onnx::Conv_864)
%onnx::Conv_903 = Identity(%onnx::Conv_864)
%onnx::Conv_900 = Identity(%onnx::Conv_864)
%onnx::Conv_897 = Identity(%onnx::Conv_864)
%onnx::Conv_894 = Identity(%onnx::Conv_864)
%onnx::Conv_891 = Identity(%onnx::Conv_864)
%onnx::Conv_888 = Identity(%onnx::Conv_864)
%onnx::Conv_885 = Identity(%onnx::Conv_864)
%onnx::Conv_882 = Identity(%onnx::Conv_864)
%onnx::Conv_879 = Identity(%onnx::Conv_864)
%onnx::Conv_876 = Identity(%onnx::Conv_864)
%onnx::Conv_873 = Identity(%onnx::Conv_864)
%onnx::Conv_870 = Identity(%onnx::Conv_864)
%onnx::Conv_867 = Identity(%onnx::Conv_864)
%onnx::Conv_861 = Identity(%onnx::Conv_771)
%onnx::Conv_858 = Identity(%onnx::Conv_771)
%onnx::Conv_855 = Identity(%onnx::Conv_771)
%onnx::Conv_852 = Identity(%onnx::Conv_771)
%onnx::Conv_849 = Identity(%onnx::Conv_771)
%onnx::Conv_846 = Identity(%onnx::Conv_771)
%onnx::Conv_843 = Identity(%onnx::Conv_771)
%onnx::Conv_840 = Identity(%onnx::Conv_771)
%onnx::Conv_837 = Identity(%onnx::Conv_771)
%onnx::Conv_834 = Identity(%onnx::Conv_771)
%onnx::Conv_831 = Identity(%onnx::Conv_771)
%onnx::Conv_828 = Identity(%onnx::Conv_771)
%onnx::Conv_825 = Identity(%onnx::Conv_771)
%onnx::Conv_822 = Identity(%onnx::Conv_771)
%onnx::Conv_819 = Identity(%onnx::Conv_771)
%onnx::Conv_816 = Identity(%onnx::Conv_774)
%onnx::Conv_813 = Identity(%onnx::Conv_774)
%onnx::Conv_810 = Identity(%onnx::Conv_774)
%onnx::Conv_807 = Identity(%onnx::Conv_774)
%onnx::Conv_804 = Identity(%onnx::Conv_774)
%onnx::Conv_801 = Identity(%onnx::Conv_774)
%onnx::Conv_798 = Identity(%onnx::Conv_774)
%onnx::Conv_795 = Identity(%onnx::Conv_774)
%onnx::Conv_792 = Identity(%onnx::Conv_774)
%onnx::Conv_789 = Identity(%onnx::Conv_774)
%onnx::Conv_786 = Identity(%onnx::Conv_774)
%onnx::Conv_783 = Identity(%onnx::Conv_774)
%onnx::Conv_780 = Identity(%onnx::Conv_774)
%onnx::Conv_777 = Identity(%onnx::Conv_774)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_770, %onnx::Conv_771)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_773, %onnx::Conv_774)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_776, %onnx::Conv_777)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_779, %onnx::Conv_780)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/Add_1_output_0, %/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_782, %onnx::Conv_783)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_785, %onnx::Conv_786)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_4_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_788, %onnx::Conv_789)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/Add_1_output_0, %/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_797, %onnx::Conv_798)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_800, %onnx::Conv_801)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_4_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_803, %onnx::Conv_804)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_806, %onnx::Conv_807)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_809, %onnx::Conv_810)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/Add_1_output_0, %/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_812, %onnx::Conv_813)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_815, %onnx::Conv_816)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_4_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_818, %onnx::Conv_819)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_821, %onnx::Conv_822)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_824, %onnx::Conv_825)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/Add_1_output_0, %/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_827, %onnx::Conv_828)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_830, %onnx::Conv_831)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_4_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_833, %onnx::Conv_834)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_836, %onnx::Conv_837)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_839, %onnx::Conv_840)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/Add_1_output_0, %/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_842, %onnx::Conv_843)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_845, %onnx::Conv_846)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_4_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_848, %onnx::Conv_849)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_851, %onnx::Conv_852)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_854, %onnx::Conv_855)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/Add_1_output_0, %/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_857, %onnx::Conv_858)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_4_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/Add_1_output_0, %/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_4_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/Add_1_output_0, %/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_4_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/Add_1_output_0, %/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_4_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.5/maxpool/MaxPool_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%768 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %768
}
|
val_accuracy
| 90.604967
| 1,120,806,912
| 3,729,162
|
{'zcp_epe_nas': 73.68162110578385, 'zcp_fisher': 62.448753356933594, 'zcp_flops': 17932910592.0, 'zcp_grad_norm': 129.97219848632812, 'zcp_grasp': -28.440673828125, 'zcp_jacov': -16.056800438540108, 'zcp_l2_norm': 843.9127807617188, 'zcp_nwot': 221.48643534981878, 'zcp_params': 3729162.0, 'zcp_plain': 0.0011754955630740001, 'zcp_snip': 792.48828125, 'zcp_synflow': 90.44926920202991, 'zcp_zen': 79.3829345703125, 'zcp_val_accuracy': 0.8840144276618951}
| |
NASBench101_288450
|
NASBench101
|
288450
|
aea0e408dde714a9afcc0e880c45083e
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_851[FLOAT, 128x3x3x3]
%onnx::Conv_852[FLOAT, 128]
%onnx::Conv_854[FLOAT, 64x128x1x1]
%onnx::Conv_855[FLOAT, 64]
%onnx::Conv_857[FLOAT, 64x64x1x1]
%onnx::Conv_860[FLOAT, 64x128x1x1]
%onnx::Conv_863[FLOAT, 64x64x1x1]
%onnx::Conv_866[FLOAT, 64x128x1x1]
%onnx::Conv_869[FLOAT, 64x64x1x1]
%onnx::Conv_872[FLOAT, 64x128x1x1]
%onnx::Conv_875[FLOAT, 64x64x1x1]
%onnx::Conv_878[FLOAT, 64x128x1x1]
%onnx::Conv_881[FLOAT, 64x64x1x1]
%onnx::Conv_884[FLOAT, 64x128x1x1]
%onnx::Conv_887[FLOAT, 64x64x1x1]
%onnx::Conv_890[FLOAT, 64x128x1x1]
%onnx::Conv_893[FLOAT, 64x64x1x1]
%onnx::Conv_896[FLOAT, 64x128x1x1]
%onnx::Conv_899[FLOAT, 64x64x1x1]
%onnx::Conv_902[FLOAT, 64x128x1x1]
%onnx::Conv_905[FLOAT, 64x64x1x1]
%onnx::Conv_908[FLOAT, 128x128x1x1]
%onnx::Conv_911[FLOAT, 128x128x1x1]
%onnx::Conv_914[FLOAT, 128x128x1x1]
%onnx::Conv_917[FLOAT, 128x128x1x1]
%onnx::Conv_920[FLOAT, 128x128x1x1]
%onnx::Conv_923[FLOAT, 128x128x1x1]
%onnx::Conv_926[FLOAT, 128x256x1x1]
%onnx::Conv_929[FLOAT, 128x128x1x1]
%onnx::Conv_932[FLOAT, 128x256x1x1]
%onnx::Conv_935[FLOAT, 128x128x1x1]
%onnx::Conv_938[FLOAT, 128x256x1x1]
%onnx::Conv_941[FLOAT, 128x128x1x1]
%onnx::Conv_944[FLOAT, 128x256x1x1]
%onnx::Conv_947[FLOAT, 128x128x1x1]
%onnx::Conv_950[FLOAT, 128x256x1x1]
%onnx::Conv_953[FLOAT, 128x128x1x1]
%onnx::Conv_956[FLOAT, 128x256x1x1]
%onnx::Conv_959[FLOAT, 128x128x1x1]
%onnx::Conv_962[FLOAT, 256x256x1x1]
%onnx::Conv_963[FLOAT, 256]
%onnx::Conv_965[FLOAT, 256x256x1x1]
%onnx::Conv_968[FLOAT, 256x256x1x1]
%onnx::Conv_971[FLOAT, 256x256x1x1]
%onnx::Conv_974[FLOAT, 256x256x1x1]
%onnx::Conv_977[FLOAT, 256x256x1x1]
%onnx::Conv_980[FLOAT, 256x512x1x1]
%onnx::Conv_983[FLOAT, 256x256x1x1]
%onnx::Conv_986[FLOAT, 256x512x1x1]
%onnx::Conv_989[FLOAT, 256x256x1x1]
%onnx::Conv_992[FLOAT, 256x512x1x1]
%onnx::Conv_995[FLOAT, 256x256x1x1]
%onnx::Conv_998[FLOAT, 256x512x1x1]
%onnx::Conv_1001[FLOAT, 256x256x1x1]
%onnx::Conv_1004[FLOAT, 256x512x1x1]
%onnx::Conv_1007[FLOAT, 256x256x1x1]
%onnx::Conv_1010[FLOAT, 256x512x1x1]
%onnx::Conv_1013[FLOAT, 256x256x1x1]
) {
%onnx::Conv_1014 = Identity(%onnx::Conv_963)
%onnx::Conv_1011 = Identity(%onnx::Conv_963)
%onnx::Conv_1008 = Identity(%onnx::Conv_963)
%onnx::Conv_1005 = Identity(%onnx::Conv_963)
%onnx::Conv_1002 = Identity(%onnx::Conv_963)
%onnx::Conv_999 = Identity(%onnx::Conv_963)
%onnx::Conv_996 = Identity(%onnx::Conv_963)
%onnx::Conv_993 = Identity(%onnx::Conv_963)
%onnx::Conv_990 = Identity(%onnx::Conv_963)
%onnx::Conv_987 = Identity(%onnx::Conv_963)
%onnx::Conv_984 = Identity(%onnx::Conv_963)
%onnx::Conv_981 = Identity(%onnx::Conv_963)
%onnx::Conv_978 = Identity(%onnx::Conv_963)
%onnx::Conv_975 = Identity(%onnx::Conv_963)
%onnx::Conv_972 = Identity(%onnx::Conv_963)
%onnx::Conv_969 = Identity(%onnx::Conv_963)
%onnx::Conv_966 = Identity(%onnx::Conv_963)
%onnx::Conv_960 = Identity(%onnx::Conv_852)
%onnx::Conv_957 = Identity(%onnx::Conv_852)
%onnx::Conv_954 = Identity(%onnx::Conv_852)
%onnx::Conv_951 = Identity(%onnx::Conv_852)
%onnx::Conv_948 = Identity(%onnx::Conv_852)
%onnx::Conv_945 = Identity(%onnx::Conv_852)
%onnx::Conv_942 = Identity(%onnx::Conv_852)
%onnx::Conv_939 = Identity(%onnx::Conv_852)
%onnx::Conv_936 = Identity(%onnx::Conv_852)
%onnx::Conv_933 = Identity(%onnx::Conv_852)
%onnx::Conv_930 = Identity(%onnx::Conv_852)
%onnx::Conv_927 = Identity(%onnx::Conv_852)
%onnx::Conv_924 = Identity(%onnx::Conv_852)
%onnx::Conv_921 = Identity(%onnx::Conv_852)
%onnx::Conv_918 = Identity(%onnx::Conv_852)
%onnx::Conv_915 = Identity(%onnx::Conv_852)
%onnx::Conv_912 = Identity(%onnx::Conv_852)
%onnx::Conv_909 = Identity(%onnx::Conv_852)
%onnx::Conv_906 = Identity(%onnx::Conv_855)
%onnx::Conv_903 = Identity(%onnx::Conv_855)
%onnx::Conv_900 = Identity(%onnx::Conv_855)
%onnx::Conv_897 = Identity(%onnx::Conv_855)
%onnx::Conv_894 = Identity(%onnx::Conv_855)
%onnx::Conv_891 = Identity(%onnx::Conv_855)
%onnx::Conv_888 = Identity(%onnx::Conv_855)
%onnx::Conv_885 = Identity(%onnx::Conv_855)
%onnx::Conv_882 = Identity(%onnx::Conv_855)
%onnx::Conv_879 = Identity(%onnx::Conv_855)
%onnx::Conv_876 = Identity(%onnx::Conv_855)
%onnx::Conv_873 = Identity(%onnx::Conv_855)
%onnx::Conv_870 = Identity(%onnx::Conv_855)
%onnx::Conv_867 = Identity(%onnx::Conv_855)
%onnx::Conv_864 = Identity(%onnx::Conv_855)
%onnx::Conv_861 = Identity(%onnx::Conv_855)
%onnx::Conv_858 = Identity(%onnx::Conv_855)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_851, %onnx::Conv_852)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_854, %onnx::Conv_855)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_857, %onnx::Conv_858)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/Add_1_output_0, %/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/Add_5_output_0 = Add(%/layers.1/Add_4_output_0, %/layers.1/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_5_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/Add_1_output_0, %/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/Add_5_output_0 = Add(%/layers.2/Add_4_output_0, %/layers.2/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_5_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/Add_1_output_0, %/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/Add_5_output_0 = Add(%/layers.3/Add_4_output_0, %/layers.3/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_5_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_908, %onnx::Conv_909)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_911, %onnx::Conv_912)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_914, %onnx::Conv_915)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/Add_1_output_0, %/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_917, %onnx::Conv_918)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_920, %onnx::Conv_921)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/Add_5_output_0 = Add(%/layers.5/Add_4_output_0, %/layers.5/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_5_output_0, %onnx::Conv_923, %onnx::Conv_924)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_926, %onnx::Conv_927)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_929, %onnx::Conv_930)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_932, %onnx::Conv_933)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/Add_1_output_0, %/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_935, %onnx::Conv_936)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_938, %onnx::Conv_939)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/Add_5_output_0 = Add(%/layers.6/Add_4_output_0, %/layers.6/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_5_output_0, %onnx::Conv_941, %onnx::Conv_942)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_944, %onnx::Conv_945)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_947, %onnx::Conv_948)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_950, %onnx::Conv_951)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/Add_1_output_0, %/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_953, %onnx::Conv_954)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_956, %onnx::Conv_957)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/Add_5_output_0 = Add(%/layers.7/Add_4_output_0, %/layers.7/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_5_output_0, %onnx::Conv_959, %onnx::Conv_960)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_962, %onnx::Conv_963)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_965, %onnx::Conv_966)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_968, %onnx::Conv_969)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/Add_1_output_0, %/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_974, %onnx::Conv_975)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/Add_5_output_0 = Add(%/layers.9/Add_4_output_0, %/layers.9/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_5_output_0, %onnx::Conv_977, %onnx::Conv_978)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/Add_1_output_0, %/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/Add_5_output_0 = Add(%/layers.10/Add_4_output_0, %/layers.10/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_5_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/Add_1_output_0, %/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/Add_5_output_0 = Add(%/layers.11/Add_4_output_0, %/layers.11/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_5_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%849 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %849
}
|
val_accuracy
| 89.853764
| 653,797,376
| 2,101,642
|
{'zcp_epe_nas': 117.99064088210987, 'zcp_fisher': 2.122584342956543, 'zcp_flops': 10460758016.0, 'zcp_grad_norm': 36.14634704589844, 'zcp_grasp': 1.866867065429687, 'zcp_jacov': -16.049159070102135, 'zcp_l2_norm': 1041.2362060546875, 'zcp_nwot': 224.52469138301748, 'zcp_params': 2101642.0, 'zcp_plain': -0.023862350732088002, 'zcp_snip': 210.60421752929688, 'zcp_synflow': 81.30097246461348, 'zcp_zen': 82.25395965576172, 'zcp_val_accuracy': 0.9423077106475831}
| |
NASBench101_147793
|
NASBench101
|
147793
|
5967c9bc70470880155140052412fa1e
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_887[FLOAT, 128x3x3x3]
%onnx::Conv_888[FLOAT, 128]
%onnx::Conv_890[FLOAT, 128x128x1x1]
%onnx::Conv_893[FLOAT, 128x128x3x3]
%onnx::Conv_896[FLOAT, 128x128x1x1]
%onnx::Conv_899[FLOAT, 128x128x3x3]
%onnx::Conv_902[FLOAT, 128x128x3x3]
%onnx::Conv_905[FLOAT, 128x128x3x3]
%onnx::Conv_908[FLOAT, 128x128x1x1]
%onnx::Conv_911[FLOAT, 128x128x3x3]
%onnx::Conv_914[FLOAT, 128x128x1x1]
%onnx::Conv_917[FLOAT, 128x128x3x3]
%onnx::Conv_920[FLOAT, 128x128x3x3]
%onnx::Conv_923[FLOAT, 128x128x3x3]
%onnx::Conv_926[FLOAT, 128x128x1x1]
%onnx::Conv_929[FLOAT, 128x128x3x3]
%onnx::Conv_932[FLOAT, 128x128x1x1]
%onnx::Conv_935[FLOAT, 128x128x3x3]
%onnx::Conv_938[FLOAT, 128x128x3x3]
%onnx::Conv_941[FLOAT, 128x128x3x3]
%onnx::Conv_944[FLOAT, 256x128x1x1]
%onnx::Conv_945[FLOAT, 256]
%onnx::Conv_947[FLOAT, 256x256x3x3]
%onnx::Conv_950[FLOAT, 256x128x1x1]
%onnx::Conv_953[FLOAT, 256x256x3x3]
%onnx::Conv_956[FLOAT, 256x256x3x3]
%onnx::Conv_959[FLOAT, 256x256x3x3]
%onnx::Conv_962[FLOAT, 256x256x1x1]
%onnx::Conv_965[FLOAT, 256x256x3x3]
%onnx::Conv_968[FLOAT, 256x256x1x1]
%onnx::Conv_971[FLOAT, 256x256x3x3]
%onnx::Conv_974[FLOAT, 256x256x3x3]
%onnx::Conv_977[FLOAT, 256x256x3x3]
%onnx::Conv_980[FLOAT, 256x256x1x1]
%onnx::Conv_983[FLOAT, 256x256x3x3]
%onnx::Conv_986[FLOAT, 256x256x1x1]
%onnx::Conv_989[FLOAT, 256x256x3x3]
%onnx::Conv_992[FLOAT, 256x256x3x3]
%onnx::Conv_995[FLOAT, 256x256x3x3]
%onnx::Conv_998[FLOAT, 512x256x1x1]
%onnx::Conv_999[FLOAT, 512]
%onnx::Conv_1001[FLOAT, 512x512x3x3]
%onnx::Conv_1004[FLOAT, 512x256x1x1]
%onnx::Conv_1007[FLOAT, 512x512x3x3]
%onnx::Conv_1010[FLOAT, 512x512x3x3]
%onnx::Conv_1013[FLOAT, 512x512x3x3]
%onnx::Conv_1016[FLOAT, 512x512x1x1]
%onnx::Conv_1019[FLOAT, 512x512x3x3]
%onnx::Conv_1022[FLOAT, 512x512x1x1]
%onnx::Conv_1025[FLOAT, 512x512x3x3]
%onnx::Conv_1028[FLOAT, 512x512x3x3]
%onnx::Conv_1031[FLOAT, 512x512x3x3]
%onnx::Conv_1034[FLOAT, 512x512x1x1]
%onnx::Conv_1037[FLOAT, 512x512x3x3]
%onnx::Conv_1040[FLOAT, 512x512x1x1]
%onnx::Conv_1043[FLOAT, 512x512x3x3]
%onnx::Conv_1046[FLOAT, 512x512x3x3]
%onnx::Conv_1049[FLOAT, 512x512x3x3]
) {
%onnx::Conv_1050 = Identity(%onnx::Conv_999)
%onnx::Conv_1047 = Identity(%onnx::Conv_999)
%onnx::Conv_1044 = Identity(%onnx::Conv_999)
%onnx::Conv_1041 = Identity(%onnx::Conv_999)
%onnx::Conv_1038 = Identity(%onnx::Conv_999)
%onnx::Conv_1035 = Identity(%onnx::Conv_999)
%onnx::Conv_1032 = Identity(%onnx::Conv_999)
%onnx::Conv_1029 = Identity(%onnx::Conv_999)
%onnx::Conv_1026 = Identity(%onnx::Conv_999)
%onnx::Conv_1023 = Identity(%onnx::Conv_999)
%onnx::Conv_1020 = Identity(%onnx::Conv_999)
%onnx::Conv_1017 = Identity(%onnx::Conv_999)
%onnx::Conv_1014 = Identity(%onnx::Conv_999)
%onnx::Conv_1011 = Identity(%onnx::Conv_999)
%onnx::Conv_1008 = Identity(%onnx::Conv_999)
%onnx::Conv_1005 = Identity(%onnx::Conv_999)
%onnx::Conv_1002 = Identity(%onnx::Conv_999)
%onnx::Conv_996 = Identity(%onnx::Conv_945)
%onnx::Conv_993 = Identity(%onnx::Conv_945)
%onnx::Conv_990 = Identity(%onnx::Conv_945)
%onnx::Conv_987 = Identity(%onnx::Conv_945)
%onnx::Conv_984 = Identity(%onnx::Conv_945)
%onnx::Conv_981 = Identity(%onnx::Conv_945)
%onnx::Conv_978 = Identity(%onnx::Conv_945)
%onnx::Conv_975 = Identity(%onnx::Conv_945)
%onnx::Conv_972 = Identity(%onnx::Conv_945)
%onnx::Conv_969 = Identity(%onnx::Conv_945)
%onnx::Conv_966 = Identity(%onnx::Conv_945)
%onnx::Conv_963 = Identity(%onnx::Conv_945)
%onnx::Conv_960 = Identity(%onnx::Conv_945)
%onnx::Conv_957 = Identity(%onnx::Conv_945)
%onnx::Conv_954 = Identity(%onnx::Conv_945)
%onnx::Conv_951 = Identity(%onnx::Conv_945)
%onnx::Conv_948 = Identity(%onnx::Conv_945)
%onnx::Conv_942 = Identity(%onnx::Conv_888)
%onnx::Conv_939 = Identity(%onnx::Conv_888)
%onnx::Conv_936 = Identity(%onnx::Conv_888)
%onnx::Conv_933 = Identity(%onnx::Conv_888)
%onnx::Conv_930 = Identity(%onnx::Conv_888)
%onnx::Conv_927 = Identity(%onnx::Conv_888)
%onnx::Conv_924 = Identity(%onnx::Conv_888)
%onnx::Conv_921 = Identity(%onnx::Conv_888)
%onnx::Conv_918 = Identity(%onnx::Conv_888)
%onnx::Conv_915 = Identity(%onnx::Conv_888)
%onnx::Conv_912 = Identity(%onnx::Conv_888)
%onnx::Conv_909 = Identity(%onnx::Conv_888)
%onnx::Conv_906 = Identity(%onnx::Conv_888)
%onnx::Conv_903 = Identity(%onnx::Conv_888)
%onnx::Conv_900 = Identity(%onnx::Conv_888)
%onnx::Conv_897 = Identity(%onnx::Conv_888)
%onnx::Conv_894 = Identity(%onnx::Conv_888)
%onnx::Conv_891 = Identity(%onnx::Conv_888)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_887, %onnx::Conv_888)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/Add_1_output_0, %/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/Add_3_output_0, %/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_5_output_0 = Add(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/Add_6_output_0 = Add(%/layers.1/Add_5_output_0, %/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_6_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_7_output_0 = Add(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_4_output_0)
%/layers.1/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_7_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_908, %onnx::Conv_909)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_911, %onnx::Conv_912)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_914, %onnx::Conv_915)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/Add_1_output_0, %/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_917, %onnx::Conv_918)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/Add_3_output_0, %/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_920, %onnx::Conv_921)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_5_output_0 = Add(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/Add_6_output_0 = Add(%/layers.2/Add_5_output_0, %/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_6_output_0, %onnx::Conv_923, %onnx::Conv_924)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_7_output_0 = Add(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_4_output_0)
%/layers.2/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_7_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_926, %onnx::Conv_927)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_929, %onnx::Conv_930)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_932, %onnx::Conv_933)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/Add_1_output_0, %/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_935, %onnx::Conv_936)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/Add_3_output_0, %/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_4_output_0, %onnx::Conv_938, %onnx::Conv_939)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_5_output_0 = Add(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/Add_6_output_0 = Add(%/layers.3/Add_5_output_0, %/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_6_output_0, %onnx::Conv_941, %onnx::Conv_942)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_7_output_0 = Add(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_4_output_0)
%/layers.3/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_7_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_944, %onnx::Conv_945)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_947, %onnx::Conv_948)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_950, %onnx::Conv_951)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/Add_1_output_0, %/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_953, %onnx::Conv_954)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/Add_3_output_0, %/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_956, %onnx::Conv_957)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_5_output_0 = Add(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/Add_6_output_0 = Add(%/layers.5/Add_5_output_0, %/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_6_output_0, %onnx::Conv_959, %onnx::Conv_960)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_7_output_0 = Add(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_4_output_0)
%/layers.5/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_7_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_962, %onnx::Conv_963)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_965, %onnx::Conv_966)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_968, %onnx::Conv_969)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/Add_1_output_0, %/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/Add_3_output_0, %/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_974, %onnx::Conv_975)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_5_output_0 = Add(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/Add_6_output_0 = Add(%/layers.6/Add_5_output_0, %/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_6_output_0, %onnx::Conv_977, %onnx::Conv_978)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_7_output_0 = Add(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_4_output_0)
%/layers.6/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_7_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/Add_1_output_0, %/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/Add_3_output_0, %/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_4_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_5_output_0 = Add(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/Add_6_output_0 = Add(%/layers.7/Add_5_output_0, %/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_6_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_7_output_0 = Add(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_4_output_0)
%/layers.7/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_7_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/Add_1_output_0, %/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/Add_3_output_0, %/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_5_output_0 = Add(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/Add_6_output_0 = Add(%/layers.9/Add_5_output_0, %/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_6_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_7_output_0 = Add(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_4_output_0)
%/layers.9/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_7_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/Add_1_output_0, %/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_1025, %onnx::Conv_1026)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/Add_3_output_0, %/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_1028, %onnx::Conv_1029)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_5_output_0 = Add(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/Add_6_output_0 = Add(%/layers.10/Add_5_output_0, %/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_6_output_0, %onnx::Conv_1031, %onnx::Conv_1032)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_7_output_0 = Add(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_4_output_0)
%/layers.10/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_7_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_1034, %onnx::Conv_1035)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_1037, %onnx::Conv_1038)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_1040, %onnx::Conv_1041)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/Add_1_output_0, %/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_1043, %onnx::Conv_1044)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/Add_3_output_0, %/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_4_output_0, %onnx::Conv_1046, %onnx::Conv_1047)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_5_output_0 = Add(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/Add_6_output_0 = Add(%/layers.11/Add_5_output_0, %/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_6_output_0, %onnx::Conv_1049, %onnx::Conv_1050)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_7_output_0 = Add(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_4_output_0)
%/layers.11/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_7_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/vertex_op.5/maxpool/MaxPool_output_0)
%885 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %885
}
|
val_accuracy
| 88.982373
| 11,449,673,728
| 38,936,714
|
{'zcp_epe_nas': 122.5295651685737, 'zcp_fisher': 6755.8076171875, 'zcp_flops': 183194779648.0, 'zcp_grad_norm': 1323.6494140625, 'zcp_grasp': 1217.28125, 'zcp_jacov': -16.04671388475012, 'zcp_l2_norm': 1242.5120849609375, 'zcp_nwot': 234.59996072842955, 'zcp_params': 38936714.0, 'zcp_plain': 0.055311806499958004, 'zcp_snip': 11167.64453125, 'zcp_synflow': 175.91948762032106, 'zcp_zen': 135.56927490234375, 'zcp_val_accuracy': 0.925480782985687}
| |
NASBench101_321861
|
NASBench101
|
321861
|
c2c2a82fae90cb9b6420698e9c062d18
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_662[FLOAT, 128x3x3x3]
%onnx::Conv_663[FLOAT, 128]
%onnx::Conv_665[FLOAT, 128x128x1x1]
%onnx::Conv_668[FLOAT, 128x128x3x3]
%onnx::Conv_671[FLOAT, 128x128x1x1]
%onnx::Conv_674[FLOAT, 128x128x3x3]
%onnx::Conv_677[FLOAT, 128x128x1x1]
%onnx::Conv_680[FLOAT, 128x128x3x3]
%onnx::Conv_683[FLOAT, 128x128x1x1]
%onnx::Conv_686[FLOAT, 128x128x3x3]
%onnx::Conv_689[FLOAT, 128x128x1x1]
%onnx::Conv_692[FLOAT, 128x128x3x3]
%onnx::Conv_695[FLOAT, 128x128x1x1]
%onnx::Conv_698[FLOAT, 128x128x3x3]
%onnx::Conv_701[FLOAT, 256x128x1x1]
%onnx::Conv_702[FLOAT, 256]
%onnx::Conv_704[FLOAT, 256x256x3x3]
%onnx::Conv_707[FLOAT, 256x256x1x1]
%onnx::Conv_710[FLOAT, 256x256x3x3]
%onnx::Conv_713[FLOAT, 256x256x1x1]
%onnx::Conv_716[FLOAT, 256x256x3x3]
%onnx::Conv_719[FLOAT, 256x256x1x1]
%onnx::Conv_722[FLOAT, 256x256x3x3]
%onnx::Conv_725[FLOAT, 256x256x1x1]
%onnx::Conv_728[FLOAT, 256x256x3x3]
%onnx::Conv_731[FLOAT, 256x256x1x1]
%onnx::Conv_734[FLOAT, 256x256x3x3]
%onnx::Conv_737[FLOAT, 512x256x1x1]
%onnx::Conv_738[FLOAT, 512]
%onnx::Conv_740[FLOAT, 512x512x3x3]
%onnx::Conv_743[FLOAT, 512x512x1x1]
%onnx::Conv_746[FLOAT, 512x512x3x3]
%onnx::Conv_749[FLOAT, 512x512x1x1]
%onnx::Conv_752[FLOAT, 512x512x3x3]
%onnx::Conv_755[FLOAT, 512x512x1x1]
%onnx::Conv_758[FLOAT, 512x512x3x3]
%onnx::Conv_761[FLOAT, 512x512x1x1]
%onnx::Conv_764[FLOAT, 512x512x3x3]
%onnx::Conv_767[FLOAT, 512x512x1x1]
%onnx::Conv_770[FLOAT, 512x512x3x3]
) {
%onnx::Conv_771 = Identity(%onnx::Conv_738)
%onnx::Conv_768 = Identity(%onnx::Conv_738)
%onnx::Conv_765 = Identity(%onnx::Conv_738)
%onnx::Conv_762 = Identity(%onnx::Conv_738)
%onnx::Conv_759 = Identity(%onnx::Conv_738)
%onnx::Conv_756 = Identity(%onnx::Conv_738)
%onnx::Conv_753 = Identity(%onnx::Conv_738)
%onnx::Conv_750 = Identity(%onnx::Conv_738)
%onnx::Conv_747 = Identity(%onnx::Conv_738)
%onnx::Conv_744 = Identity(%onnx::Conv_738)
%onnx::Conv_741 = Identity(%onnx::Conv_738)
%onnx::Conv_735 = Identity(%onnx::Conv_702)
%onnx::Conv_732 = Identity(%onnx::Conv_702)
%onnx::Conv_729 = Identity(%onnx::Conv_702)
%onnx::Conv_726 = Identity(%onnx::Conv_702)
%onnx::Conv_723 = Identity(%onnx::Conv_702)
%onnx::Conv_720 = Identity(%onnx::Conv_702)
%onnx::Conv_717 = Identity(%onnx::Conv_702)
%onnx::Conv_714 = Identity(%onnx::Conv_702)
%onnx::Conv_711 = Identity(%onnx::Conv_702)
%onnx::Conv_708 = Identity(%onnx::Conv_702)
%onnx::Conv_705 = Identity(%onnx::Conv_702)
%onnx::Conv_699 = Identity(%onnx::Conv_663)
%onnx::Conv_696 = Identity(%onnx::Conv_663)
%onnx::Conv_693 = Identity(%onnx::Conv_663)
%onnx::Conv_690 = Identity(%onnx::Conv_663)
%onnx::Conv_687 = Identity(%onnx::Conv_663)
%onnx::Conv_684 = Identity(%onnx::Conv_663)
%onnx::Conv_681 = Identity(%onnx::Conv_663)
%onnx::Conv_678 = Identity(%onnx::Conv_663)
%onnx::Conv_675 = Identity(%onnx::Conv_663)
%onnx::Conv_672 = Identity(%onnx::Conv_663)
%onnx::Conv_669 = Identity(%onnx::Conv_663)
%onnx::Conv_666 = Identity(%onnx::Conv_663)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_662, %onnx::Conv_663)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_665, %onnx::Conv_666)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_668, %onnx::Conv_669)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/Add_2_output_0, %/layers.1/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.1/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_671, %onnx::Conv_672)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %/layers.1/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.1/Add_5_output_0 = Add(%/layers.1/Add_4_output_0, %/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_5_output_0, %onnx::Conv_674, %onnx::Conv_675)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_677, %onnx::Conv_678)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_680, %onnx::Conv_681)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/Add_2_output_0, %/layers.2/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.2/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_683, %onnx::Conv_684)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %/layers.2/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.2/Add_5_output_0 = Add(%/layers.2/Add_4_output_0, %/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_5_output_0, %onnx::Conv_686, %onnx::Conv_687)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_689, %onnx::Conv_690)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_692, %onnx::Conv_693)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/Add_2_output_0, %/layers.3/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.3/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_695, %onnx::Conv_696)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %/layers.3/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.3/Add_5_output_0 = Add(%/layers.3/Add_4_output_0, %/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_5_output_0, %onnx::Conv_698, %onnx::Conv_699)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_701, %onnx::Conv_702)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_704, %onnx::Conv_705)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/Add_2_output_0, %/layers.5/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.5/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_707, %onnx::Conv_708)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %/layers.5/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.5/Add_5_output_0 = Add(%/layers.5/Add_4_output_0, %/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_5_output_0, %onnx::Conv_710, %onnx::Conv_711)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_713, %onnx::Conv_714)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_716, %onnx::Conv_717)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/Add_2_output_0, %/layers.6/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.6/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_719, %onnx::Conv_720)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %/layers.6/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.6/Add_5_output_0 = Add(%/layers.6/Add_4_output_0, %/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_5_output_0, %onnx::Conv_722, %onnx::Conv_723)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_725, %onnx::Conv_726)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_728, %onnx::Conv_729)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/Add_2_output_0, %/layers.7/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.7/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_731, %onnx::Conv_732)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %/layers.7/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.7/Add_5_output_0 = Add(%/layers.7/Add_4_output_0, %/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_5_output_0, %onnx::Conv_734, %onnx::Conv_735)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_737, %onnx::Conv_738)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_740, %onnx::Conv_741)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/Add_2_output_0, %/layers.9/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.9/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_743, %onnx::Conv_744)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %/layers.9/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.9/Add_5_output_0 = Add(%/layers.9/Add_4_output_0, %/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_5_output_0, %onnx::Conv_746, %onnx::Conv_747)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_749, %onnx::Conv_750)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_752, %onnx::Conv_753)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/Add_2_output_0, %/layers.10/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.10/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_755, %onnx::Conv_756)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %/layers.10/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.10/Add_5_output_0 = Add(%/layers.10/Add_4_output_0, %/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_5_output_0, %onnx::Conv_758, %onnx::Conv_759)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_761, %onnx::Conv_762)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_764, %onnx::Conv_765)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/Add_2_output_0, %/layers.11/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.11/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/vertex_op.3/maxpool/MaxPool_output_0, %onnx::Conv_767, %onnx::Conv_768)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %/layers.11/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.11/Add_5_output_0 = Add(%/layers.11/Add_4_output_0, %/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_5_output_0, %onnx::Conv_770, %onnx::Conv_771)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%660 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %660
}
|
val_accuracy
| 91.015625
| 6,036,400,128
| 20,510,346
|
{'zcp_epe_nas': 143.51258151455127, 'zcp_fisher': 92.5037841796875, 'zcp_flops': 96582402048.0, 'zcp_grad_norm': 126.34672546386719, 'zcp_grasp': -7.3829345703125, 'zcp_jacov': -16.04755661917148, 'zcp_l2_norm': 834.7434692382812, 'zcp_nwot': 228.41695553683874, 'zcp_params': 20510346.0, 'zcp_plain': 0.056475777179002006, 'zcp_snip': 1138.342529296875, 'zcp_synflow': 136.58863246993297, 'zcp_zen': 97.16220092773438, 'zcp_val_accuracy': 0.9358974099159241}
| |
NASBench101_214313
|
NASBench101
|
214313
|
81d531cfc07e7b1fd9c1e0d23fdcfe1f
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_545[FLOAT, 128x3x3x3]
%onnx::Conv_546[FLOAT, 128]
%onnx::Conv_548[FLOAT, 64x128x1x1]
%onnx::Conv_549[FLOAT, 64]
%onnx::Conv_551[FLOAT, 64x128x1x1]
%onnx::Conv_554[FLOAT, 64x64x3x3]
%onnx::Conv_557[FLOAT, 64x128x1x1]
%onnx::Conv_560[FLOAT, 64x128x1x1]
%onnx::Conv_563[FLOAT, 64x64x3x3]
%onnx::Conv_566[FLOAT, 64x128x1x1]
%onnx::Conv_569[FLOAT, 64x128x1x1]
%onnx::Conv_572[FLOAT, 64x64x3x3]
%onnx::Conv_575[FLOAT, 128x128x1x1]
%onnx::Conv_578[FLOAT, 128x128x1x1]
%onnx::Conv_581[FLOAT, 128x128x3x3]
%onnx::Conv_584[FLOAT, 128x256x1x1]
%onnx::Conv_587[FLOAT, 128x256x1x1]
%onnx::Conv_590[FLOAT, 128x128x3x3]
%onnx::Conv_593[FLOAT, 128x256x1x1]
%onnx::Conv_596[FLOAT, 128x256x1x1]
%onnx::Conv_599[FLOAT, 128x128x3x3]
%onnx::Conv_602[FLOAT, 256x256x1x1]
%onnx::Conv_603[FLOAT, 256]
%onnx::Conv_605[FLOAT, 256x256x1x1]
%onnx::Conv_608[FLOAT, 256x256x3x3]
%onnx::Conv_611[FLOAT, 256x512x1x1]
%onnx::Conv_614[FLOAT, 256x512x1x1]
%onnx::Conv_617[FLOAT, 256x256x3x3]
%onnx::Conv_620[FLOAT, 256x512x1x1]
%onnx::Conv_623[FLOAT, 256x512x1x1]
%onnx::Conv_626[FLOAT, 256x256x3x3]
) {
%onnx::Conv_627 = Identity(%onnx::Conv_603)
%onnx::Conv_624 = Identity(%onnx::Conv_603)
%onnx::Conv_621 = Identity(%onnx::Conv_603)
%onnx::Conv_618 = Identity(%onnx::Conv_603)
%onnx::Conv_615 = Identity(%onnx::Conv_603)
%onnx::Conv_612 = Identity(%onnx::Conv_603)
%onnx::Conv_609 = Identity(%onnx::Conv_603)
%onnx::Conv_606 = Identity(%onnx::Conv_603)
%onnx::Conv_600 = Identity(%onnx::Conv_546)
%onnx::Conv_597 = Identity(%onnx::Conv_546)
%onnx::Conv_594 = Identity(%onnx::Conv_546)
%onnx::Conv_591 = Identity(%onnx::Conv_546)
%onnx::Conv_588 = Identity(%onnx::Conv_546)
%onnx::Conv_585 = Identity(%onnx::Conv_546)
%onnx::Conv_582 = Identity(%onnx::Conv_546)
%onnx::Conv_579 = Identity(%onnx::Conv_546)
%onnx::Conv_576 = Identity(%onnx::Conv_546)
%onnx::Conv_573 = Identity(%onnx::Conv_549)
%onnx::Conv_570 = Identity(%onnx::Conv_549)
%onnx::Conv_567 = Identity(%onnx::Conv_549)
%onnx::Conv_564 = Identity(%onnx::Conv_549)
%onnx::Conv_561 = Identity(%onnx::Conv_549)
%onnx::Conv_558 = Identity(%onnx::Conv_549)
%onnx::Conv_555 = Identity(%onnx::Conv_549)
%onnx::Conv_552 = Identity(%onnx::Conv_549)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_545, %onnx::Conv_546)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_548, %onnx::Conv_549)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_551, %onnx::Conv_552)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_554, %onnx::Conv_555)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.1/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_557, %onnx::Conv_558)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_560, %onnx::Conv_561)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_563, %onnx::Conv_564)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.2/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_566, %onnx::Conv_567)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_569, %onnx::Conv_570)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_572, %onnx::Conv_573)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.3/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_575, %onnx::Conv_576)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_578, %onnx::Conv_579)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_581, %onnx::Conv_582)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.5/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_584, %onnx::Conv_585)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_587, %onnx::Conv_588)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_590, %onnx::Conv_591)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.6/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_593, %onnx::Conv_594)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_596, %onnx::Conv_597)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_599, %onnx::Conv_600)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.7/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_602, %onnx::Conv_603)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_605, %onnx::Conv_606)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_608, %onnx::Conv_609)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.9/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_611, %onnx::Conv_612)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_614, %onnx::Conv_615)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_617, %onnx::Conv_618)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.10/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_620, %onnx::Conv_621)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_623, %onnx::Conv_624)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_626, %onnx::Conv_627)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.11/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/vertex_op.5/maxpool/MaxPool_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%543 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %543
}
|
val_accuracy
| 90.274441
| 964,306,944
| 3,207,690
|
{'zcp_epe_nas': 86.6512746632789, 'zcp_fisher': 15.742091178894043, 'zcp_flops': 15428911104.0, 'zcp_grad_norm': 64.03255462646484, 'zcp_grasp': -24.8416748046875, 'zcp_jacov': -16.05222594238579, 'zcp_l2_norm': 545.4808959960938, 'zcp_nwot': 213.17911199559683, 'zcp_params': 3207690.0, 'zcp_plain': 0.185451656579971, 'zcp_snip': 388.9736022949219, 'zcp_synflow': 67.43730359573391, 'zcp_zen': 60.70955276489258, 'zcp_val_accuracy': 0.9082531929016111}
| |
NASBench101_121387
|
NASBench101
|
121387
|
4956bca6bdf76a9d0a3e1c9b407119b0
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_734[FLOAT, 128x3x3x3]
%onnx::Conv_735[FLOAT, 128]
%onnx::Conv_737[FLOAT, 128x128x1x1]
%onnx::Conv_740[FLOAT, 128x128x1x1]
%onnx::Conv_743[FLOAT, 128x128x3x3]
%onnx::Conv_746[FLOAT, 128x128x1x1]
%onnx::Conv_749[FLOAT, 128x128x3x3]
%onnx::Conv_752[FLOAT, 128x128x1x1]
%onnx::Conv_755[FLOAT, 128x128x1x1]
%onnx::Conv_758[FLOAT, 128x128x3x3]
%onnx::Conv_761[FLOAT, 128x128x1x1]
%onnx::Conv_764[FLOAT, 128x128x3x3]
%onnx::Conv_767[FLOAT, 128x128x1x1]
%onnx::Conv_770[FLOAT, 128x128x1x1]
%onnx::Conv_773[FLOAT, 128x128x3x3]
%onnx::Conv_776[FLOAT, 128x128x1x1]
%onnx::Conv_779[FLOAT, 128x128x3x3]
%onnx::Conv_782[FLOAT, 256x128x1x1]
%onnx::Conv_783[FLOAT, 256]
%onnx::Conv_785[FLOAT, 256x256x1x1]
%onnx::Conv_788[FLOAT, 256x256x3x3]
%onnx::Conv_791[FLOAT, 256x256x1x1]
%onnx::Conv_794[FLOAT, 256x256x3x3]
%onnx::Conv_797[FLOAT, 256x256x1x1]
%onnx::Conv_800[FLOAT, 256x256x1x1]
%onnx::Conv_803[FLOAT, 256x256x3x3]
%onnx::Conv_806[FLOAT, 256x256x1x1]
%onnx::Conv_809[FLOAT, 256x256x3x3]
%onnx::Conv_812[FLOAT, 256x256x1x1]
%onnx::Conv_815[FLOAT, 256x256x1x1]
%onnx::Conv_818[FLOAT, 256x256x3x3]
%onnx::Conv_821[FLOAT, 256x256x1x1]
%onnx::Conv_824[FLOAT, 256x256x3x3]
%onnx::Conv_827[FLOAT, 512x256x1x1]
%onnx::Conv_828[FLOAT, 512]
%onnx::Conv_830[FLOAT, 512x512x1x1]
%onnx::Conv_833[FLOAT, 512x512x3x3]
%onnx::Conv_836[FLOAT, 512x512x1x1]
%onnx::Conv_839[FLOAT, 512x512x3x3]
%onnx::Conv_842[FLOAT, 512x512x1x1]
%onnx::Conv_845[FLOAT, 512x512x1x1]
%onnx::Conv_848[FLOAT, 512x512x3x3]
%onnx::Conv_851[FLOAT, 512x512x1x1]
%onnx::Conv_854[FLOAT, 512x512x3x3]
%onnx::Conv_857[FLOAT, 512x512x1x1]
%onnx::Conv_860[FLOAT, 512x512x1x1]
%onnx::Conv_863[FLOAT, 512x512x3x3]
%onnx::Conv_866[FLOAT, 512x512x1x1]
%onnx::Conv_869[FLOAT, 512x512x3x3]
) {
%onnx::Conv_870 = Identity(%onnx::Conv_828)
%onnx::Conv_867 = Identity(%onnx::Conv_828)
%onnx::Conv_864 = Identity(%onnx::Conv_828)
%onnx::Conv_861 = Identity(%onnx::Conv_828)
%onnx::Conv_858 = Identity(%onnx::Conv_828)
%onnx::Conv_855 = Identity(%onnx::Conv_828)
%onnx::Conv_852 = Identity(%onnx::Conv_828)
%onnx::Conv_849 = Identity(%onnx::Conv_828)
%onnx::Conv_846 = Identity(%onnx::Conv_828)
%onnx::Conv_843 = Identity(%onnx::Conv_828)
%onnx::Conv_840 = Identity(%onnx::Conv_828)
%onnx::Conv_837 = Identity(%onnx::Conv_828)
%onnx::Conv_834 = Identity(%onnx::Conv_828)
%onnx::Conv_831 = Identity(%onnx::Conv_828)
%onnx::Conv_825 = Identity(%onnx::Conv_783)
%onnx::Conv_822 = Identity(%onnx::Conv_783)
%onnx::Conv_819 = Identity(%onnx::Conv_783)
%onnx::Conv_816 = Identity(%onnx::Conv_783)
%onnx::Conv_813 = Identity(%onnx::Conv_783)
%onnx::Conv_810 = Identity(%onnx::Conv_783)
%onnx::Conv_807 = Identity(%onnx::Conv_783)
%onnx::Conv_804 = Identity(%onnx::Conv_783)
%onnx::Conv_801 = Identity(%onnx::Conv_783)
%onnx::Conv_798 = Identity(%onnx::Conv_783)
%onnx::Conv_795 = Identity(%onnx::Conv_783)
%onnx::Conv_792 = Identity(%onnx::Conv_783)
%onnx::Conv_789 = Identity(%onnx::Conv_783)
%onnx::Conv_786 = Identity(%onnx::Conv_783)
%onnx::Conv_780 = Identity(%onnx::Conv_735)
%onnx::Conv_777 = Identity(%onnx::Conv_735)
%onnx::Conv_774 = Identity(%onnx::Conv_735)
%onnx::Conv_771 = Identity(%onnx::Conv_735)
%onnx::Conv_768 = Identity(%onnx::Conv_735)
%onnx::Conv_765 = Identity(%onnx::Conv_735)
%onnx::Conv_762 = Identity(%onnx::Conv_735)
%onnx::Conv_759 = Identity(%onnx::Conv_735)
%onnx::Conv_756 = Identity(%onnx::Conv_735)
%onnx::Conv_753 = Identity(%onnx::Conv_735)
%onnx::Conv_750 = Identity(%onnx::Conv_735)
%onnx::Conv_747 = Identity(%onnx::Conv_735)
%onnx::Conv_744 = Identity(%onnx::Conv_735)
%onnx::Conv_741 = Identity(%onnx::Conv_735)
%onnx::Conv_738 = Identity(%onnx::Conv_735)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_734, %onnx::Conv_735)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_737, %onnx::Conv_738)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_740, %onnx::Conv_741)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_743, %onnx::Conv_744)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/Add_1_output_0, %/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_746, %onnx::Conv_747)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/Add_3_output_0, %/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_749, %onnx::Conv_750)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_752, %onnx::Conv_753)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_755, %onnx::Conv_756)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_758, %onnx::Conv_759)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/Add_1_output_0, %/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_761, %onnx::Conv_762)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/Add_3_output_0, %/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_764, %onnx::Conv_765)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_767, %onnx::Conv_768)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_770, %onnx::Conv_771)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_773, %onnx::Conv_774)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/Add_1_output_0, %/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_776, %onnx::Conv_777)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/Add_3_output_0, %/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_4_output_0, %onnx::Conv_779, %onnx::Conv_780)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_782, %onnx::Conv_783)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_785, %onnx::Conv_786)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_788, %onnx::Conv_789)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/Add_1_output_0, %/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/Add_3_output_0, %/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_797, %onnx::Conv_798)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_800, %onnx::Conv_801)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_803, %onnx::Conv_804)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/Add_1_output_0, %/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_806, %onnx::Conv_807)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/Add_3_output_0, %/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_809, %onnx::Conv_810)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_812, %onnx::Conv_813)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_815, %onnx::Conv_816)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_818, %onnx::Conv_819)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/Add_1_output_0, %/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_821, %onnx::Conv_822)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/Add_3_output_0, %/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_4_output_0, %onnx::Conv_824, %onnx::Conv_825)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_827, %onnx::Conv_828)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_830, %onnx::Conv_831)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_833, %onnx::Conv_834)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/Add_1_output_0, %/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_836, %onnx::Conv_837)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/Add_3_output_0, %/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_839, %onnx::Conv_840)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_842, %onnx::Conv_843)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_845, %onnx::Conv_846)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_848, %onnx::Conv_849)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/Add_1_output_0, %/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_851, %onnx::Conv_852)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/Add_3_output_0, %/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_854, %onnx::Conv_855)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_857, %onnx::Conv_858)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/Add_1_output_0, %/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/Add_3_output_0, %/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_4_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%732 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %732
}
|
val_accuracy
| 92.027241
| 6,343,895,040
| 21,547,914
|
{'zcp_epe_nas': 77.30577054003916, 'zcp_fisher': 34.62464141845703, 'zcp_flops': 101502320640.0, 'zcp_grad_norm': 92.59884643554688, 'zcp_grasp': -2.4559326171875, 'zcp_jacov': -16.062129906422708, 'zcp_l2_norm': 1046.9134521484375, 'zcp_nwot': 231.74439090040636, 'zcp_params': 21547914.0, 'zcp_plain': 0.013203138485550001, 'zcp_snip': 800.2980346679688, 'zcp_synflow': 132.22063346801482, 'zcp_zen': 104.3379898071289, 'zcp_val_accuracy': 0.903145015239715}
| |
NASBench101_197577
|
NASBench101
|
197577
|
7793a4afbd52fc2ed0037682800a8dba
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_896[FLOAT, 128x3x3x3]
%onnx::Conv_897[FLOAT, 128]
%onnx::Conv_899[FLOAT, 128x128x1x1]
%onnx::Conv_902[FLOAT, 128x128x1x1]
%onnx::Conv_905[FLOAT, 128x128x1x1]
%onnx::Conv_908[FLOAT, 128x128x3x3]
%onnx::Conv_911[FLOAT, 128x128x3x3]
%onnx::Conv_914[FLOAT, 128x128x1x1]
%onnx::Conv_917[FLOAT, 128x128x1x1]
%onnx::Conv_920[FLOAT, 128x128x1x1]
%onnx::Conv_923[FLOAT, 128x128x1x1]
%onnx::Conv_926[FLOAT, 128x128x3x3]
%onnx::Conv_929[FLOAT, 128x128x3x3]
%onnx::Conv_932[FLOAT, 128x128x1x1]
%onnx::Conv_935[FLOAT, 128x128x1x1]
%onnx::Conv_938[FLOAT, 128x128x1x1]
%onnx::Conv_941[FLOAT, 128x128x1x1]
%onnx::Conv_944[FLOAT, 128x128x3x3]
%onnx::Conv_947[FLOAT, 128x128x3x3]
%onnx::Conv_950[FLOAT, 128x128x1x1]
%onnx::Conv_953[FLOAT, 256x128x1x1]
%onnx::Conv_954[FLOAT, 256]
%onnx::Conv_956[FLOAT, 256x256x1x1]
%onnx::Conv_959[FLOAT, 256x128x1x1]
%onnx::Conv_962[FLOAT, 256x256x3x3]
%onnx::Conv_965[FLOAT, 256x256x3x3]
%onnx::Conv_968[FLOAT, 256x128x1x1]
%onnx::Conv_971[FLOAT, 256x256x1x1]
%onnx::Conv_974[FLOAT, 256x256x1x1]
%onnx::Conv_977[FLOAT, 256x256x1x1]
%onnx::Conv_980[FLOAT, 256x256x3x3]
%onnx::Conv_983[FLOAT, 256x256x3x3]
%onnx::Conv_986[FLOAT, 256x256x1x1]
%onnx::Conv_989[FLOAT, 256x256x1x1]
%onnx::Conv_992[FLOAT, 256x256x1x1]
%onnx::Conv_995[FLOAT, 256x256x1x1]
%onnx::Conv_998[FLOAT, 256x256x3x3]
%onnx::Conv_1001[FLOAT, 256x256x3x3]
%onnx::Conv_1004[FLOAT, 256x256x1x1]
%onnx::Conv_1007[FLOAT, 512x256x1x1]
%onnx::Conv_1008[FLOAT, 512]
%onnx::Conv_1010[FLOAT, 512x512x1x1]
%onnx::Conv_1013[FLOAT, 512x256x1x1]
%onnx::Conv_1016[FLOAT, 512x512x3x3]
%onnx::Conv_1019[FLOAT, 512x512x3x3]
%onnx::Conv_1022[FLOAT, 512x256x1x1]
%onnx::Conv_1025[FLOAT, 512x512x1x1]
%onnx::Conv_1028[FLOAT, 512x512x1x1]
%onnx::Conv_1031[FLOAT, 512x512x1x1]
%onnx::Conv_1034[FLOAT, 512x512x3x3]
%onnx::Conv_1037[FLOAT, 512x512x3x3]
%onnx::Conv_1040[FLOAT, 512x512x1x1]
%onnx::Conv_1043[FLOAT, 512x512x1x1]
%onnx::Conv_1046[FLOAT, 512x512x1x1]
%onnx::Conv_1049[FLOAT, 512x512x1x1]
%onnx::Conv_1052[FLOAT, 512x512x3x3]
%onnx::Conv_1055[FLOAT, 512x512x3x3]
%onnx::Conv_1058[FLOAT, 512x512x1x1]
) {
%onnx::Conv_1059 = Identity(%onnx::Conv_1008)
%onnx::Conv_1056 = Identity(%onnx::Conv_1008)
%onnx::Conv_1053 = Identity(%onnx::Conv_1008)
%onnx::Conv_1050 = Identity(%onnx::Conv_1008)
%onnx::Conv_1047 = Identity(%onnx::Conv_1008)
%onnx::Conv_1044 = Identity(%onnx::Conv_1008)
%onnx::Conv_1041 = Identity(%onnx::Conv_1008)
%onnx::Conv_1038 = Identity(%onnx::Conv_1008)
%onnx::Conv_1035 = Identity(%onnx::Conv_1008)
%onnx::Conv_1032 = Identity(%onnx::Conv_1008)
%onnx::Conv_1029 = Identity(%onnx::Conv_1008)
%onnx::Conv_1026 = Identity(%onnx::Conv_1008)
%onnx::Conv_1023 = Identity(%onnx::Conv_1008)
%onnx::Conv_1020 = Identity(%onnx::Conv_1008)
%onnx::Conv_1017 = Identity(%onnx::Conv_1008)
%onnx::Conv_1014 = Identity(%onnx::Conv_1008)
%onnx::Conv_1011 = Identity(%onnx::Conv_1008)
%onnx::Conv_1005 = Identity(%onnx::Conv_954)
%onnx::Conv_1002 = Identity(%onnx::Conv_954)
%onnx::Conv_999 = Identity(%onnx::Conv_954)
%onnx::Conv_996 = Identity(%onnx::Conv_954)
%onnx::Conv_993 = Identity(%onnx::Conv_954)
%onnx::Conv_990 = Identity(%onnx::Conv_954)
%onnx::Conv_987 = Identity(%onnx::Conv_954)
%onnx::Conv_984 = Identity(%onnx::Conv_954)
%onnx::Conv_981 = Identity(%onnx::Conv_954)
%onnx::Conv_978 = Identity(%onnx::Conv_954)
%onnx::Conv_975 = Identity(%onnx::Conv_954)
%onnx::Conv_972 = Identity(%onnx::Conv_954)
%onnx::Conv_969 = Identity(%onnx::Conv_954)
%onnx::Conv_966 = Identity(%onnx::Conv_954)
%onnx::Conv_963 = Identity(%onnx::Conv_954)
%onnx::Conv_960 = Identity(%onnx::Conv_954)
%onnx::Conv_957 = Identity(%onnx::Conv_954)
%onnx::Conv_951 = Identity(%onnx::Conv_897)
%onnx::Conv_948 = Identity(%onnx::Conv_897)
%onnx::Conv_945 = Identity(%onnx::Conv_897)
%onnx::Conv_942 = Identity(%onnx::Conv_897)
%onnx::Conv_939 = Identity(%onnx::Conv_897)
%onnx::Conv_936 = Identity(%onnx::Conv_897)
%onnx::Conv_933 = Identity(%onnx::Conv_897)
%onnx::Conv_930 = Identity(%onnx::Conv_897)
%onnx::Conv_927 = Identity(%onnx::Conv_897)
%onnx::Conv_924 = Identity(%onnx::Conv_897)
%onnx::Conv_921 = Identity(%onnx::Conv_897)
%onnx::Conv_918 = Identity(%onnx::Conv_897)
%onnx::Conv_915 = Identity(%onnx::Conv_897)
%onnx::Conv_912 = Identity(%onnx::Conv_897)
%onnx::Conv_909 = Identity(%onnx::Conv_897)
%onnx::Conv_906 = Identity(%onnx::Conv_897)
%onnx::Conv_903 = Identity(%onnx::Conv_897)
%onnx::Conv_900 = Identity(%onnx::Conv_897)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_896, %onnx::Conv_897)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_908, %onnx::Conv_909)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_911, %onnx::Conv_912)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0)
%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_914, %onnx::Conv_915)
%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_4_output_0)
%/layers.1/Add_5_output_0 = Add(%/layers.1/Add_4_output_0, %/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/Add_6_output_0 = Add(%/layers.1/Add_5_output_0, %/layers.1/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.1/Add_7_output_0 = Add(%/layers.1/Add_6_output_0, %/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_7_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_917, %onnx::Conv_918)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_920, %onnx::Conv_921)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_923, %onnx::Conv_924)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_926, %onnx::Conv_927)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_929, %onnx::Conv_930)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0)
%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_932, %onnx::Conv_933)
%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_4_output_0)
%/layers.2/Add_5_output_0 = Add(%/layers.2/Add_4_output_0, %/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/Add_6_output_0 = Add(%/layers.2/Add_5_output_0, %/layers.2/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.2/Add_7_output_0 = Add(%/layers.2/Add_6_output_0, %/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_7_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_935, %onnx::Conv_936)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_938, %onnx::Conv_939)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_941, %onnx::Conv_942)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_944, %onnx::Conv_945)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_947, %onnx::Conv_948)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0)
%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_950, %onnx::Conv_951)
%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_4_output_0)
%/layers.3/Add_5_output_0 = Add(%/layers.3/Add_4_output_0, %/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/Add_6_output_0 = Add(%/layers.3/Add_5_output_0, %/layers.3/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.3/Add_7_output_0 = Add(%/layers.3/Add_6_output_0, %/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_7_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_953, %onnx::Conv_954)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_956, %onnx::Conv_957)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_959, %onnx::Conv_960)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_962, %onnx::Conv_963)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_965, %onnx::Conv_966)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0)
%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_968, %onnx::Conv_969)
%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_4_output_0)
%/layers.5/Add_5_output_0 = Add(%/layers.5/Add_4_output_0, %/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/Add_6_output_0 = Add(%/layers.5/Add_5_output_0, %/layers.5/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.5/Add_7_output_0 = Add(%/layers.5/Add_6_output_0, %/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_7_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_974, %onnx::Conv_975)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_977, %onnx::Conv_978)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0)
%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_4_output_0)
%/layers.6/Add_5_output_0 = Add(%/layers.6/Add_4_output_0, %/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/Add_6_output_0 = Add(%/layers.6/Add_5_output_0, %/layers.6/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.6/Add_7_output_0 = Add(%/layers.6/Add_6_output_0, %/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_7_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0)
%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_4_output_0)
%/layers.7/Add_5_output_0 = Add(%/layers.7/Add_4_output_0, %/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/Add_6_output_0 = Add(%/layers.7/Add_5_output_0, %/layers.7/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.7/Add_7_output_0 = Add(%/layers.7/Add_6_output_0, %/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_7_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0)
%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_4_output_0)
%/layers.9/Add_5_output_0 = Add(%/layers.9/Add_4_output_0, %/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/Add_6_output_0 = Add(%/layers.9/Add_5_output_0, %/layers.9/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.9/Add_7_output_0 = Add(%/layers.9/Add_6_output_0, %/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_7_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_1025, %onnx::Conv_1026)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_1028, %onnx::Conv_1029)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_1031, %onnx::Conv_1032)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_1034, %onnx::Conv_1035)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_1037, %onnx::Conv_1038)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0)
%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_1040, %onnx::Conv_1041)
%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_4_output_0)
%/layers.10/Add_5_output_0 = Add(%/layers.10/Add_4_output_0, %/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/Add_6_output_0 = Add(%/layers.10/Add_5_output_0, %/layers.10/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.10/Add_7_output_0 = Add(%/layers.10/Add_6_output_0, %/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_7_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_1043, %onnx::Conv_1044)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_1046, %onnx::Conv_1047)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_1049, %onnx::Conv_1050)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_1052, %onnx::Conv_1053)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_1055, %onnx::Conv_1056)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0)
%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/maxpool/MaxPool_output_0, %onnx::Conv_1058, %onnx::Conv_1059)
%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_4_output_0)
%/layers.11/Add_5_output_0 = Add(%/layers.11/Add_4_output_0, %/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/Add_6_output_0 = Add(%/layers.11/Add_5_output_0, %/layers.11/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.11/Add_7_output_0 = Add(%/layers.11/Add_6_output_0, %/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_7_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/vertex_op.5/maxpool/MaxPool_output_0)
%894 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %894
}
|
val_accuracy
| 91.286057
| 6,584,281,088
| 22,257,802
|
{'zcp_epe_nas': 140.7329210817495, 'zcp_fisher': 289.6521301269531, 'zcp_flops': 105348497408.0, 'zcp_grad_norm': 352.0042419433594, 'zcp_grasp': -452.9384765625, 'zcp_jacov': -16.06692269804088, 'zcp_l2_norm': 1226.4107666015625, 'zcp_nwot': 235.04640973069988, 'zcp_params': 22257802.0, 'zcp_plain': 0.266954720020294, 'zcp_snip': 2853.099609375, 'zcp_synflow': 94.2438897708618, 'zcp_zen': 122.0214614868164, 'zcp_val_accuracy': 0.9130609035491941}
| |
NASBench101_421372
|
NASBench101
|
421372
|
fea080e153fd656f77487f1c4ad37124
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_851[FLOAT, 128x3x3x3]
%onnx::Conv_852[FLOAT, 128]
%onnx::Conv_854[FLOAT, 128x128x1x1]
%onnx::Conv_857[FLOAT, 128x128x1x1]
%onnx::Conv_860[FLOAT, 128x128x1x1]
%onnx::Conv_863[FLOAT, 128x128x1x1]
%onnx::Conv_866[FLOAT, 128x128x1x1]
%onnx::Conv_869[FLOAT, 128x128x1x1]
%onnx::Conv_872[FLOAT, 128x128x1x1]
%onnx::Conv_875[FLOAT, 128x128x1x1]
%onnx::Conv_878[FLOAT, 128x128x1x1]
%onnx::Conv_881[FLOAT, 128x128x1x1]
%onnx::Conv_884[FLOAT, 128x128x1x1]
%onnx::Conv_887[FLOAT, 128x128x1x1]
%onnx::Conv_890[FLOAT, 128x128x1x1]
%onnx::Conv_893[FLOAT, 128x128x1x1]
%onnx::Conv_896[FLOAT, 128x128x1x1]
%onnx::Conv_899[FLOAT, 128x128x1x1]
%onnx::Conv_902[FLOAT, 128x128x1x1]
%onnx::Conv_905[FLOAT, 128x128x1x1]
%onnx::Conv_908[FLOAT, 256x128x1x1]
%onnx::Conv_909[FLOAT, 256]
%onnx::Conv_911[FLOAT, 256x256x1x1]
%onnx::Conv_914[FLOAT, 256x256x1x1]
%onnx::Conv_917[FLOAT, 256x128x1x1]
%onnx::Conv_920[FLOAT, 256x256x1x1]
%onnx::Conv_923[FLOAT, 256x128x1x1]
%onnx::Conv_926[FLOAT, 256x256x1x1]
%onnx::Conv_929[FLOAT, 256x256x1x1]
%onnx::Conv_932[FLOAT, 256x256x1x1]
%onnx::Conv_935[FLOAT, 256x256x1x1]
%onnx::Conv_938[FLOAT, 256x256x1x1]
%onnx::Conv_941[FLOAT, 256x256x1x1]
%onnx::Conv_944[FLOAT, 256x256x1x1]
%onnx::Conv_947[FLOAT, 256x256x1x1]
%onnx::Conv_950[FLOAT, 256x256x1x1]
%onnx::Conv_953[FLOAT, 256x256x1x1]
%onnx::Conv_956[FLOAT, 256x256x1x1]
%onnx::Conv_959[FLOAT, 256x256x1x1]
%onnx::Conv_962[FLOAT, 512x256x1x1]
%onnx::Conv_963[FLOAT, 512]
%onnx::Conv_965[FLOAT, 512x512x1x1]
%onnx::Conv_968[FLOAT, 512x512x1x1]
%onnx::Conv_971[FLOAT, 512x256x1x1]
%onnx::Conv_974[FLOAT, 512x512x1x1]
%onnx::Conv_977[FLOAT, 512x256x1x1]
%onnx::Conv_980[FLOAT, 512x512x1x1]
%onnx::Conv_983[FLOAT, 512x512x1x1]
%onnx::Conv_986[FLOAT, 512x512x1x1]
%onnx::Conv_989[FLOAT, 512x512x1x1]
%onnx::Conv_992[FLOAT, 512x512x1x1]
%onnx::Conv_995[FLOAT, 512x512x1x1]
%onnx::Conv_998[FLOAT, 512x512x1x1]
%onnx::Conv_1001[FLOAT, 512x512x1x1]
%onnx::Conv_1004[FLOAT, 512x512x1x1]
%onnx::Conv_1007[FLOAT, 512x512x1x1]
%onnx::Conv_1010[FLOAT, 512x512x1x1]
%onnx::Conv_1013[FLOAT, 512x512x1x1]
) {
%onnx::Conv_1014 = Identity(%onnx::Conv_963)
%onnx::Conv_1011 = Identity(%onnx::Conv_963)
%onnx::Conv_1008 = Identity(%onnx::Conv_963)
%onnx::Conv_1005 = Identity(%onnx::Conv_963)
%onnx::Conv_1002 = Identity(%onnx::Conv_963)
%onnx::Conv_999 = Identity(%onnx::Conv_963)
%onnx::Conv_996 = Identity(%onnx::Conv_963)
%onnx::Conv_993 = Identity(%onnx::Conv_963)
%onnx::Conv_990 = Identity(%onnx::Conv_963)
%onnx::Conv_987 = Identity(%onnx::Conv_963)
%onnx::Conv_984 = Identity(%onnx::Conv_963)
%onnx::Conv_981 = Identity(%onnx::Conv_963)
%onnx::Conv_978 = Identity(%onnx::Conv_963)
%onnx::Conv_975 = Identity(%onnx::Conv_963)
%onnx::Conv_972 = Identity(%onnx::Conv_963)
%onnx::Conv_969 = Identity(%onnx::Conv_963)
%onnx::Conv_966 = Identity(%onnx::Conv_963)
%onnx::Conv_960 = Identity(%onnx::Conv_909)
%onnx::Conv_957 = Identity(%onnx::Conv_909)
%onnx::Conv_954 = Identity(%onnx::Conv_909)
%onnx::Conv_951 = Identity(%onnx::Conv_909)
%onnx::Conv_948 = Identity(%onnx::Conv_909)
%onnx::Conv_945 = Identity(%onnx::Conv_909)
%onnx::Conv_942 = Identity(%onnx::Conv_909)
%onnx::Conv_939 = Identity(%onnx::Conv_909)
%onnx::Conv_936 = Identity(%onnx::Conv_909)
%onnx::Conv_933 = Identity(%onnx::Conv_909)
%onnx::Conv_930 = Identity(%onnx::Conv_909)
%onnx::Conv_927 = Identity(%onnx::Conv_909)
%onnx::Conv_924 = Identity(%onnx::Conv_909)
%onnx::Conv_921 = Identity(%onnx::Conv_909)
%onnx::Conv_918 = Identity(%onnx::Conv_909)
%onnx::Conv_915 = Identity(%onnx::Conv_909)
%onnx::Conv_912 = Identity(%onnx::Conv_909)
%onnx::Conv_906 = Identity(%onnx::Conv_852)
%onnx::Conv_903 = Identity(%onnx::Conv_852)
%onnx::Conv_900 = Identity(%onnx::Conv_852)
%onnx::Conv_897 = Identity(%onnx::Conv_852)
%onnx::Conv_894 = Identity(%onnx::Conv_852)
%onnx::Conv_891 = Identity(%onnx::Conv_852)
%onnx::Conv_888 = Identity(%onnx::Conv_852)
%onnx::Conv_885 = Identity(%onnx::Conv_852)
%onnx::Conv_882 = Identity(%onnx::Conv_852)
%onnx::Conv_879 = Identity(%onnx::Conv_852)
%onnx::Conv_876 = Identity(%onnx::Conv_852)
%onnx::Conv_873 = Identity(%onnx::Conv_852)
%onnx::Conv_870 = Identity(%onnx::Conv_852)
%onnx::Conv_867 = Identity(%onnx::Conv_852)
%onnx::Conv_864 = Identity(%onnx::Conv_852)
%onnx::Conv_861 = Identity(%onnx::Conv_852)
%onnx::Conv_858 = Identity(%onnx::Conv_852)
%onnx::Conv_855 = Identity(%onnx::Conv_852)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_851, %onnx::Conv_852)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_854, %onnx::Conv_855)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_857, %onnx::Conv_858)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/Add_2_output_0, %/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/Add_5_output_0 = Add(%/layers.1/Add_4_output_0, %/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/Add_6_output_0 = Add(%/layers.1/Add_5_output_0, %/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_6_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/Add_2_output_0, %/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/Add_5_output_0 = Add(%/layers.2/Add_4_output_0, %/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/Add_6_output_0 = Add(%/layers.2/Add_5_output_0, %/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_6_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/Add_2_output_0, %/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/Add_5_output_0 = Add(%/layers.3/Add_4_output_0, %/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/Add_6_output_0 = Add(%/layers.3/Add_5_output_0, %/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_6_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_908, %onnx::Conv_909)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_911, %onnx::Conv_912)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_914, %onnx::Conv_915)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_917, %onnx::Conv_918)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/Add_2_output_0, %/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_920, %onnx::Conv_921)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_923, %onnx::Conv_924)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/Add_5_output_0 = Add(%/layers.5/Add_4_output_0, %/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/Add_6_output_0 = Add(%/layers.5/Add_5_output_0, %/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_6_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_926, %onnx::Conv_927)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_929, %onnx::Conv_930)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_932, %onnx::Conv_933)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_935, %onnx::Conv_936)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/Add_2_output_0, %/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_938, %onnx::Conv_939)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_941, %onnx::Conv_942)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/Add_5_output_0 = Add(%/layers.6/Add_4_output_0, %/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/Add_6_output_0 = Add(%/layers.6/Add_5_output_0, %/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_6_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_944, %onnx::Conv_945)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_947, %onnx::Conv_948)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_950, %onnx::Conv_951)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_953, %onnx::Conv_954)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/Add_2_output_0, %/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_956, %onnx::Conv_957)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_959, %onnx::Conv_960)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/Add_5_output_0 = Add(%/layers.7/Add_4_output_0, %/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/Add_6_output_0 = Add(%/layers.7/Add_5_output_0, %/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_6_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_962, %onnx::Conv_963)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_965, %onnx::Conv_966)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_968, %onnx::Conv_969)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/Add_2_output_0, %/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_974, %onnx::Conv_975)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_977, %onnx::Conv_978)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/Add_5_output_0 = Add(%/layers.9/Add_4_output_0, %/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/Add_6_output_0 = Add(%/layers.9/Add_5_output_0, %/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_6_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/Add_2_output_0, %/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/Add_5_output_0 = Add(%/layers.10/Add_4_output_0, %/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/Add_6_output_0 = Add(%/layers.10/Add_5_output_0, %/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_6_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/Add_2_output_0, %/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/Add_5_output_0 = Add(%/layers.11/Add_4_output_0, %/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/Add_6_output_0 = Add(%/layers.11/Add_5_output_0, %/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_6_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/vertex_op.4/maxpool/MaxPool_output_0)
%849 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %849
}
|
val_accuracy
| 88.341343
| 1,752,442,880
| 5,742,730
|
{'zcp_epe_nas': 111.39278595791328, 'zcp_fisher': 194.7716522216797, 'zcp_flops': 28039086080.0, 'zcp_grad_norm': 331.2353515625, 'zcp_grasp': 41.8740234375, 'zcp_jacov': -16.05363472682599, 'zcp_l2_norm': 1226.695556640625, 'zcp_nwot': 235.60926730572777, 'zcp_params': 5742730.0, 'zcp_plain': -0.07780267298221501, 'zcp_snip': 2280.14599609375, 'zcp_synflow': 110.50585382593536, 'zcp_zen': 101.396728515625, 'zcp_val_accuracy': 0.931490361690521}
| |
NASBench101_125389
|
NASBench101
|
125389
|
4bc1eea94bef86bb978b4c4858351331
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_689[FLOAT, 128x3x3x3]
%onnx::Conv_690[FLOAT, 128]
%onnx::Conv_692[FLOAT, 43x128x1x1]
%onnx::Conv_693[FLOAT, 43]
%onnx::Conv_695[FLOAT, 43x43x3x3]
%onnx::Conv_698[FLOAT, 43x128x1x1]
%onnx::Conv_701[FLOAT, 42x128x1x1]
%onnx::Conv_702[FLOAT, 42]
%onnx::Conv_704[FLOAT, 43x128x1x1]
%onnx::Conv_707[FLOAT, 43x43x3x3]
%onnx::Conv_710[FLOAT, 43x128x1x1]
%onnx::Conv_713[FLOAT, 42x128x1x1]
%onnx::Conv_716[FLOAT, 43x128x1x1]
%onnx::Conv_719[FLOAT, 43x43x3x3]
%onnx::Conv_722[FLOAT, 43x128x1x1]
%onnx::Conv_725[FLOAT, 42x128x1x1]
%onnx::Conv_728[FLOAT, 86x128x1x1]
%onnx::Conv_729[FLOAT, 86]
%onnx::Conv_731[FLOAT, 86x86x3x3]
%onnx::Conv_734[FLOAT, 85x128x1x1]
%onnx::Conv_735[FLOAT, 85]
%onnx::Conv_737[FLOAT, 85x128x1x1]
%onnx::Conv_740[FLOAT, 86x256x1x1]
%onnx::Conv_743[FLOAT, 86x86x3x3]
%onnx::Conv_746[FLOAT, 85x256x1x1]
%onnx::Conv_749[FLOAT, 85x256x1x1]
%onnx::Conv_752[FLOAT, 86x256x1x1]
%onnx::Conv_755[FLOAT, 86x86x3x3]
%onnx::Conv_758[FLOAT, 85x256x1x1]
%onnx::Conv_761[FLOAT, 85x256x1x1]
%onnx::Conv_764[FLOAT, 171x256x1x1]
%onnx::Conv_765[FLOAT, 171]
%onnx::Conv_767[FLOAT, 171x171x3x3]
%onnx::Conv_770[FLOAT, 171x256x1x1]
%onnx::Conv_773[FLOAT, 170x256x1x1]
%onnx::Conv_774[FLOAT, 170]
%onnx::Conv_776[FLOAT, 171x512x1x1]
%onnx::Conv_779[FLOAT, 171x171x3x3]
%onnx::Conv_782[FLOAT, 171x512x1x1]
%onnx::Conv_785[FLOAT, 170x512x1x1]
%onnx::Conv_788[FLOAT, 171x512x1x1]
%onnx::Conv_791[FLOAT, 171x171x3x3]
%onnx::Conv_794[FLOAT, 171x512x1x1]
%onnx::Conv_797[FLOAT, 170x512x1x1]
) {
%onnx::Conv_798 = Identity(%onnx::Conv_774)
%onnx::Conv_795 = Identity(%onnx::Conv_765)
%onnx::Conv_792 = Identity(%onnx::Conv_765)
%onnx::Conv_789 = Identity(%onnx::Conv_765)
%onnx::Conv_786 = Identity(%onnx::Conv_774)
%onnx::Conv_783 = Identity(%onnx::Conv_765)
%onnx::Conv_780 = Identity(%onnx::Conv_765)
%onnx::Conv_777 = Identity(%onnx::Conv_765)
%onnx::Conv_771 = Identity(%onnx::Conv_765)
%onnx::Conv_768 = Identity(%onnx::Conv_765)
%onnx::Conv_762 = Identity(%onnx::Conv_735)
%onnx::Conv_759 = Identity(%onnx::Conv_735)
%onnx::Conv_756 = Identity(%onnx::Conv_729)
%onnx::Conv_753 = Identity(%onnx::Conv_729)
%onnx::Conv_750 = Identity(%onnx::Conv_735)
%onnx::Conv_747 = Identity(%onnx::Conv_735)
%onnx::Conv_744 = Identity(%onnx::Conv_729)
%onnx::Conv_741 = Identity(%onnx::Conv_729)
%onnx::Conv_738 = Identity(%onnx::Conv_735)
%onnx::Conv_732 = Identity(%onnx::Conv_729)
%onnx::Conv_726 = Identity(%onnx::Conv_702)
%onnx::Conv_723 = Identity(%onnx::Conv_693)
%onnx::Conv_720 = Identity(%onnx::Conv_693)
%onnx::Conv_717 = Identity(%onnx::Conv_693)
%onnx::Conv_714 = Identity(%onnx::Conv_702)
%onnx::Conv_711 = Identity(%onnx::Conv_693)
%onnx::Conv_708 = Identity(%onnx::Conv_693)
%onnx::Conv_705 = Identity(%onnx::Conv_693)
%onnx::Conv_699 = Identity(%onnx::Conv_693)
%onnx::Conv_696 = Identity(%onnx::Conv_693)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_689, %onnx::Conv_690)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_692, %onnx::Conv_693)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_695, %onnx::Conv_696)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_698, %onnx::Conv_699)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.1/Slice_output_0 = Slice(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0, %/layers.1/Constant_4_output_0, %/layers.1/Constant_2_output_0, %/layers.1/Constant_5_output_0)
%/layers.1/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/Slice_output_0, %/layers.1/Constant_6_output_0)
%/layers.1/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_701, %onnx::Conv_702)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.3/maxpool/MaxPool_output_0, %/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %/layers.1/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_704, %onnx::Conv_705)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_707, %onnx::Conv_708)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_710, %onnx::Conv_711)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.2/Slice_output_0 = Slice(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0, %/layers.2/Constant_4_output_0, %/layers.2/Constant_2_output_0, %/layers.2/Constant_5_output_0)
%/layers.2/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/Slice_output_0, %/layers.2/Constant_6_output_0)
%/layers.2/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_713, %onnx::Conv_714)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.3/maxpool/MaxPool_output_0, %/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %/layers.2/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_716, %onnx::Conv_717)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_719, %onnx::Conv_720)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_722, %onnx::Conv_723)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.3/Slice_output_0 = Slice(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0, %/layers.3/Constant_4_output_0, %/layers.3/Constant_2_output_0, %/layers.3/Constant_5_output_0)
%/layers.3/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/Slice_output_0, %/layers.3/Constant_6_output_0)
%/layers.3/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_725, %onnx::Conv_726)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.3/maxpool/MaxPool_output_0, %/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %/layers.3/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_728, %onnx::Conv_729)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_731, %onnx::Conv_732)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_734, %onnx::Conv_735)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.5/Slice_output_0 = Slice(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0, %/layers.5/Constant_4_output_0, %/layers.5/Constant_2_output_0, %/layers.5/Constant_5_output_0)
%/layers.5/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/Slice_output_0, %/layers.5/Constant_6_output_0)
%/layers.5/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_737, %onnx::Conv_738)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.3/maxpool/MaxPool_output_0, %/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %/layers.5/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_740, %onnx::Conv_741)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_743, %onnx::Conv_744)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_746, %onnx::Conv_747)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.6/Slice_output_0 = Slice(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0, %/layers.6/Constant_4_output_0, %/layers.6/Constant_2_output_0, %/layers.6/Constant_5_output_0)
%/layers.6/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/Slice_output_0, %/layers.6/Constant_6_output_0)
%/layers.6/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_749, %onnx::Conv_750)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.3/maxpool/MaxPool_output_0, %/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %/layers.6/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_752, %onnx::Conv_753)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_755, %onnx::Conv_756)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_758, %onnx::Conv_759)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.7/Slice_output_0 = Slice(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0, %/layers.7/Constant_4_output_0, %/layers.7/Constant_2_output_0, %/layers.7/Constant_5_output_0)
%/layers.7/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/Slice_output_0, %/layers.7/Constant_6_output_0)
%/layers.7/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_761, %onnx::Conv_762)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.3/maxpool/MaxPool_output_0, %/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %/layers.7/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_764, %onnx::Conv_765)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_767, %onnx::Conv_768)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_770, %onnx::Conv_771)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.9/Slice_output_0 = Slice(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0, %/layers.9/Constant_4_output_0, %/layers.9/Constant_2_output_0, %/layers.9/Constant_5_output_0)
%/layers.9/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/Slice_output_0, %/layers.9/Constant_6_output_0)
%/layers.9/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_773, %onnx::Conv_774)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.3/maxpool/MaxPool_output_0, %/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %/layers.9/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_776, %onnx::Conv_777)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_779, %onnx::Conv_780)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_782, %onnx::Conv_783)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.10/Slice_output_0 = Slice(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0, %/layers.10/Constant_4_output_0, %/layers.10/Constant_2_output_0, %/layers.10/Constant_5_output_0)
%/layers.10/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/Slice_output_0, %/layers.10/Constant_6_output_0)
%/layers.10/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_785, %onnx::Conv_786)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.3/maxpool/MaxPool_output_0, %/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %/layers.10/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_788, %onnx::Conv_789)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.11/Slice_output_0 = Slice(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0, %/layers.11/Constant_4_output_0, %/layers.11/Constant_2_output_0, %/layers.11/Constant_5_output_0)
%/layers.11/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/Slice_output_0, %/layers.11/Constant_6_output_0)
%/layers.11/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_797, %onnx::Conv_798)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.3/maxpool/MaxPool_output_0, %/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %/layers.11/vertex_op.4/maxpool/MaxPool_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%687 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %687
}
|
val_accuracy
| 91.40625
| 589,466,496
| 1,923,492
|
{'zcp_epe_nas': 131.04198754693206, 'zcp_fisher': 5.117156028747559, 'zcp_flops': 9431463936.0, 'zcp_grad_norm': 44.10383224487305, 'zcp_grasp': -5.8883056640625, 'zcp_jacov': -16.04140579882982, 'zcp_l2_norm': 714.5977783203125, 'zcp_nwot': 212.49711860574072, 'zcp_params': 1923492.0, 'zcp_plain': 0.050067059695720006, 'zcp_snip': 235.5615234375, 'zcp_synflow': 63.71108859281378, 'zcp_zen': 67.71805572509766, 'zcp_val_accuracy': 0.8964343070983881}
| |
NASBench101_158555
|
NASBench101
|
158555
|
5ffa7929e344314fc7b70896b7a28d4c
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_719[FLOAT, 128x3x3x3]
%onnx::Conv_720[FLOAT, 128]
%onnx::Conv_722[FLOAT, 43x128x1x1]
%onnx::Conv_723[FLOAT, 43]
%onnx::Conv_725[FLOAT, 43x43x3x3]
%onnx::Conv_728[FLOAT, 42x42x3x3]
%onnx::Conv_729[FLOAT, 42]
%onnx::Conv_731[FLOAT, 128x128x1x1]
%onnx::Conv_734[FLOAT, 43x128x1x1]
%onnx::Conv_737[FLOAT, 43x43x3x3]
%onnx::Conv_740[FLOAT, 42x42x3x3]
%onnx::Conv_743[FLOAT, 128x128x1x1]
%onnx::Conv_746[FLOAT, 43x128x1x1]
%onnx::Conv_749[FLOAT, 43x43x3x3]
%onnx::Conv_752[FLOAT, 42x42x3x3]
%onnx::Conv_755[FLOAT, 128x128x1x1]
%onnx::Conv_758[FLOAT, 86x128x1x1]
%onnx::Conv_759[FLOAT, 86]
%onnx::Conv_761[FLOAT, 85x85x3x3]
%onnx::Conv_762[FLOAT, 85]
%onnx::Conv_764[FLOAT, 85x85x3x3]
%onnx::Conv_767[FLOAT, 256x128x1x1]
%onnx::Conv_768[FLOAT, 256]
%onnx::Conv_770[FLOAT, 86x256x1x1]
%onnx::Conv_773[FLOAT, 85x85x3x3]
%onnx::Conv_776[FLOAT, 85x85x3x3]
%onnx::Conv_779[FLOAT, 256x256x1x1]
%onnx::Conv_782[FLOAT, 86x256x1x1]
%onnx::Conv_785[FLOAT, 85x85x3x3]
%onnx::Conv_788[FLOAT, 85x85x3x3]
%onnx::Conv_791[FLOAT, 256x256x1x1]
%onnx::Conv_794[FLOAT, 171x256x1x1]
%onnx::Conv_795[FLOAT, 171]
%onnx::Conv_797[FLOAT, 171x171x3x3]
%onnx::Conv_800[FLOAT, 170x170x3x3]
%onnx::Conv_801[FLOAT, 170]
%onnx::Conv_803[FLOAT, 512x256x1x1]
%onnx::Conv_804[FLOAT, 512]
%onnx::Conv_806[FLOAT, 171x512x1x1]
%onnx::Conv_809[FLOAT, 171x171x3x3]
%onnx::Conv_812[FLOAT, 170x170x3x3]
%onnx::Conv_815[FLOAT, 512x512x1x1]
%onnx::Conv_818[FLOAT, 171x512x1x1]
%onnx::Conv_821[FLOAT, 171x171x3x3]
%onnx::Conv_824[FLOAT, 170x170x3x3]
%onnx::Conv_827[FLOAT, 512x512x1x1]
) {
%onnx::Conv_828 = Identity(%onnx::Conv_804)
%onnx::Conv_825 = Identity(%onnx::Conv_801)
%onnx::Conv_822 = Identity(%onnx::Conv_795)
%onnx::Conv_819 = Identity(%onnx::Conv_795)
%onnx::Conv_816 = Identity(%onnx::Conv_804)
%onnx::Conv_813 = Identity(%onnx::Conv_801)
%onnx::Conv_810 = Identity(%onnx::Conv_795)
%onnx::Conv_807 = Identity(%onnx::Conv_795)
%onnx::Conv_798 = Identity(%onnx::Conv_795)
%onnx::Conv_792 = Identity(%onnx::Conv_768)
%onnx::Conv_789 = Identity(%onnx::Conv_762)
%onnx::Conv_786 = Identity(%onnx::Conv_762)
%onnx::Conv_783 = Identity(%onnx::Conv_759)
%onnx::Conv_780 = Identity(%onnx::Conv_768)
%onnx::Conv_777 = Identity(%onnx::Conv_762)
%onnx::Conv_774 = Identity(%onnx::Conv_762)
%onnx::Conv_771 = Identity(%onnx::Conv_759)
%onnx::Conv_765 = Identity(%onnx::Conv_762)
%onnx::Conv_756 = Identity(%onnx::Conv_720)
%onnx::Conv_753 = Identity(%onnx::Conv_729)
%onnx::Conv_750 = Identity(%onnx::Conv_723)
%onnx::Conv_747 = Identity(%onnx::Conv_723)
%onnx::Conv_744 = Identity(%onnx::Conv_720)
%onnx::Conv_741 = Identity(%onnx::Conv_729)
%onnx::Conv_738 = Identity(%onnx::Conv_723)
%onnx::Conv_735 = Identity(%onnx::Conv_723)
%onnx::Conv_732 = Identity(%onnx::Conv_720)
%onnx::Conv_726 = Identity(%onnx::Conv_723)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_719, %onnx::Conv_720)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_722, %onnx::Conv_723)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_725, %onnx::Conv_726)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.1/Slice_output_0 = Slice(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0, %/layers.1/Constant_4_output_0, %/layers.1/Constant_2_output_0, %/layers.1/Constant_5_output_0)
%/layers.1/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.1/Slice_1_output_0 = Slice(%/layers.1/vertex_op.3/maxpool/MaxPool_output_0, %/layers.1/Constant_7_output_0, %/layers.1/Constant_8_output_0, %/layers.1/Constant_6_output_0, %/layers.1/Constant_9_output_0)
%/layers.1/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/Slice_output_0, %/layers.1/Constant_10_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/Add_2_output_0, %/layers.1/Slice_1_output_0)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_728, %onnx::Conv_729)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/vertex_op.3/maxpool/MaxPool_output_0, %/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_731, %onnx::Conv_732)
%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/Concat_output_0, %/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_734, %onnx::Conv_735)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_737, %onnx::Conv_738)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.2/Slice_output_0 = Slice(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0, %/layers.2/Constant_4_output_0, %/layers.2/Constant_2_output_0, %/layers.2/Constant_5_output_0)
%/layers.2/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.2/Slice_1_output_0 = Slice(%/layers.2/vertex_op.3/maxpool/MaxPool_output_0, %/layers.2/Constant_7_output_0, %/layers.2/Constant_8_output_0, %/layers.2/Constant_6_output_0, %/layers.2/Constant_9_output_0)
%/layers.2/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/Slice_output_0, %/layers.2/Constant_10_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/Add_2_output_0, %/layers.2/Slice_1_output_0)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_740, %onnx::Conv_741)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/vertex_op.3/maxpool/MaxPool_output_0, %/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_743, %onnx::Conv_744)
%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/Concat_output_0, %/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_746, %onnx::Conv_747)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_749, %onnx::Conv_750)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.3/Slice_output_0 = Slice(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0, %/layers.3/Constant_4_output_0, %/layers.3/Constant_2_output_0, %/layers.3/Constant_5_output_0)
%/layers.3/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.3/Slice_1_output_0 = Slice(%/layers.3/vertex_op.3/maxpool/MaxPool_output_0, %/layers.3/Constant_7_output_0, %/layers.3/Constant_8_output_0, %/layers.3/Constant_6_output_0, %/layers.3/Constant_9_output_0)
%/layers.3/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/Slice_output_0, %/layers.3/Constant_10_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/Add_2_output_0, %/layers.3/Slice_1_output_0)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_752, %onnx::Conv_753)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/vertex_op.3/maxpool/MaxPool_output_0, %/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_755, %onnx::Conv_756)
%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/Concat_output_0, %/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Add_4_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_758, %onnx::Conv_759)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.5/Slice_output_0 = Slice(%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/Constant_2_output_0, %/layers.5/Constant_3_output_0, %/layers.5/Constant_1_output_0, %/layers.5/Constant_4_output_0)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Slice_output_0, %onnx::Conv_761, %onnx::Conv_762)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_5_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_5_output_0)
%/layers.5/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_6_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/Add_2_output_0, %/layers.5/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_764, %onnx::Conv_765)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/vertex_op.3/maxpool/MaxPool_output_0, %/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_767, %onnx::Conv_768)
%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/Concat_output_0, %/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_770, %onnx::Conv_771)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.6/Slice_output_0 = Slice(%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/Constant_2_output_0, %/layers.6/Constant_3_output_0, %/layers.6/Constant_1_output_0, %/layers.6/Constant_4_output_0)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Slice_output_0, %onnx::Conv_773, %onnx::Conv_774)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_5_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_5_output_0)
%/layers.6/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_6_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/Add_2_output_0, %/layers.6/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_776, %onnx::Conv_777)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/vertex_op.3/maxpool/MaxPool_output_0, %/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_779, %onnx::Conv_780)
%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/Concat_output_0, %/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_782, %onnx::Conv_783)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.7/Slice_output_0 = Slice(%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/Constant_2_output_0, %/layers.7/Constant_3_output_0, %/layers.7/Constant_1_output_0, %/layers.7/Constant_4_output_0)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Slice_output_0, %onnx::Conv_785, %onnx::Conv_786)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_5_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_5_output_0)
%/layers.7/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/Constant_6_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_6_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/Add_2_output_0, %/layers.7/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_788, %onnx::Conv_789)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/vertex_op.3/maxpool/MaxPool_output_0, %/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/Concat_output_0, %/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Add_4_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_797, %onnx::Conv_798)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.9/Slice_output_0 = Slice(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0, %/layers.9/Constant_4_output_0, %/layers.9/Constant_2_output_0, %/layers.9/Constant_5_output_0)
%/layers.9/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.9/Slice_1_output_0 = Slice(%/layers.9/vertex_op.3/maxpool/MaxPool_output_0, %/layers.9/Constant_7_output_0, %/layers.9/Constant_8_output_0, %/layers.9/Constant_6_output_0, %/layers.9/Constant_9_output_0)
%/layers.9/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/Slice_output_0, %/layers.9/Constant_10_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/Add_2_output_0, %/layers.9/Slice_1_output_0)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_800, %onnx::Conv_801)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/vertex_op.3/maxpool/MaxPool_output_0, %/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_803, %onnx::Conv_804)
%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/Concat_output_0, %/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_806, %onnx::Conv_807)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_809, %onnx::Conv_810)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.10/Slice_output_0 = Slice(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0, %/layers.10/Constant_4_output_0, %/layers.10/Constant_2_output_0, %/layers.10/Constant_5_output_0)
%/layers.10/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.10/Slice_1_output_0 = Slice(%/layers.10/vertex_op.3/maxpool/MaxPool_output_0, %/layers.10/Constant_7_output_0, %/layers.10/Constant_8_output_0, %/layers.10/Constant_6_output_0, %/layers.10/Constant_9_output_0)
%/layers.10/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/Slice_output_0, %/layers.10/Constant_10_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/Add_2_output_0, %/layers.10/Slice_1_output_0)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_812, %onnx::Conv_813)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/vertex_op.3/maxpool/MaxPool_output_0, %/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_815, %onnx::Conv_816)
%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/Concat_output_0, %/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_818, %onnx::Conv_819)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_821, %onnx::Conv_822)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.11/Slice_output_0 = Slice(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0, %/layers.11/Constant_4_output_0, %/layers.11/Constant_2_output_0, %/layers.11/Constant_5_output_0)
%/layers.11/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.11/Slice_1_output_0 = Slice(%/layers.11/vertex_op.3/maxpool/MaxPool_output_0, %/layers.11/Constant_7_output_0, %/layers.11/Constant_8_output_0, %/layers.11/Constant_6_output_0, %/layers.11/Constant_9_output_0)
%/layers.11/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/Slice_output_0, %/layers.11/Constant_10_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/Add_2_output_0, %/layers.11/Slice_1_output_0)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_824, %onnx::Conv_825)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/vertex_op.3/maxpool/MaxPool_output_0, %/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_827, %onnx::Conv_828)
%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/Concat_output_0, %/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Add_4_output_0)
%717 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %717
}
|
val_accuracy
| 90.414661
| 978,072,448
| 3,235,886
|
{'zcp_epe_nas': 106.99211493210738, 'zcp_fisher': 3.234248161315918, 'zcp_flops': 15649159168.0, 'zcp_grad_norm': 36.57662582397461, 'zcp_grasp': 2.910858154296875, 'zcp_jacov': -16.065425730477898, 'zcp_l2_norm': 639.3207397460938, 'zcp_nwot': 218.1015016779268, 'zcp_params': 3235886.0, 'zcp_plain': -0.08127563446760101, 'zcp_snip': 212.7371063232422, 'zcp_synflow': 94.7113379445774, 'zcp_zen': 81.87544250488281, 'zcp_val_accuracy': 0.9318910241127011}
| |
NASBench101_153001
|
NASBench101
|
153001
|
5c9445d43e83498c3524f7ded8b633d4
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_770[FLOAT, 128x3x3x3]
%onnx::Conv_771[FLOAT, 128]
%onnx::Conv_773[FLOAT, 128x128x1x1]
%onnx::Conv_776[FLOAT, 128x128x3x3]
%onnx::Conv_779[FLOAT, 128x128x1x1]
%onnx::Conv_782[FLOAT, 128x128x3x3]
%onnx::Conv_785[FLOAT, 128x128x1x1]
%onnx::Conv_788[FLOAT, 128x128x1x1]
%onnx::Conv_791[FLOAT, 128x128x3x3]
%onnx::Conv_794[FLOAT, 128x128x1x1]
%onnx::Conv_797[FLOAT, 128x128x3x3]
%onnx::Conv_800[FLOAT, 128x128x1x1]
%onnx::Conv_803[FLOAT, 128x128x1x1]
%onnx::Conv_806[FLOAT, 128x128x3x3]
%onnx::Conv_809[FLOAT, 128x128x1x1]
%onnx::Conv_812[FLOAT, 128x128x3x3]
%onnx::Conv_815[FLOAT, 128x128x1x1]
%onnx::Conv_818[FLOAT, 256x128x1x1]
%onnx::Conv_819[FLOAT, 256]
%onnx::Conv_821[FLOAT, 256x256x3x3]
%onnx::Conv_824[FLOAT, 256x128x1x1]
%onnx::Conv_827[FLOAT, 256x256x3x3]
%onnx::Conv_830[FLOAT, 256x128x1x1]
%onnx::Conv_833[FLOAT, 256x256x1x1]
%onnx::Conv_836[FLOAT, 256x256x3x3]
%onnx::Conv_839[FLOAT, 256x256x1x1]
%onnx::Conv_842[FLOAT, 256x256x3x3]
%onnx::Conv_845[FLOAT, 256x256x1x1]
%onnx::Conv_848[FLOAT, 256x256x1x1]
%onnx::Conv_851[FLOAT, 256x256x3x3]
%onnx::Conv_854[FLOAT, 256x256x1x1]
%onnx::Conv_857[FLOAT, 256x256x3x3]
%onnx::Conv_860[FLOAT, 256x256x1x1]
%onnx::Conv_863[FLOAT, 512x256x1x1]
%onnx::Conv_864[FLOAT, 512]
%onnx::Conv_866[FLOAT, 512x512x3x3]
%onnx::Conv_869[FLOAT, 512x256x1x1]
%onnx::Conv_872[FLOAT, 512x512x3x3]
%onnx::Conv_875[FLOAT, 512x256x1x1]
%onnx::Conv_878[FLOAT, 512x512x1x1]
%onnx::Conv_881[FLOAT, 512x512x3x3]
%onnx::Conv_884[FLOAT, 512x512x1x1]
%onnx::Conv_887[FLOAT, 512x512x3x3]
%onnx::Conv_890[FLOAT, 512x512x1x1]
%onnx::Conv_893[FLOAT, 512x512x1x1]
%onnx::Conv_896[FLOAT, 512x512x3x3]
%onnx::Conv_899[FLOAT, 512x512x1x1]
%onnx::Conv_902[FLOAT, 512x512x3x3]
%onnx::Conv_905[FLOAT, 512x512x1x1]
) {
%onnx::Conv_906 = Identity(%onnx::Conv_864)
%onnx::Conv_903 = Identity(%onnx::Conv_864)
%onnx::Conv_900 = Identity(%onnx::Conv_864)
%onnx::Conv_897 = Identity(%onnx::Conv_864)
%onnx::Conv_894 = Identity(%onnx::Conv_864)
%onnx::Conv_891 = Identity(%onnx::Conv_864)
%onnx::Conv_888 = Identity(%onnx::Conv_864)
%onnx::Conv_885 = Identity(%onnx::Conv_864)
%onnx::Conv_882 = Identity(%onnx::Conv_864)
%onnx::Conv_879 = Identity(%onnx::Conv_864)
%onnx::Conv_876 = Identity(%onnx::Conv_864)
%onnx::Conv_873 = Identity(%onnx::Conv_864)
%onnx::Conv_870 = Identity(%onnx::Conv_864)
%onnx::Conv_867 = Identity(%onnx::Conv_864)
%onnx::Conv_861 = Identity(%onnx::Conv_819)
%onnx::Conv_858 = Identity(%onnx::Conv_819)
%onnx::Conv_855 = Identity(%onnx::Conv_819)
%onnx::Conv_852 = Identity(%onnx::Conv_819)
%onnx::Conv_849 = Identity(%onnx::Conv_819)
%onnx::Conv_846 = Identity(%onnx::Conv_819)
%onnx::Conv_843 = Identity(%onnx::Conv_819)
%onnx::Conv_840 = Identity(%onnx::Conv_819)
%onnx::Conv_837 = Identity(%onnx::Conv_819)
%onnx::Conv_834 = Identity(%onnx::Conv_819)
%onnx::Conv_831 = Identity(%onnx::Conv_819)
%onnx::Conv_828 = Identity(%onnx::Conv_819)
%onnx::Conv_825 = Identity(%onnx::Conv_819)
%onnx::Conv_822 = Identity(%onnx::Conv_819)
%onnx::Conv_816 = Identity(%onnx::Conv_771)
%onnx::Conv_813 = Identity(%onnx::Conv_771)
%onnx::Conv_810 = Identity(%onnx::Conv_771)
%onnx::Conv_807 = Identity(%onnx::Conv_771)
%onnx::Conv_804 = Identity(%onnx::Conv_771)
%onnx::Conv_801 = Identity(%onnx::Conv_771)
%onnx::Conv_798 = Identity(%onnx::Conv_771)
%onnx::Conv_795 = Identity(%onnx::Conv_771)
%onnx::Conv_792 = Identity(%onnx::Conv_771)
%onnx::Conv_789 = Identity(%onnx::Conv_771)
%onnx::Conv_786 = Identity(%onnx::Conv_771)
%onnx::Conv_783 = Identity(%onnx::Conv_771)
%onnx::Conv_780 = Identity(%onnx::Conv_771)
%onnx::Conv_777 = Identity(%onnx::Conv_771)
%onnx::Conv_774 = Identity(%onnx::Conv_771)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_770, %onnx::Conv_771)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_773, %onnx::Conv_774)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_776, %onnx::Conv_777)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_779, %onnx::Conv_780)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/Add_2_output_0, %/layers.1/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.1/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/Add_5_output_0 = Add(%/layers.1/Add_4_output_0, %/layers.1/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.1/Add_6_output_0 = Add(%/layers.1/Add_5_output_0, %/layers.1/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_6_output_0, %onnx::Conv_782, %onnx::Conv_783)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_785, %onnx::Conv_786)
%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_7_output_0 = Add(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_7_output_0, %onnx::Conv_788, %onnx::Conv_789)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_7_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/Add_2_output_0, %/layers.2/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.2/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/Add_5_output_0 = Add(%/layers.2/Add_4_output_0, %/layers.2/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.2/Add_6_output_0 = Add(%/layers.2/Add_5_output_0, %/layers.2/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_6_output_0, %onnx::Conv_797, %onnx::Conv_798)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_7_output_0, %onnx::Conv_800, %onnx::Conv_801)
%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_7_output_0 = Add(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_7_output_0, %onnx::Conv_803, %onnx::Conv_804)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_806, %onnx::Conv_807)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_7_output_0, %onnx::Conv_809, %onnx::Conv_810)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/Add_2_output_0, %/layers.3/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.3/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/Add_5_output_0 = Add(%/layers.3/Add_4_output_0, %/layers.3/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.3/Add_6_output_0 = Add(%/layers.3/Add_5_output_0, %/layers.3/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_6_output_0, %onnx::Conv_812, %onnx::Conv_813)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_7_output_0, %onnx::Conv_815, %onnx::Conv_816)
%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_7_output_0 = Add(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Add_7_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_818, %onnx::Conv_819)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_821, %onnx::Conv_822)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_824, %onnx::Conv_825)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/Add_2_output_0, %/layers.5/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.5/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/Add_5_output_0 = Add(%/layers.5/Add_4_output_0, %/layers.5/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.5/Add_6_output_0 = Add(%/layers.5/Add_5_output_0, %/layers.5/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_6_output_0, %onnx::Conv_827, %onnx::Conv_828)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_830, %onnx::Conv_831)
%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_7_output_0 = Add(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_7_output_0, %onnx::Conv_833, %onnx::Conv_834)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_836, %onnx::Conv_837)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_7_output_0, %onnx::Conv_839, %onnx::Conv_840)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/Add_2_output_0, %/layers.6/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.6/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/Add_5_output_0 = Add(%/layers.6/Add_4_output_0, %/layers.6/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.6/Add_6_output_0 = Add(%/layers.6/Add_5_output_0, %/layers.6/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_6_output_0, %onnx::Conv_842, %onnx::Conv_843)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_7_output_0, %onnx::Conv_845, %onnx::Conv_846)
%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_7_output_0 = Add(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_7_output_0, %onnx::Conv_848, %onnx::Conv_849)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_851, %onnx::Conv_852)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_7_output_0, %onnx::Conv_854, %onnx::Conv_855)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/Add_2_output_0, %/layers.7/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.7/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/Add_5_output_0 = Add(%/layers.7/Add_4_output_0, %/layers.7/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.7/Add_6_output_0 = Add(%/layers.7/Add_5_output_0, %/layers.7/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_6_output_0, %onnx::Conv_857, %onnx::Conv_858)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_7_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_7_output_0 = Add(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Add_7_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/Add_2_output_0, %/layers.9/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.9/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/Add_5_output_0 = Add(%/layers.9/Add_4_output_0, %/layers.9/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.9/Add_6_output_0 = Add(%/layers.9/Add_5_output_0, %/layers.9/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_6_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_7_output_0 = Add(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_7_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_7_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/Add_2_output_0, %/layers.10/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.10/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/Add_5_output_0 = Add(%/layers.10/Add_4_output_0, %/layers.10/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.10/Add_6_output_0 = Add(%/layers.10/Add_5_output_0, %/layers.10/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_6_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_7_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_7_output_0 = Add(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_7_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_7_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/Add_2_output_0, %/layers.11/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.11/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/Add_5_output_0 = Add(%/layers.11/Add_4_output_0, %/layers.11/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.11/Add_6_output_0 = Add(%/layers.11/Add_5_output_0, %/layers.11/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_6_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_7_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_7_output_0 = Add(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Add_7_output_0)
%768 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %768
}
|
val_accuracy
| 93.229169
| 6,276,786,176
| 21,220,234
|
{'zcp_epe_nas': 162.12086664151454, 'zcp_fisher': 2.505191326141357, 'zcp_flops': 100428578816.0, 'zcp_grad_norm': 29.234172821044922, 'zcp_grasp': -1.6302471160888672, 'zcp_jacov': -16.046301399432537, 'zcp_l2_norm': 1014.662109375, 'zcp_nwot': 231.85761343741484, 'zcp_params': 21220234.0, 'zcp_plain': 0.067835576832294, 'zcp_snip': 265.3919982910156, 'zcp_synflow': 109.48105537492872, 'zcp_zen': 106.97577667236328, 'zcp_val_accuracy': 0.9003405570983881}
| |
NASBench101_221861
|
NASBench101
|
221861
|
8675eae2f37fcbf0a989fa5f65071a11
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_860[FLOAT, 128x3x3x3]
%onnx::Conv_861[FLOAT, 128]
%onnx::Conv_863[FLOAT, 32x128x1x1]
%onnx::Conv_864[FLOAT, 32]
%onnx::Conv_866[FLOAT, 32x32x1x1]
%onnx::Conv_869[FLOAT, 32x32x3x3]
%onnx::Conv_872[FLOAT, 32x32x3x3]
%onnx::Conv_875[FLOAT, 32x32x3x3]
%onnx::Conv_878[FLOAT, 32x32x3x3]
%onnx::Conv_881[FLOAT, 32x128x1x1]
%onnx::Conv_884[FLOAT, 32x32x1x1]
%onnx::Conv_887[FLOAT, 32x32x3x3]
%onnx::Conv_890[FLOAT, 32x32x3x3]
%onnx::Conv_893[FLOAT, 32x32x3x3]
%onnx::Conv_896[FLOAT, 32x32x3x3]
%onnx::Conv_899[FLOAT, 32x128x1x1]
%onnx::Conv_902[FLOAT, 32x32x1x1]
%onnx::Conv_905[FLOAT, 32x32x3x3]
%onnx::Conv_908[FLOAT, 32x32x3x3]
%onnx::Conv_911[FLOAT, 32x32x3x3]
%onnx::Conv_914[FLOAT, 32x32x3x3]
%onnx::Conv_917[FLOAT, 64x128x1x1]
%onnx::Conv_918[FLOAT, 64]
%onnx::Conv_920[FLOAT, 64x64x1x1]
%onnx::Conv_923[FLOAT, 64x64x3x3]
%onnx::Conv_926[FLOAT, 64x64x3x3]
%onnx::Conv_929[FLOAT, 64x64x3x3]
%onnx::Conv_932[FLOAT, 64x64x3x3]
%onnx::Conv_935[FLOAT, 64x256x1x1]
%onnx::Conv_938[FLOAT, 64x64x1x1]
%onnx::Conv_941[FLOAT, 64x64x3x3]
%onnx::Conv_944[FLOAT, 64x64x3x3]
%onnx::Conv_947[FLOAT, 64x64x3x3]
%onnx::Conv_950[FLOAT, 64x64x3x3]
%onnx::Conv_953[FLOAT, 64x256x1x1]
%onnx::Conv_956[FLOAT, 64x64x1x1]
%onnx::Conv_959[FLOAT, 64x64x3x3]
%onnx::Conv_962[FLOAT, 64x64x3x3]
%onnx::Conv_965[FLOAT, 64x64x3x3]
%onnx::Conv_968[FLOAT, 64x64x3x3]
%onnx::Conv_971[FLOAT, 128x256x1x1]
%onnx::Conv_974[FLOAT, 128x128x1x1]
%onnx::Conv_977[FLOAT, 128x128x3x3]
%onnx::Conv_980[FLOAT, 128x128x3x3]
%onnx::Conv_983[FLOAT, 128x128x3x3]
%onnx::Conv_986[FLOAT, 128x128x3x3]
%onnx::Conv_989[FLOAT, 128x512x1x1]
%onnx::Conv_992[FLOAT, 128x128x1x1]
%onnx::Conv_995[FLOAT, 128x128x3x3]
%onnx::Conv_998[FLOAT, 128x128x3x3]
%onnx::Conv_1001[FLOAT, 128x128x3x3]
%onnx::Conv_1004[FLOAT, 128x128x3x3]
%onnx::Conv_1007[FLOAT, 128x512x1x1]
%onnx::Conv_1010[FLOAT, 128x128x1x1]
%onnx::Conv_1013[FLOAT, 128x128x3x3]
%onnx::Conv_1016[FLOAT, 128x128x3x3]
%onnx::Conv_1019[FLOAT, 128x128x3x3]
%onnx::Conv_1022[FLOAT, 128x128x3x3]
) {
%onnx::Conv_1023 = Identity(%onnx::Conv_861)
%onnx::Conv_1020 = Identity(%onnx::Conv_861)
%onnx::Conv_1017 = Identity(%onnx::Conv_861)
%onnx::Conv_1014 = Identity(%onnx::Conv_861)
%onnx::Conv_1011 = Identity(%onnx::Conv_861)
%onnx::Conv_1008 = Identity(%onnx::Conv_861)
%onnx::Conv_1005 = Identity(%onnx::Conv_861)
%onnx::Conv_1002 = Identity(%onnx::Conv_861)
%onnx::Conv_999 = Identity(%onnx::Conv_861)
%onnx::Conv_996 = Identity(%onnx::Conv_861)
%onnx::Conv_993 = Identity(%onnx::Conv_861)
%onnx::Conv_990 = Identity(%onnx::Conv_861)
%onnx::Conv_987 = Identity(%onnx::Conv_861)
%onnx::Conv_984 = Identity(%onnx::Conv_861)
%onnx::Conv_981 = Identity(%onnx::Conv_861)
%onnx::Conv_978 = Identity(%onnx::Conv_861)
%onnx::Conv_975 = Identity(%onnx::Conv_861)
%onnx::Conv_972 = Identity(%onnx::Conv_861)
%onnx::Conv_969 = Identity(%onnx::Conv_918)
%onnx::Conv_966 = Identity(%onnx::Conv_918)
%onnx::Conv_963 = Identity(%onnx::Conv_918)
%onnx::Conv_960 = Identity(%onnx::Conv_918)
%onnx::Conv_957 = Identity(%onnx::Conv_918)
%onnx::Conv_954 = Identity(%onnx::Conv_918)
%onnx::Conv_951 = Identity(%onnx::Conv_918)
%onnx::Conv_948 = Identity(%onnx::Conv_918)
%onnx::Conv_945 = Identity(%onnx::Conv_918)
%onnx::Conv_942 = Identity(%onnx::Conv_918)
%onnx::Conv_939 = Identity(%onnx::Conv_918)
%onnx::Conv_936 = Identity(%onnx::Conv_918)
%onnx::Conv_933 = Identity(%onnx::Conv_918)
%onnx::Conv_930 = Identity(%onnx::Conv_918)
%onnx::Conv_927 = Identity(%onnx::Conv_918)
%onnx::Conv_924 = Identity(%onnx::Conv_918)
%onnx::Conv_921 = Identity(%onnx::Conv_918)
%onnx::Conv_915 = Identity(%onnx::Conv_864)
%onnx::Conv_912 = Identity(%onnx::Conv_864)
%onnx::Conv_909 = Identity(%onnx::Conv_864)
%onnx::Conv_906 = Identity(%onnx::Conv_864)
%onnx::Conv_903 = Identity(%onnx::Conv_864)
%onnx::Conv_900 = Identity(%onnx::Conv_864)
%onnx::Conv_897 = Identity(%onnx::Conv_864)
%onnx::Conv_894 = Identity(%onnx::Conv_864)
%onnx::Conv_891 = Identity(%onnx::Conv_864)
%onnx::Conv_888 = Identity(%onnx::Conv_864)
%onnx::Conv_885 = Identity(%onnx::Conv_864)
%onnx::Conv_882 = Identity(%onnx::Conv_864)
%onnx::Conv_879 = Identity(%onnx::Conv_864)
%onnx::Conv_876 = Identity(%onnx::Conv_864)
%onnx::Conv_873 = Identity(%onnx::Conv_864)
%onnx::Conv_870 = Identity(%onnx::Conv_864)
%onnx::Conv_867 = Identity(%onnx::Conv_864)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_860, %onnx::Conv_861)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_4_output_0)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_4_output_0)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_908, %onnx::Conv_909)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_911, %onnx::Conv_912)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_4_output_0)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_4_output_0, %onnx::Conv_914, %onnx::Conv_915)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_917, %onnx::Conv_918)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_920, %onnx::Conv_921)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_923, %onnx::Conv_924)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_926, %onnx::Conv_927)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_929, %onnx::Conv_930)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_4_output_0)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_932, %onnx::Conv_933)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_935, %onnx::Conv_936)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_938, %onnx::Conv_939)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_941, %onnx::Conv_942)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_944, %onnx::Conv_945)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_947, %onnx::Conv_948)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_4_output_0)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_950, %onnx::Conv_951)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_953, %onnx::Conv_954)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_956, %onnx::Conv_957)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_959, %onnx::Conv_960)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_962, %onnx::Conv_963)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_965, %onnx::Conv_966)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_4_output_0)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_4_output_0, %onnx::Conv_968, %onnx::Conv_969)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_974, %onnx::Conv_975)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_977, %onnx::Conv_978)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_4_output_0)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_4_output_0)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_4_output_0)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_4_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%858 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %858
}
|
val_accuracy
| 91.997194
| 781,854,720
| 2,620,938
|
{'zcp_epe_nas': 64.48448264858972, 'zcp_fisher': 87.80416870117188, 'zcp_flops': 12509675520.0, 'zcp_grad_norm': 269.6516418457031, 'zcp_grasp': 776.947265625, 'zcp_jacov': -16.054678473421575, 'zcp_l2_norm': 728.9632568359375, 'zcp_nwot': 214.5185135623255, 'zcp_params': 2620938.0, 'zcp_plain': 0.005857831332832001, 'zcp_snip': 911.6822509765625, 'zcp_synflow': 133.96001990013585, 'zcp_zen': 91.77005767822266, 'zcp_val_accuracy': 0.915865361690521}
| |
NASBench101_102166
|
NASBench101
|
102166
|
3dce3bc07d2701d6aa94d2e4c91f9fe1
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_743[FLOAT, 128x3x3x3]
%onnx::Conv_744[FLOAT, 128]
%onnx::Conv_746[FLOAT, 128x128x1x1]
%onnx::Conv_749[FLOAT, 128x128x1x1]
%onnx::Conv_752[FLOAT, 128x128x1x1]
%onnx::Conv_755[FLOAT, 128x128x1x1]
%onnx::Conv_758[FLOAT, 128x128x3x3]
%onnx::Conv_761[FLOAT, 128x128x1x1]
%onnx::Conv_764[FLOAT, 128x128x1x1]
%onnx::Conv_767[FLOAT, 128x128x1x1]
%onnx::Conv_770[FLOAT, 128x128x1x1]
%onnx::Conv_773[FLOAT, 128x128x3x3]
%onnx::Conv_776[FLOAT, 128x128x1x1]
%onnx::Conv_779[FLOAT, 128x128x1x1]
%onnx::Conv_782[FLOAT, 128x128x1x1]
%onnx::Conv_785[FLOAT, 128x128x1x1]
%onnx::Conv_788[FLOAT, 128x128x3x3]
%onnx::Conv_791[FLOAT, 256x128x1x1]
%onnx::Conv_792[FLOAT, 256]
%onnx::Conv_794[FLOAT, 256x256x1x1]
%onnx::Conv_797[FLOAT, 256x256x1x1]
%onnx::Conv_800[FLOAT, 256x128x1x1]
%onnx::Conv_803[FLOAT, 256x256x3x3]
%onnx::Conv_806[FLOAT, 256x256x1x1]
%onnx::Conv_809[FLOAT, 256x256x1x1]
%onnx::Conv_812[FLOAT, 256x256x1x1]
%onnx::Conv_815[FLOAT, 256x256x1x1]
%onnx::Conv_818[FLOAT, 256x256x3x3]
%onnx::Conv_821[FLOAT, 256x256x1x1]
%onnx::Conv_824[FLOAT, 256x256x1x1]
%onnx::Conv_827[FLOAT, 256x256x1x1]
%onnx::Conv_830[FLOAT, 256x256x1x1]
%onnx::Conv_833[FLOAT, 256x256x3x3]
%onnx::Conv_836[FLOAT, 512x256x1x1]
%onnx::Conv_837[FLOAT, 512]
%onnx::Conv_839[FLOAT, 512x512x1x1]
%onnx::Conv_842[FLOAT, 512x512x1x1]
%onnx::Conv_845[FLOAT, 512x256x1x1]
%onnx::Conv_848[FLOAT, 512x512x3x3]
%onnx::Conv_851[FLOAT, 512x512x1x1]
%onnx::Conv_854[FLOAT, 512x512x1x1]
%onnx::Conv_857[FLOAT, 512x512x1x1]
%onnx::Conv_860[FLOAT, 512x512x1x1]
%onnx::Conv_863[FLOAT, 512x512x3x3]
%onnx::Conv_866[FLOAT, 512x512x1x1]
%onnx::Conv_869[FLOAT, 512x512x1x1]
%onnx::Conv_872[FLOAT, 512x512x1x1]
%onnx::Conv_875[FLOAT, 512x512x1x1]
%onnx::Conv_878[FLOAT, 512x512x3x3]
) {
%onnx::Conv_879 = Identity(%onnx::Conv_837)
%onnx::Conv_876 = Identity(%onnx::Conv_837)
%onnx::Conv_873 = Identity(%onnx::Conv_837)
%onnx::Conv_870 = Identity(%onnx::Conv_837)
%onnx::Conv_867 = Identity(%onnx::Conv_837)
%onnx::Conv_864 = Identity(%onnx::Conv_837)
%onnx::Conv_861 = Identity(%onnx::Conv_837)
%onnx::Conv_858 = Identity(%onnx::Conv_837)
%onnx::Conv_855 = Identity(%onnx::Conv_837)
%onnx::Conv_852 = Identity(%onnx::Conv_837)
%onnx::Conv_849 = Identity(%onnx::Conv_837)
%onnx::Conv_846 = Identity(%onnx::Conv_837)
%onnx::Conv_843 = Identity(%onnx::Conv_837)
%onnx::Conv_840 = Identity(%onnx::Conv_837)
%onnx::Conv_834 = Identity(%onnx::Conv_792)
%onnx::Conv_831 = Identity(%onnx::Conv_792)
%onnx::Conv_828 = Identity(%onnx::Conv_792)
%onnx::Conv_825 = Identity(%onnx::Conv_792)
%onnx::Conv_822 = Identity(%onnx::Conv_792)
%onnx::Conv_819 = Identity(%onnx::Conv_792)
%onnx::Conv_816 = Identity(%onnx::Conv_792)
%onnx::Conv_813 = Identity(%onnx::Conv_792)
%onnx::Conv_810 = Identity(%onnx::Conv_792)
%onnx::Conv_807 = Identity(%onnx::Conv_792)
%onnx::Conv_804 = Identity(%onnx::Conv_792)
%onnx::Conv_801 = Identity(%onnx::Conv_792)
%onnx::Conv_798 = Identity(%onnx::Conv_792)
%onnx::Conv_795 = Identity(%onnx::Conv_792)
%onnx::Conv_789 = Identity(%onnx::Conv_744)
%onnx::Conv_786 = Identity(%onnx::Conv_744)
%onnx::Conv_783 = Identity(%onnx::Conv_744)
%onnx::Conv_780 = Identity(%onnx::Conv_744)
%onnx::Conv_777 = Identity(%onnx::Conv_744)
%onnx::Conv_774 = Identity(%onnx::Conv_744)
%onnx::Conv_771 = Identity(%onnx::Conv_744)
%onnx::Conv_768 = Identity(%onnx::Conv_744)
%onnx::Conv_765 = Identity(%onnx::Conv_744)
%onnx::Conv_762 = Identity(%onnx::Conv_744)
%onnx::Conv_759 = Identity(%onnx::Conv_744)
%onnx::Conv_756 = Identity(%onnx::Conv_744)
%onnx::Conv_753 = Identity(%onnx::Conv_744)
%onnx::Conv_750 = Identity(%onnx::Conv_744)
%onnx::Conv_747 = Identity(%onnx::Conv_744)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_743, %onnx::Conv_744)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_746, %onnx::Conv_747)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_749, %onnx::Conv_750)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_752, %onnx::Conv_753)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_755, %onnx::Conv_756)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/Add_2_output_0, %/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_758, %onnx::Conv_759)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_761, %onnx::Conv_762)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_764, %onnx::Conv_765)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_767, %onnx::Conv_768)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_770, %onnx::Conv_771)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/Add_2_output_0, %/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_773, %onnx::Conv_774)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_776, %onnx::Conv_777)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_779, %onnx::Conv_780)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_782, %onnx::Conv_783)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_785, %onnx::Conv_786)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/Add_2_output_0, %/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_4_output_0, %onnx::Conv_788, %onnx::Conv_789)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_797, %onnx::Conv_798)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_800, %onnx::Conv_801)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/Add_2_output_0, %/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_803, %onnx::Conv_804)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_806, %onnx::Conv_807)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_809, %onnx::Conv_810)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_812, %onnx::Conv_813)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_815, %onnx::Conv_816)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/Add_2_output_0, %/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_818, %onnx::Conv_819)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_821, %onnx::Conv_822)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_824, %onnx::Conv_825)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_827, %onnx::Conv_828)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_830, %onnx::Conv_831)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/Add_2_output_0, %/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_4_output_0, %onnx::Conv_833, %onnx::Conv_834)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_836, %onnx::Conv_837)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_839, %onnx::Conv_840)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_842, %onnx::Conv_843)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_845, %onnx::Conv_846)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/Add_2_output_0, %/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_848, %onnx::Conv_849)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_851, %onnx::Conv_852)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_854, %onnx::Conv_855)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_857, %onnx::Conv_858)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/Add_2_output_0, %/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/Add_2_output_0, %/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_4_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%741 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %741
}
|
val_accuracy
| 92.277646
| 3,894,421,504
| 13,126,538
|
{'zcp_epe_nas': 101.42994704680608, 'zcp_fisher': 9.839851379394531, 'zcp_flops': 62310744064.0, 'zcp_grad_norm': 47.98204040527344, 'zcp_grasp': -1.241317749023437, 'zcp_jacov': -16.04751972714398, 'zcp_l2_norm': 1030.082275390625, 'zcp_nwot': 231.2550862902182, 'zcp_params': 13126538.0, 'zcp_plain': 0.0072635333053760005, 'zcp_snip': 402.7557678222656, 'zcp_synflow': 120.2333641880466, 'zcp_zen': 94.64735412597656, 'zcp_val_accuracy': 0.926582515239715}
| |
NASBench101_403569
|
NASBench101
|
403569
|
f3f7ff032f2c0d28a5a30265c662cd72
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_743[FLOAT, 128x3x3x3]
%onnx::Conv_744[FLOAT, 128]
%onnx::Conv_746[FLOAT, 43x128x1x1]
%onnx::Conv_747[FLOAT, 43]
%onnx::Conv_749[FLOAT, 43x43x1x1]
%onnx::Conv_752[FLOAT, 43x43x1x1]
%onnx::Conv_755[FLOAT, 42x42x1x1]
%onnx::Conv_756[FLOAT, 42]
%onnx::Conv_758[FLOAT, 43x128x1x1]
%onnx::Conv_761[FLOAT, 43x43x1x1]
%onnx::Conv_764[FLOAT, 43x43x1x1]
%onnx::Conv_767[FLOAT, 42x42x1x1]
%onnx::Conv_770[FLOAT, 43x128x1x1]
%onnx::Conv_773[FLOAT, 43x43x1x1]
%onnx::Conv_776[FLOAT, 43x43x1x1]
%onnx::Conv_779[FLOAT, 42x42x1x1]
%onnx::Conv_782[FLOAT, 86x128x1x1]
%onnx::Conv_783[FLOAT, 86]
%onnx::Conv_785[FLOAT, 86x86x1x1]
%onnx::Conv_788[FLOAT, 85x85x1x1]
%onnx::Conv_789[FLOAT, 85]
%onnx::Conv_791[FLOAT, 85x85x1x1]
%onnx::Conv_794[FLOAT, 86x256x1x1]
%onnx::Conv_797[FLOAT, 86x86x1x1]
%onnx::Conv_800[FLOAT, 85x85x1x1]
%onnx::Conv_803[FLOAT, 85x85x1x1]
%onnx::Conv_806[FLOAT, 86x256x1x1]
%onnx::Conv_809[FLOAT, 86x86x1x1]
%onnx::Conv_812[FLOAT, 85x85x1x1]
%onnx::Conv_815[FLOAT, 85x85x1x1]
%onnx::Conv_818[FLOAT, 171x256x1x1]
%onnx::Conv_819[FLOAT, 171]
%onnx::Conv_821[FLOAT, 171x171x1x1]
%onnx::Conv_824[FLOAT, 171x171x1x1]
%onnx::Conv_827[FLOAT, 170x170x1x1]
%onnx::Conv_828[FLOAT, 170]
%onnx::Conv_830[FLOAT, 171x512x1x1]
%onnx::Conv_833[FLOAT, 171x171x1x1]
%onnx::Conv_836[FLOAT, 171x171x1x1]
%onnx::Conv_839[FLOAT, 170x170x1x1]
%onnx::Conv_842[FLOAT, 171x512x1x1]
%onnx::Conv_845[FLOAT, 171x171x1x1]
%onnx::Conv_848[FLOAT, 171x171x1x1]
%onnx::Conv_851[FLOAT, 170x170x1x1]
) {
%onnx::Conv_852 = Identity(%onnx::Conv_828)
%onnx::Conv_849 = Identity(%onnx::Conv_819)
%onnx::Conv_846 = Identity(%onnx::Conv_819)
%onnx::Conv_843 = Identity(%onnx::Conv_819)
%onnx::Conv_840 = Identity(%onnx::Conv_828)
%onnx::Conv_837 = Identity(%onnx::Conv_819)
%onnx::Conv_834 = Identity(%onnx::Conv_819)
%onnx::Conv_831 = Identity(%onnx::Conv_819)
%onnx::Conv_825 = Identity(%onnx::Conv_819)
%onnx::Conv_822 = Identity(%onnx::Conv_819)
%onnx::Conv_816 = Identity(%onnx::Conv_789)
%onnx::Conv_813 = Identity(%onnx::Conv_789)
%onnx::Conv_810 = Identity(%onnx::Conv_783)
%onnx::Conv_807 = Identity(%onnx::Conv_783)
%onnx::Conv_804 = Identity(%onnx::Conv_789)
%onnx::Conv_801 = Identity(%onnx::Conv_789)
%onnx::Conv_798 = Identity(%onnx::Conv_783)
%onnx::Conv_795 = Identity(%onnx::Conv_783)
%onnx::Conv_792 = Identity(%onnx::Conv_789)
%onnx::Conv_786 = Identity(%onnx::Conv_783)
%onnx::Conv_780 = Identity(%onnx::Conv_756)
%onnx::Conv_777 = Identity(%onnx::Conv_747)
%onnx::Conv_774 = Identity(%onnx::Conv_747)
%onnx::Conv_771 = Identity(%onnx::Conv_747)
%onnx::Conv_768 = Identity(%onnx::Conv_756)
%onnx::Conv_765 = Identity(%onnx::Conv_747)
%onnx::Conv_762 = Identity(%onnx::Conv_747)
%onnx::Conv_759 = Identity(%onnx::Conv_747)
%onnx::Conv_753 = Identity(%onnx::Conv_747)
%onnx::Conv_750 = Identity(%onnx::Conv_747)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_743, %onnx::Conv_744)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_746, %onnx::Conv_747)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_749, %onnx::Conv_750)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_752, %onnx::Conv_753)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.1/Slice_output_0 = Slice(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0, %/layers.1/Constant_4_output_0, %/layers.1/Constant_2_output_0, %/layers.1/Constant_5_output_0)
%/layers.1/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.1/Slice_1_output_0 = Slice(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_7_output_0, %/layers.1/Constant_8_output_0, %/layers.1/Constant_6_output_0, %/layers.1/Constant_9_output_0)
%/layers.1/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/Slice_output_0, %/layers.1/Constant_10_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/Add_2_output_0, %/layers.1/Slice_1_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_755, %onnx::Conv_756)
%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_758, %onnx::Conv_759)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_761, %onnx::Conv_762)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_764, %onnx::Conv_765)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.2/Slice_output_0 = Slice(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0, %/layers.2/Constant_4_output_0, %/layers.2/Constant_2_output_0, %/layers.2/Constant_5_output_0)
%/layers.2/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.2/Slice_1_output_0 = Slice(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_7_output_0, %/layers.2/Constant_8_output_0, %/layers.2/Constant_6_output_0, %/layers.2/Constant_9_output_0)
%/layers.2/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/Slice_output_0, %/layers.2/Constant_10_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/Add_2_output_0, %/layers.2/Slice_1_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_767, %onnx::Conv_768)
%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_770, %onnx::Conv_771)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_773, %onnx::Conv_774)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_776, %onnx::Conv_777)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.3/Slice_output_0 = Slice(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0, %/layers.3/Constant_4_output_0, %/layers.3/Constant_2_output_0, %/layers.3/Constant_5_output_0)
%/layers.3/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.3/Slice_1_output_0 = Slice(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_7_output_0, %/layers.3/Constant_8_output_0, %/layers.3/Constant_6_output_0, %/layers.3/Constant_9_output_0)
%/layers.3/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/Slice_output_0, %/layers.3/Constant_10_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/Add_2_output_0, %/layers.3/Slice_1_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_779, %onnx::Conv_780)
%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_782, %onnx::Conv_783)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_785, %onnx::Conv_786)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.5/Slice_output_0 = Slice(%/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %/layers.5/Constant_3_output_0, %/layers.5/Constant_4_output_0, %/layers.5/Constant_2_output_0, %/layers.5/Constant_5_output_0)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Slice_output_0, %onnx::Conv_788, %onnx::Conv_789)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.5/Slice_1_output_0 = Slice(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_7_output_0, %/layers.5/Constant_8_output_0, %/layers.5/Constant_6_output_0, %/layers.5/Constant_9_output_0)
%/layers.5/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/Slice_1_output_0, %/layers.5/Constant_10_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/Add_2_output_0, %/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_797, %onnx::Conv_798)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.6/Slice_output_0 = Slice(%/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %/layers.6/Constant_3_output_0, %/layers.6/Constant_4_output_0, %/layers.6/Constant_2_output_0, %/layers.6/Constant_5_output_0)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Slice_output_0, %onnx::Conv_800, %onnx::Conv_801)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.6/Slice_1_output_0 = Slice(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_7_output_0, %/layers.6/Constant_8_output_0, %/layers.6/Constant_6_output_0, %/layers.6/Constant_9_output_0)
%/layers.6/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/Slice_1_output_0, %/layers.6/Constant_10_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/Add_2_output_0, %/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_803, %onnx::Conv_804)
%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_806, %onnx::Conv_807)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_809, %onnx::Conv_810)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.7/Slice_output_0 = Slice(%/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %/layers.7/Constant_3_output_0, %/layers.7/Constant_4_output_0, %/layers.7/Constant_2_output_0, %/layers.7/Constant_5_output_0)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Slice_output_0, %onnx::Conv_812, %onnx::Conv_813)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.7/Slice_1_output_0 = Slice(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_7_output_0, %/layers.7/Constant_8_output_0, %/layers.7/Constant_6_output_0, %/layers.7/Constant_9_output_0)
%/layers.7/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/Slice_1_output_0, %/layers.7/Constant_10_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/Add_2_output_0, %/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_815, %onnx::Conv_816)
%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_818, %onnx::Conv_819)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_821, %onnx::Conv_822)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_824, %onnx::Conv_825)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.9/Slice_output_0 = Slice(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0, %/layers.9/Constant_4_output_0, %/layers.9/Constant_2_output_0, %/layers.9/Constant_5_output_0)
%/layers.9/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.9/Slice_1_output_0 = Slice(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_7_output_0, %/layers.9/Constant_8_output_0, %/layers.9/Constant_6_output_0, %/layers.9/Constant_9_output_0)
%/layers.9/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/Slice_output_0, %/layers.9/Constant_10_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/Add_2_output_0, %/layers.9/Slice_1_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_827, %onnx::Conv_828)
%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_830, %onnx::Conv_831)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_833, %onnx::Conv_834)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_836, %onnx::Conv_837)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.10/Slice_output_0 = Slice(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0, %/layers.10/Constant_4_output_0, %/layers.10/Constant_2_output_0, %/layers.10/Constant_5_output_0)
%/layers.10/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.10/Slice_1_output_0 = Slice(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_7_output_0, %/layers.10/Constant_8_output_0, %/layers.10/Constant_6_output_0, %/layers.10/Constant_9_output_0)
%/layers.10/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/Slice_output_0, %/layers.10/Constant_10_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/Add_2_output_0, %/layers.10/Slice_1_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_839, %onnx::Conv_840)
%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_842, %onnx::Conv_843)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_845, %onnx::Conv_846)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %onnx::Conv_848, %onnx::Conv_849)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.11/Slice_output_0 = Slice(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0, %/layers.11/Constant_4_output_0, %/layers.11/Constant_2_output_0, %/layers.11/Constant_5_output_0)
%/layers.11/Constant_6_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_7_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_8_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_9_output_0 = Constant[value = <Tensor>]()
%/layers.11/Slice_1_output_0 = Slice(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_7_output_0, %/layers.11/Constant_8_output_0, %/layers.11/Constant_6_output_0, %/layers.11/Constant_9_output_0)
%/layers.11/Constant_10_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/Slice_output_0, %/layers.11/Constant_10_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/Add_2_output_0, %/layers.11/Slice_1_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_851, %onnx::Conv_852)
%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.5/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%741 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %741
}
|
val_accuracy
| 85.957533
| 206,171,904
| 650,520
|
{'zcp_epe_nas': 102.50468012661595, 'zcp_fisher': 48.40120315551758, 'zcp_flops': 3298750464.0, 'zcp_grad_norm': 126.39688110351562, 'zcp_grasp': 8.55908203125, 'zcp_jacov': -16.046377664360218, 'zcp_l2_norm': 564.0869140625, 'zcp_nwot': 213.03798197079422, 'zcp_params': 650520.0, 'zcp_plain': -0.018146451562643003, 'zcp_snip': 528.3807983398438, 'zcp_synflow': 91.63341519120034, 'zcp_zen': 51.6834602355957, 'zcp_val_accuracy': 0.9237780570983881}
| |
NASBench101_44162
|
NASBench101
|
44162
|
1acc218dee85fb3b584e026796c1e603
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_887[FLOAT, 128x3x3x3]
%onnx::Conv_888[FLOAT, 128]
%onnx::Conv_890[FLOAT, 64x128x1x1]
%onnx::Conv_891[FLOAT, 64]
%onnx::Conv_893[FLOAT, 64x64x3x3]
%onnx::Conv_896[FLOAT, 64x64x3x3]
%onnx::Conv_899[FLOAT, 64x64x1x1]
%onnx::Conv_902[FLOAT, 64x64x3x3]
%onnx::Conv_905[FLOAT, 128x128x1x1]
%onnx::Conv_908[FLOAT, 64x128x1x1]
%onnx::Conv_911[FLOAT, 64x64x3x3]
%onnx::Conv_914[FLOAT, 64x64x3x3]
%onnx::Conv_917[FLOAT, 64x64x1x1]
%onnx::Conv_920[FLOAT, 64x64x3x3]
%onnx::Conv_923[FLOAT, 128x128x1x1]
%onnx::Conv_926[FLOAT, 64x128x1x1]
%onnx::Conv_929[FLOAT, 64x64x3x3]
%onnx::Conv_932[FLOAT, 64x64x3x3]
%onnx::Conv_935[FLOAT, 64x64x1x1]
%onnx::Conv_938[FLOAT, 64x64x3x3]
%onnx::Conv_941[FLOAT, 128x128x1x1]
%onnx::Conv_944[FLOAT, 128x128x1x1]
%onnx::Conv_947[FLOAT, 128x128x3x3]
%onnx::Conv_950[FLOAT, 128x128x3x3]
%onnx::Conv_953[FLOAT, 128x128x1x1]
%onnx::Conv_956[FLOAT, 128x128x3x3]
%onnx::Conv_959[FLOAT, 256x128x1x1]
%onnx::Conv_960[FLOAT, 256]
%onnx::Conv_962[FLOAT, 128x256x1x1]
%onnx::Conv_965[FLOAT, 128x128x3x3]
%onnx::Conv_968[FLOAT, 128x128x3x3]
%onnx::Conv_971[FLOAT, 128x128x1x1]
%onnx::Conv_974[FLOAT, 128x128x3x3]
%onnx::Conv_977[FLOAT, 256x256x1x1]
%onnx::Conv_980[FLOAT, 128x256x1x1]
%onnx::Conv_983[FLOAT, 128x128x3x3]
%onnx::Conv_986[FLOAT, 128x128x3x3]
%onnx::Conv_989[FLOAT, 128x128x1x1]
%onnx::Conv_992[FLOAT, 128x128x3x3]
%onnx::Conv_995[FLOAT, 256x256x1x1]
%onnx::Conv_998[FLOAT, 256x256x1x1]
%onnx::Conv_1001[FLOAT, 256x256x3x3]
%onnx::Conv_1004[FLOAT, 256x256x3x3]
%onnx::Conv_1007[FLOAT, 256x256x1x1]
%onnx::Conv_1010[FLOAT, 256x256x3x3]
%onnx::Conv_1013[FLOAT, 512x256x1x1]
%onnx::Conv_1014[FLOAT, 512]
%onnx::Conv_1016[FLOAT, 256x512x1x1]
%onnx::Conv_1019[FLOAT, 256x256x3x3]
%onnx::Conv_1022[FLOAT, 256x256x3x3]
%onnx::Conv_1025[FLOAT, 256x256x1x1]
%onnx::Conv_1028[FLOAT, 256x256x3x3]
%onnx::Conv_1031[FLOAT, 512x512x1x1]
%onnx::Conv_1034[FLOAT, 256x512x1x1]
%onnx::Conv_1037[FLOAT, 256x256x3x3]
%onnx::Conv_1040[FLOAT, 256x256x3x3]
%onnx::Conv_1043[FLOAT, 256x256x1x1]
%onnx::Conv_1046[FLOAT, 256x256x3x3]
%onnx::Conv_1049[FLOAT, 512x512x1x1]
) {
%onnx::Conv_1050 = Identity(%onnx::Conv_1014)
%onnx::Conv_1047 = Identity(%onnx::Conv_960)
%onnx::Conv_1044 = Identity(%onnx::Conv_960)
%onnx::Conv_1041 = Identity(%onnx::Conv_960)
%onnx::Conv_1038 = Identity(%onnx::Conv_960)
%onnx::Conv_1035 = Identity(%onnx::Conv_960)
%onnx::Conv_1032 = Identity(%onnx::Conv_1014)
%onnx::Conv_1029 = Identity(%onnx::Conv_960)
%onnx::Conv_1026 = Identity(%onnx::Conv_960)
%onnx::Conv_1023 = Identity(%onnx::Conv_960)
%onnx::Conv_1020 = Identity(%onnx::Conv_960)
%onnx::Conv_1017 = Identity(%onnx::Conv_960)
%onnx::Conv_1011 = Identity(%onnx::Conv_960)
%onnx::Conv_1008 = Identity(%onnx::Conv_960)
%onnx::Conv_1005 = Identity(%onnx::Conv_960)
%onnx::Conv_1002 = Identity(%onnx::Conv_960)
%onnx::Conv_999 = Identity(%onnx::Conv_960)
%onnx::Conv_996 = Identity(%onnx::Conv_960)
%onnx::Conv_993 = Identity(%onnx::Conv_888)
%onnx::Conv_990 = Identity(%onnx::Conv_888)
%onnx::Conv_987 = Identity(%onnx::Conv_888)
%onnx::Conv_984 = Identity(%onnx::Conv_888)
%onnx::Conv_981 = Identity(%onnx::Conv_888)
%onnx::Conv_978 = Identity(%onnx::Conv_960)
%onnx::Conv_975 = Identity(%onnx::Conv_888)
%onnx::Conv_972 = Identity(%onnx::Conv_888)
%onnx::Conv_969 = Identity(%onnx::Conv_888)
%onnx::Conv_966 = Identity(%onnx::Conv_888)
%onnx::Conv_963 = Identity(%onnx::Conv_888)
%onnx::Conv_957 = Identity(%onnx::Conv_888)
%onnx::Conv_954 = Identity(%onnx::Conv_888)
%onnx::Conv_951 = Identity(%onnx::Conv_888)
%onnx::Conv_948 = Identity(%onnx::Conv_888)
%onnx::Conv_945 = Identity(%onnx::Conv_888)
%onnx::Conv_942 = Identity(%onnx::Conv_888)
%onnx::Conv_939 = Identity(%onnx::Conv_891)
%onnx::Conv_936 = Identity(%onnx::Conv_891)
%onnx::Conv_933 = Identity(%onnx::Conv_891)
%onnx::Conv_930 = Identity(%onnx::Conv_891)
%onnx::Conv_927 = Identity(%onnx::Conv_891)
%onnx::Conv_924 = Identity(%onnx::Conv_888)
%onnx::Conv_921 = Identity(%onnx::Conv_891)
%onnx::Conv_918 = Identity(%onnx::Conv_891)
%onnx::Conv_915 = Identity(%onnx::Conv_891)
%onnx::Conv_912 = Identity(%onnx::Conv_891)
%onnx::Conv_909 = Identity(%onnx::Conv_891)
%onnx::Conv_906 = Identity(%onnx::Conv_888)
%onnx::Conv_903 = Identity(%onnx::Conv_891)
%onnx::Conv_900 = Identity(%onnx::Conv_891)
%onnx::Conv_897 = Identity(%onnx::Conv_891)
%onnx::Conv_894 = Identity(%onnx::Conv_891)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_887, %onnx::Conv_888)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/Add_3_output_0, %/layers.1/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_4_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_5_output_0 = Add(%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_4_output_0)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_5_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.3/maxpool/MaxPool_output_0, %/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_6_output_0 = Add(%/layers.1/Concat_output_0, %/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_6_output_0, %onnx::Conv_908, %onnx::Conv_909)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_911, %onnx::Conv_912)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_914, %onnx::Conv_915)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/Add_3_output_0, %/layers.2/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_4_output_0, %onnx::Conv_917, %onnx::Conv_918)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_5_output_0 = Add(%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_4_output_0)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_5_output_0, %onnx::Conv_920, %onnx::Conv_921)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.3/maxpool/MaxPool_output_0, %/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_6_output_0, %onnx::Conv_923, %onnx::Conv_924)
%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_6_output_0 = Add(%/layers.2/Concat_output_0, %/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_6_output_0, %onnx::Conv_926, %onnx::Conv_927)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_929, %onnx::Conv_930)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_932, %onnx::Conv_933)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/Add_3_output_0, %/layers.3/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_4_output_0, %onnx::Conv_935, %onnx::Conv_936)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_5_output_0 = Add(%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_4_output_0)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_5_output_0, %onnx::Conv_938, %onnx::Conv_939)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.3/maxpool/MaxPool_output_0, %/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_6_output_0, %onnx::Conv_941, %onnx::Conv_942)
%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_6_output_0 = Add(%/layers.3/Concat_output_0, %/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Add_6_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_944, %onnx::Conv_945)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_947, %onnx::Conv_948)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_950, %onnx::Conv_951)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/Add_3_output_0, %/layers.5/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_4_output_0, %onnx::Conv_953, %onnx::Conv_954)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_5_output_0 = Add(%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_4_output_0)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_5_output_0, %onnx::Conv_956, %onnx::Conv_957)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.3/maxpool/MaxPool_output_0, %/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_959, %onnx::Conv_960)
%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_6_output_0 = Add(%/layers.5/Concat_output_0, %/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_6_output_0, %onnx::Conv_962, %onnx::Conv_963)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_965, %onnx::Conv_966)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_968, %onnx::Conv_969)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/Add_3_output_0, %/layers.6/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_4_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_5_output_0 = Add(%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_4_output_0)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_5_output_0, %onnx::Conv_974, %onnx::Conv_975)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.3/maxpool/MaxPool_output_0, %/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_6_output_0, %onnx::Conv_977, %onnx::Conv_978)
%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_6_output_0 = Add(%/layers.6/Concat_output_0, %/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_6_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/Add_3_output_0, %/layers.7/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_4_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_5_output_0 = Add(%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_4_output_0)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_5_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.3/maxpool/MaxPool_output_0, %/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_6_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_6_output_0 = Add(%/layers.7/Concat_output_0, %/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Add_6_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/Add_3_output_0, %/layers.9/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_4_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_5_output_0 = Add(%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_4_output_0)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_5_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.3/maxpool/MaxPool_output_0, %/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_6_output_0 = Add(%/layers.9/Concat_output_0, %/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_6_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/Add_3_output_0, %/layers.10/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_4_output_0, %onnx::Conv_1025, %onnx::Conv_1026)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_5_output_0 = Add(%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_4_output_0)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_5_output_0, %onnx::Conv_1028, %onnx::Conv_1029)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.3/maxpool/MaxPool_output_0, %/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_6_output_0, %onnx::Conv_1031, %onnx::Conv_1032)
%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_6_output_0 = Add(%/layers.10/Concat_output_0, %/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_6_output_0, %onnx::Conv_1034, %onnx::Conv_1035)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_1037, %onnx::Conv_1038)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_1040, %onnx::Conv_1041)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/Add_3_output_0, %/layers.11/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_4_output_0, %onnx::Conv_1043, %onnx::Conv_1044)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_5_output_0 = Add(%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_4_output_0)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_5_output_0, %onnx::Conv_1046, %onnx::Conv_1047)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.3/maxpool/MaxPool_output_0, %/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_6_output_0, %onnx::Conv_1049, %onnx::Conv_1050)
%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_6_output_0 = Add(%/layers.11/Concat_output_0, %/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Add_6_output_0)
%885 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %885
}
|
val_accuracy
| 92.538059
| 2,543,986,688
| 8,555,530
|
{'zcp_epe_nas': 141.13786474449415, 'zcp_fisher': 148.565185546875, 'zcp_flops': 40703787008.0, 'zcp_grad_norm': 237.42330932617188, 'zcp_grasp': 2.23974609375, 'zcp_jacov': -16.053043804385833, 'zcp_l2_norm': 994.1105346679688, 'zcp_nwot': 226.72685939677373, 'zcp_params': 8555530.0, 'zcp_plain': 0.05536214262247, 'zcp_snip': 1470.6585693359375, 'zcp_synflow': 155.97044981981372, 'zcp_zen': 112.21358489990234, 'zcp_val_accuracy': 0.913661837577819}
| |
NASBench101_381979
|
NASBench101
|
381979
|
e6f1dbc0987071ab4969bfe45316dc16
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_878[FLOAT, 128x3x3x3]
%onnx::Conv_879[FLOAT, 128]
%onnx::Conv_881[FLOAT, 128x128x1x1]
%onnx::Conv_884[FLOAT, 128x128x1x1]
%onnx::Conv_887[FLOAT, 128x128x1x1]
%onnx::Conv_890[FLOAT, 128x128x1x1]
%onnx::Conv_893[FLOAT, 128x128x3x3]
%onnx::Conv_896[FLOAT, 128x128x1x1]
%onnx::Conv_899[FLOAT, 128x128x1x1]
%onnx::Conv_902[FLOAT, 128x128x1x1]
%onnx::Conv_905[FLOAT, 128x128x1x1]
%onnx::Conv_908[FLOAT, 128x128x1x1]
%onnx::Conv_911[FLOAT, 128x128x3x3]
%onnx::Conv_914[FLOAT, 128x128x1x1]
%onnx::Conv_917[FLOAT, 128x128x1x1]
%onnx::Conv_920[FLOAT, 128x128x1x1]
%onnx::Conv_923[FLOAT, 128x128x1x1]
%onnx::Conv_926[FLOAT, 128x128x1x1]
%onnx::Conv_929[FLOAT, 128x128x3x3]
%onnx::Conv_932[FLOAT, 128x128x1x1]
%onnx::Conv_935[FLOAT, 256x128x1x1]
%onnx::Conv_936[FLOAT, 256]
%onnx::Conv_938[FLOAT, 256x256x1x1]
%onnx::Conv_941[FLOAT, 256x128x1x1]
%onnx::Conv_944[FLOAT, 256x256x1x1]
%onnx::Conv_947[FLOAT, 256x256x3x3]
%onnx::Conv_950[FLOAT, 256x128x1x1]
%onnx::Conv_953[FLOAT, 256x256x1x1]
%onnx::Conv_956[FLOAT, 256x256x1x1]
%onnx::Conv_959[FLOAT, 256x256x1x1]
%onnx::Conv_962[FLOAT, 256x256x1x1]
%onnx::Conv_965[FLOAT, 256x256x3x3]
%onnx::Conv_968[FLOAT, 256x256x1x1]
%onnx::Conv_971[FLOAT, 256x256x1x1]
%onnx::Conv_974[FLOAT, 256x256x1x1]
%onnx::Conv_977[FLOAT, 256x256x1x1]
%onnx::Conv_980[FLOAT, 256x256x1x1]
%onnx::Conv_983[FLOAT, 256x256x3x3]
%onnx::Conv_986[FLOAT, 256x256x1x1]
%onnx::Conv_989[FLOAT, 512x256x1x1]
%onnx::Conv_990[FLOAT, 512]
%onnx::Conv_992[FLOAT, 512x512x1x1]
%onnx::Conv_995[FLOAT, 512x256x1x1]
%onnx::Conv_998[FLOAT, 512x512x1x1]
%onnx::Conv_1001[FLOAT, 512x512x3x3]
%onnx::Conv_1004[FLOAT, 512x256x1x1]
%onnx::Conv_1007[FLOAT, 512x512x1x1]
%onnx::Conv_1010[FLOAT, 512x512x1x1]
%onnx::Conv_1013[FLOAT, 512x512x1x1]
%onnx::Conv_1016[FLOAT, 512x512x1x1]
%onnx::Conv_1019[FLOAT, 512x512x3x3]
%onnx::Conv_1022[FLOAT, 512x512x1x1]
%onnx::Conv_1025[FLOAT, 512x512x1x1]
%onnx::Conv_1028[FLOAT, 512x512x1x1]
%onnx::Conv_1031[FLOAT, 512x512x1x1]
%onnx::Conv_1034[FLOAT, 512x512x1x1]
%onnx::Conv_1037[FLOAT, 512x512x3x3]
%onnx::Conv_1040[FLOAT, 512x512x1x1]
) {
%onnx::Conv_1041 = Identity(%onnx::Conv_990)
%onnx::Conv_1038 = Identity(%onnx::Conv_990)
%onnx::Conv_1035 = Identity(%onnx::Conv_990)
%onnx::Conv_1032 = Identity(%onnx::Conv_990)
%onnx::Conv_1029 = Identity(%onnx::Conv_990)
%onnx::Conv_1026 = Identity(%onnx::Conv_990)
%onnx::Conv_1023 = Identity(%onnx::Conv_990)
%onnx::Conv_1020 = Identity(%onnx::Conv_990)
%onnx::Conv_1017 = Identity(%onnx::Conv_990)
%onnx::Conv_1014 = Identity(%onnx::Conv_990)
%onnx::Conv_1011 = Identity(%onnx::Conv_990)
%onnx::Conv_1008 = Identity(%onnx::Conv_990)
%onnx::Conv_1005 = Identity(%onnx::Conv_990)
%onnx::Conv_1002 = Identity(%onnx::Conv_990)
%onnx::Conv_999 = Identity(%onnx::Conv_990)
%onnx::Conv_996 = Identity(%onnx::Conv_990)
%onnx::Conv_993 = Identity(%onnx::Conv_990)
%onnx::Conv_987 = Identity(%onnx::Conv_936)
%onnx::Conv_984 = Identity(%onnx::Conv_936)
%onnx::Conv_981 = Identity(%onnx::Conv_936)
%onnx::Conv_978 = Identity(%onnx::Conv_936)
%onnx::Conv_975 = Identity(%onnx::Conv_936)
%onnx::Conv_972 = Identity(%onnx::Conv_936)
%onnx::Conv_969 = Identity(%onnx::Conv_936)
%onnx::Conv_966 = Identity(%onnx::Conv_936)
%onnx::Conv_963 = Identity(%onnx::Conv_936)
%onnx::Conv_960 = Identity(%onnx::Conv_936)
%onnx::Conv_957 = Identity(%onnx::Conv_936)
%onnx::Conv_954 = Identity(%onnx::Conv_936)
%onnx::Conv_951 = Identity(%onnx::Conv_936)
%onnx::Conv_948 = Identity(%onnx::Conv_936)
%onnx::Conv_945 = Identity(%onnx::Conv_936)
%onnx::Conv_942 = Identity(%onnx::Conv_936)
%onnx::Conv_939 = Identity(%onnx::Conv_936)
%onnx::Conv_933 = Identity(%onnx::Conv_879)
%onnx::Conv_930 = Identity(%onnx::Conv_879)
%onnx::Conv_927 = Identity(%onnx::Conv_879)
%onnx::Conv_924 = Identity(%onnx::Conv_879)
%onnx::Conv_921 = Identity(%onnx::Conv_879)
%onnx::Conv_918 = Identity(%onnx::Conv_879)
%onnx::Conv_915 = Identity(%onnx::Conv_879)
%onnx::Conv_912 = Identity(%onnx::Conv_879)
%onnx::Conv_909 = Identity(%onnx::Conv_879)
%onnx::Conv_906 = Identity(%onnx::Conv_879)
%onnx::Conv_903 = Identity(%onnx::Conv_879)
%onnx::Conv_900 = Identity(%onnx::Conv_879)
%onnx::Conv_897 = Identity(%onnx::Conv_879)
%onnx::Conv_894 = Identity(%onnx::Conv_879)
%onnx::Conv_891 = Identity(%onnx::Conv_879)
%onnx::Conv_888 = Identity(%onnx::Conv_879)
%onnx::Conv_885 = Identity(%onnx::Conv_879)
%onnx::Conv_882 = Identity(%onnx::Conv_879)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_878, %onnx::Conv_879)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/Add_2_output_0, %/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/Add_5_output_0 = Add(%/layers.1/Add_4_output_0, %/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_5_output_0)
%/layers.1/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_6_output_0 = Add(%/layers.1/vertex_op.5/maxpool/MaxPool_output_0, %/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_6_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_6_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_908, %onnx::Conv_909)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/Add_2_output_0, %/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_911, %onnx::Conv_912)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/Add_5_output_0 = Add(%/layers.2/Add_4_output_0, %/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_5_output_0)
%/layers.2/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_6_output_0, %onnx::Conv_914, %onnx::Conv_915)
%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_6_output_0 = Add(%/layers.2/vertex_op.5/maxpool/MaxPool_output_0, %/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_6_output_0, %onnx::Conv_917, %onnx::Conv_918)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_920, %onnx::Conv_921)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_6_output_0, %onnx::Conv_923, %onnx::Conv_924)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_926, %onnx::Conv_927)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/Add_2_output_0, %/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_929, %onnx::Conv_930)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/Add_5_output_0 = Add(%/layers.3/Add_4_output_0, %/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_5_output_0)
%/layers.3/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_6_output_0, %onnx::Conv_932, %onnx::Conv_933)
%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_6_output_0 = Add(%/layers.3/vertex_op.5/maxpool/MaxPool_output_0, %/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Add_6_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_935, %onnx::Conv_936)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_938, %onnx::Conv_939)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_941, %onnx::Conv_942)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_944, %onnx::Conv_945)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/Add_2_output_0, %/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_947, %onnx::Conv_948)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/Add_5_output_0 = Add(%/layers.5/Add_4_output_0, %/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_5_output_0)
%/layers.5/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_950, %onnx::Conv_951)
%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_6_output_0 = Add(%/layers.5/vertex_op.5/maxpool/MaxPool_output_0, %/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_6_output_0, %onnx::Conv_953, %onnx::Conv_954)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_956, %onnx::Conv_957)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_6_output_0, %onnx::Conv_959, %onnx::Conv_960)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_962, %onnx::Conv_963)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/Add_2_output_0, %/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_965, %onnx::Conv_966)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/Add_5_output_0 = Add(%/layers.6/Add_4_output_0, %/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_5_output_0)
%/layers.6/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_6_output_0, %onnx::Conv_968, %onnx::Conv_969)
%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_6_output_0 = Add(%/layers.6/vertex_op.5/maxpool/MaxPool_output_0, %/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_6_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_974, %onnx::Conv_975)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_6_output_0, %onnx::Conv_977, %onnx::Conv_978)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/Add_2_output_0, %/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/Add_5_output_0 = Add(%/layers.7/Add_4_output_0, %/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_5_output_0)
%/layers.7/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_6_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_6_output_0 = Add(%/layers.7/vertex_op.5/maxpool/MaxPool_output_0, %/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Add_6_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/Add_2_output_0, %/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/Add_5_output_0 = Add(%/layers.9/Add_4_output_0, %/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_5_output_0)
%/layers.9/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_6_output_0 = Add(%/layers.9/vertex_op.5/maxpool/MaxPool_output_0, %/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_6_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_6_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/Add_2_output_0, %/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/Add_5_output_0 = Add(%/layers.10/Add_4_output_0, %/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_5_output_0)
%/layers.10/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_6_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_6_output_0 = Add(%/layers.10/vertex_op.5/maxpool/MaxPool_output_0, %/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_6_output_0, %onnx::Conv_1025, %onnx::Conv_1026)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_1028, %onnx::Conv_1029)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_6_output_0, %onnx::Conv_1031, %onnx::Conv_1032)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_1034, %onnx::Conv_1035)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/Add_2_output_0, %/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_1037, %onnx::Conv_1038)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/Add_5_output_0 = Add(%/layers.11/Add_4_output_0, %/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_5_output_0)
%/layers.11/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_6_output_0, %onnx::Conv_1040, %onnx::Conv_1041)
%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_6_output_0 = Add(%/layers.11/vertex_op.5/maxpool/MaxPool_output_0, %/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Add_6_output_0)
%876 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %876
}
|
val_accuracy
| 91.656649
| 4,168,361,984
| 14,000,266
|
{'zcp_epe_nas': 111.52524274780156, 'zcp_fisher': 230.58438110351562, 'zcp_flops': 66693791744.0, 'zcp_grad_norm': 293.84539794921875, 'zcp_grasp': -137.662109375, 'zcp_jacov': -16.049090196413516, 'zcp_l2_norm': 1225.93359375, 'zcp_nwot': 234.79125421536853, 'zcp_params': 14000266.0, 'zcp_plain': 0.089739255607128, 'zcp_snip': 2299.9326171875, 'zcp_synflow': 99.89202514438875, 'zcp_zen': 112.00395202636719, 'zcp_val_accuracy': 0.927383840084075}
| |
NASBench101_388587
|
NASBench101
|
388587
|
eadc9271099c7e838bcdb564d0282f54
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_968[FLOAT, 128x3x3x3]
%onnx::Conv_969[FLOAT, 128]
%onnx::Conv_971[FLOAT, 128x128x1x1]
%onnx::Conv_974[FLOAT, 128x128x1x1]
%onnx::Conv_977[FLOAT, 128x128x1x1]
%onnx::Conv_980[FLOAT, 128x128x3x3]
%onnx::Conv_983[FLOAT, 128x128x1x1]
%onnx::Conv_986[FLOAT, 128x128x1x1]
%onnx::Conv_989[FLOAT, 128x128x3x3]
%onnx::Conv_992[FLOAT, 128x128x1x1]
%onnx::Conv_995[FLOAT, 128x128x1x1]
%onnx::Conv_998[FLOAT, 128x128x1x1]
%onnx::Conv_1001[FLOAT, 128x128x3x3]
%onnx::Conv_1004[FLOAT, 128x128x1x1]
%onnx::Conv_1007[FLOAT, 128x128x1x1]
%onnx::Conv_1010[FLOAT, 128x128x3x3]
%onnx::Conv_1013[FLOAT, 128x128x1x1]
%onnx::Conv_1016[FLOAT, 128x128x1x1]
%onnx::Conv_1019[FLOAT, 128x128x1x1]
%onnx::Conv_1022[FLOAT, 128x128x3x3]
%onnx::Conv_1025[FLOAT, 128x128x1x1]
%onnx::Conv_1028[FLOAT, 128x128x1x1]
%onnx::Conv_1031[FLOAT, 128x128x3x3]
%onnx::Conv_1034[FLOAT, 256x128x1x1]
%onnx::Conv_1035[FLOAT, 256]
%onnx::Conv_1037[FLOAT, 256x128x1x1]
%onnx::Conv_1040[FLOAT, 256x256x1x1]
%onnx::Conv_1043[FLOAT, 256x256x3x3]
%onnx::Conv_1046[FLOAT, 256x256x1x1]
%onnx::Conv_1049[FLOAT, 256x128x1x1]
%onnx::Conv_1052[FLOAT, 256x256x3x3]
%onnx::Conv_1055[FLOAT, 256x256x1x1]
%onnx::Conv_1058[FLOAT, 256x256x1x1]
%onnx::Conv_1061[FLOAT, 256x256x1x1]
%onnx::Conv_1064[FLOAT, 256x256x3x3]
%onnx::Conv_1067[FLOAT, 256x256x1x1]
%onnx::Conv_1070[FLOAT, 256x256x1x1]
%onnx::Conv_1073[FLOAT, 256x256x3x3]
%onnx::Conv_1076[FLOAT, 256x256x1x1]
%onnx::Conv_1079[FLOAT, 256x256x1x1]
%onnx::Conv_1082[FLOAT, 256x256x1x1]
%onnx::Conv_1085[FLOAT, 256x256x3x3]
%onnx::Conv_1088[FLOAT, 256x256x1x1]
%onnx::Conv_1091[FLOAT, 256x256x1x1]
%onnx::Conv_1094[FLOAT, 256x256x3x3]
%onnx::Conv_1097[FLOAT, 512x256x1x1]
%onnx::Conv_1098[FLOAT, 512]
%onnx::Conv_1100[FLOAT, 512x256x1x1]
%onnx::Conv_1103[FLOAT, 512x512x1x1]
%onnx::Conv_1106[FLOAT, 512x512x3x3]
%onnx::Conv_1109[FLOAT, 512x512x1x1]
%onnx::Conv_1112[FLOAT, 512x256x1x1]
%onnx::Conv_1115[FLOAT, 512x512x3x3]
%onnx::Conv_1118[FLOAT, 512x512x1x1]
%onnx::Conv_1121[FLOAT, 512x512x1x1]
%onnx::Conv_1124[FLOAT, 512x512x1x1]
%onnx::Conv_1127[FLOAT, 512x512x3x3]
%onnx::Conv_1130[FLOAT, 512x512x1x1]
%onnx::Conv_1133[FLOAT, 512x512x1x1]
%onnx::Conv_1136[FLOAT, 512x512x3x3]
%onnx::Conv_1139[FLOAT, 512x512x1x1]
%onnx::Conv_1142[FLOAT, 512x512x1x1]
%onnx::Conv_1145[FLOAT, 512x512x1x1]
%onnx::Conv_1148[FLOAT, 512x512x3x3]
%onnx::Conv_1151[FLOAT, 512x512x1x1]
%onnx::Conv_1154[FLOAT, 512x512x1x1]
%onnx::Conv_1157[FLOAT, 512x512x3x3]
) {
%onnx::Conv_1158 = Identity(%onnx::Conv_1098)
%onnx::Conv_1155 = Identity(%onnx::Conv_1098)
%onnx::Conv_1152 = Identity(%onnx::Conv_1098)
%onnx::Conv_1149 = Identity(%onnx::Conv_1098)
%onnx::Conv_1146 = Identity(%onnx::Conv_1098)
%onnx::Conv_1143 = Identity(%onnx::Conv_1098)
%onnx::Conv_1140 = Identity(%onnx::Conv_1098)
%onnx::Conv_1137 = Identity(%onnx::Conv_1098)
%onnx::Conv_1134 = Identity(%onnx::Conv_1098)
%onnx::Conv_1131 = Identity(%onnx::Conv_1098)
%onnx::Conv_1128 = Identity(%onnx::Conv_1098)
%onnx::Conv_1125 = Identity(%onnx::Conv_1098)
%onnx::Conv_1122 = Identity(%onnx::Conv_1098)
%onnx::Conv_1119 = Identity(%onnx::Conv_1098)
%onnx::Conv_1116 = Identity(%onnx::Conv_1098)
%onnx::Conv_1113 = Identity(%onnx::Conv_1098)
%onnx::Conv_1110 = Identity(%onnx::Conv_1098)
%onnx::Conv_1107 = Identity(%onnx::Conv_1098)
%onnx::Conv_1104 = Identity(%onnx::Conv_1098)
%onnx::Conv_1101 = Identity(%onnx::Conv_1098)
%onnx::Conv_1095 = Identity(%onnx::Conv_1035)
%onnx::Conv_1092 = Identity(%onnx::Conv_1035)
%onnx::Conv_1089 = Identity(%onnx::Conv_1035)
%onnx::Conv_1086 = Identity(%onnx::Conv_1035)
%onnx::Conv_1083 = Identity(%onnx::Conv_1035)
%onnx::Conv_1080 = Identity(%onnx::Conv_1035)
%onnx::Conv_1077 = Identity(%onnx::Conv_1035)
%onnx::Conv_1074 = Identity(%onnx::Conv_1035)
%onnx::Conv_1071 = Identity(%onnx::Conv_1035)
%onnx::Conv_1068 = Identity(%onnx::Conv_1035)
%onnx::Conv_1065 = Identity(%onnx::Conv_1035)
%onnx::Conv_1062 = Identity(%onnx::Conv_1035)
%onnx::Conv_1059 = Identity(%onnx::Conv_1035)
%onnx::Conv_1056 = Identity(%onnx::Conv_1035)
%onnx::Conv_1053 = Identity(%onnx::Conv_1035)
%onnx::Conv_1050 = Identity(%onnx::Conv_1035)
%onnx::Conv_1047 = Identity(%onnx::Conv_1035)
%onnx::Conv_1044 = Identity(%onnx::Conv_1035)
%onnx::Conv_1041 = Identity(%onnx::Conv_1035)
%onnx::Conv_1038 = Identity(%onnx::Conv_1035)
%onnx::Conv_1032 = Identity(%onnx::Conv_969)
%onnx::Conv_1029 = Identity(%onnx::Conv_969)
%onnx::Conv_1026 = Identity(%onnx::Conv_969)
%onnx::Conv_1023 = Identity(%onnx::Conv_969)
%onnx::Conv_1020 = Identity(%onnx::Conv_969)
%onnx::Conv_1017 = Identity(%onnx::Conv_969)
%onnx::Conv_1014 = Identity(%onnx::Conv_969)
%onnx::Conv_1011 = Identity(%onnx::Conv_969)
%onnx::Conv_1008 = Identity(%onnx::Conv_969)
%onnx::Conv_1005 = Identity(%onnx::Conv_969)
%onnx::Conv_1002 = Identity(%onnx::Conv_969)
%onnx::Conv_999 = Identity(%onnx::Conv_969)
%onnx::Conv_996 = Identity(%onnx::Conv_969)
%onnx::Conv_993 = Identity(%onnx::Conv_969)
%onnx::Conv_990 = Identity(%onnx::Conv_969)
%onnx::Conv_987 = Identity(%onnx::Conv_969)
%onnx::Conv_984 = Identity(%onnx::Conv_969)
%onnx::Conv_981 = Identity(%onnx::Conv_969)
%onnx::Conv_978 = Identity(%onnx::Conv_969)
%onnx::Conv_975 = Identity(%onnx::Conv_969)
%onnx::Conv_972 = Identity(%onnx::Conv_969)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_968, %onnx::Conv_969)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_974, %onnx::Conv_975)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_977, %onnx::Conv_978)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/Add_5_output_0 = Add(%/layers.1/Add_4_output_0, %/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/Add_6_output_0 = Add(%/layers.1/Add_5_output_0, %/layers.1/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_6_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/Add_5_output_0 = Add(%/layers.2/Add_4_output_0, %/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/Add_6_output_0 = Add(%/layers.2/Add_5_output_0, %/layers.2/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_6_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_1025, %onnx::Conv_1026)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1028, %onnx::Conv_1029)
%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/Add_5_output_0 = Add(%/layers.3/Add_4_output_0, %/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/Add_6_output_0 = Add(%/layers.3/Add_5_output_0, %/layers.3/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_6_output_0, %onnx::Conv_1031, %onnx::Conv_1032)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1034, %onnx::Conv_1035)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1037, %onnx::Conv_1038)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_1040, %onnx::Conv_1041)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_1043, %onnx::Conv_1044)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_1046, %onnx::Conv_1047)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1049, %onnx::Conv_1050)
%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/Add_5_output_0 = Add(%/layers.5/Add_4_output_0, %/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/Add_6_output_0 = Add(%/layers.5/Add_5_output_0, %/layers.5/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_6_output_0, %onnx::Conv_1052, %onnx::Conv_1053)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1055, %onnx::Conv_1056)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1058, %onnx::Conv_1059)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_1061, %onnx::Conv_1062)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_1064, %onnx::Conv_1065)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_1067, %onnx::Conv_1068)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1070, %onnx::Conv_1071)
%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/Add_5_output_0 = Add(%/layers.6/Add_4_output_0, %/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/Add_6_output_0 = Add(%/layers.6/Add_5_output_0, %/layers.6/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_6_output_0, %onnx::Conv_1073, %onnx::Conv_1074)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1076, %onnx::Conv_1077)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1079, %onnx::Conv_1080)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_1082, %onnx::Conv_1083)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_1085, %onnx::Conv_1086)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_1088, %onnx::Conv_1089)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1091, %onnx::Conv_1092)
%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/Add_5_output_0 = Add(%/layers.7/Add_4_output_0, %/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/Add_6_output_0 = Add(%/layers.7/Add_5_output_0, %/layers.7/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_6_output_0, %onnx::Conv_1094, %onnx::Conv_1095)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1097, %onnx::Conv_1098)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1100, %onnx::Conv_1101)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_1103, %onnx::Conv_1104)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_1106, %onnx::Conv_1107)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_1109, %onnx::Conv_1110)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1112, %onnx::Conv_1113)
%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/Add_5_output_0 = Add(%/layers.9/Add_4_output_0, %/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/Add_6_output_0 = Add(%/layers.9/Add_5_output_0, %/layers.9/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_6_output_0, %onnx::Conv_1115, %onnx::Conv_1116)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1118, %onnx::Conv_1119)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1121, %onnx::Conv_1122)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_1124, %onnx::Conv_1125)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_1127, %onnx::Conv_1128)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_1130, %onnx::Conv_1131)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1133, %onnx::Conv_1134)
%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/Add_5_output_0 = Add(%/layers.10/Add_4_output_0, %/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/Add_6_output_0 = Add(%/layers.10/Add_5_output_0, %/layers.10/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_6_output_0, %onnx::Conv_1136, %onnx::Conv_1137)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1139, %onnx::Conv_1140)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1142, %onnx::Conv_1143)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_1145, %onnx::Conv_1146)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_1148, %onnx::Conv_1149)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_1151, %onnx::Conv_1152)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1154, %onnx::Conv_1155)
%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/Add_5_output_0 = Add(%/layers.11/Add_4_output_0, %/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/Add_6_output_0 = Add(%/layers.11/Add_5_output_0, %/layers.11/input_op.5/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_6_output_0, %onnx::Conv_1157, %onnx::Conv_1158)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%966 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %966
}
|
val_accuracy
| 93.159056
| 6,891,776,000
| 23,295,370
|
{'zcp_epe_nas': 115.69063393747989, 'zcp_fisher': 13.45583724975586, 'zcp_flops': 110268416000.0, 'zcp_grad_norm': 63.833370208740234, 'zcp_grasp': -1.17730712890625, 'zcp_jacov': -16.05256272599359, 'zcp_l2_norm': 1439.210693359375, 'zcp_nwot': 236.95237508868135, 'zcp_params': 23295370.0, 'zcp_plain': 0.029327739030122, 'zcp_snip': 552.277099609375, 'zcp_synflow': 131.65620077278317, 'zcp_zen': 122.8867416381836, 'zcp_val_accuracy': 0.8578726053237911}
| |
NASBench101_128654
|
NASBench101
|
128654
|
4dc4db89b249dbd5276d2b72276677df
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_977[FLOAT, 128x3x3x3]
%onnx::Conv_978[FLOAT, 128]
%onnx::Conv_980[FLOAT, 128x128x1x1]
%onnx::Conv_983[FLOAT, 128x128x1x1]
%onnx::Conv_986[FLOAT, 128x128x1x1]
%onnx::Conv_989[FLOAT, 128x128x1x1]
%onnx::Conv_992[FLOAT, 128x128x1x1]
%onnx::Conv_995[FLOAT, 128x128x1x1]
%onnx::Conv_998[FLOAT, 128x128x3x3]
%onnx::Conv_1001[FLOAT, 128x128x1x1]
%onnx::Conv_1004[FLOAT, 128x128x1x1]
%onnx::Conv_1007[FLOAT, 128x128x1x1]
%onnx::Conv_1010[FLOAT, 128x128x1x1]
%onnx::Conv_1013[FLOAT, 128x128x1x1]
%onnx::Conv_1016[FLOAT, 128x128x1x1]
%onnx::Conv_1019[FLOAT, 128x128x3x3]
%onnx::Conv_1022[FLOAT, 128x128x1x1]
%onnx::Conv_1025[FLOAT, 128x128x1x1]
%onnx::Conv_1028[FLOAT, 128x128x1x1]
%onnx::Conv_1031[FLOAT, 128x128x1x1]
%onnx::Conv_1034[FLOAT, 128x128x1x1]
%onnx::Conv_1037[FLOAT, 128x128x1x1]
%onnx::Conv_1040[FLOAT, 128x128x3x3]
%onnx::Conv_1043[FLOAT, 256x128x1x1]
%onnx::Conv_1044[FLOAT, 256]
%onnx::Conv_1046[FLOAT, 256x256x1x1]
%onnx::Conv_1049[FLOAT, 256x128x1x1]
%onnx::Conv_1052[FLOAT, 256x256x1x1]
%onnx::Conv_1055[FLOAT, 256x128x1x1]
%onnx::Conv_1058[FLOAT, 256x128x1x1]
%onnx::Conv_1061[FLOAT, 256x256x3x3]
%onnx::Conv_1064[FLOAT, 256x256x1x1]
%onnx::Conv_1067[FLOAT, 256x256x1x1]
%onnx::Conv_1070[FLOAT, 256x256x1x1]
%onnx::Conv_1073[FLOAT, 256x256x1x1]
%onnx::Conv_1076[FLOAT, 256x256x1x1]
%onnx::Conv_1079[FLOAT, 256x256x1x1]
%onnx::Conv_1082[FLOAT, 256x256x3x3]
%onnx::Conv_1085[FLOAT, 256x256x1x1]
%onnx::Conv_1088[FLOAT, 256x256x1x1]
%onnx::Conv_1091[FLOAT, 256x256x1x1]
%onnx::Conv_1094[FLOAT, 256x256x1x1]
%onnx::Conv_1097[FLOAT, 256x256x1x1]
%onnx::Conv_1100[FLOAT, 256x256x1x1]
%onnx::Conv_1103[FLOAT, 256x256x3x3]
%onnx::Conv_1106[FLOAT, 512x256x1x1]
%onnx::Conv_1107[FLOAT, 512]
%onnx::Conv_1109[FLOAT, 512x512x1x1]
%onnx::Conv_1112[FLOAT, 512x256x1x1]
%onnx::Conv_1115[FLOAT, 512x512x1x1]
%onnx::Conv_1118[FLOAT, 512x256x1x1]
%onnx::Conv_1121[FLOAT, 512x256x1x1]
%onnx::Conv_1124[FLOAT, 512x512x3x3]
%onnx::Conv_1127[FLOAT, 512x512x1x1]
%onnx::Conv_1130[FLOAT, 512x512x1x1]
%onnx::Conv_1133[FLOAT, 512x512x1x1]
%onnx::Conv_1136[FLOAT, 512x512x1x1]
%onnx::Conv_1139[FLOAT, 512x512x1x1]
%onnx::Conv_1142[FLOAT, 512x512x1x1]
%onnx::Conv_1145[FLOAT, 512x512x3x3]
%onnx::Conv_1148[FLOAT, 512x512x1x1]
%onnx::Conv_1151[FLOAT, 512x512x1x1]
%onnx::Conv_1154[FLOAT, 512x512x1x1]
%onnx::Conv_1157[FLOAT, 512x512x1x1]
%onnx::Conv_1160[FLOAT, 512x512x1x1]
%onnx::Conv_1163[FLOAT, 512x512x1x1]
%onnx::Conv_1166[FLOAT, 512x512x3x3]
) {
%onnx::Conv_1167 = Identity(%onnx::Conv_1107)
%onnx::Conv_1164 = Identity(%onnx::Conv_1107)
%onnx::Conv_1161 = Identity(%onnx::Conv_1107)
%onnx::Conv_1158 = Identity(%onnx::Conv_1107)
%onnx::Conv_1155 = Identity(%onnx::Conv_1107)
%onnx::Conv_1152 = Identity(%onnx::Conv_1107)
%onnx::Conv_1149 = Identity(%onnx::Conv_1107)
%onnx::Conv_1146 = Identity(%onnx::Conv_1107)
%onnx::Conv_1143 = Identity(%onnx::Conv_1107)
%onnx::Conv_1140 = Identity(%onnx::Conv_1107)
%onnx::Conv_1137 = Identity(%onnx::Conv_1107)
%onnx::Conv_1134 = Identity(%onnx::Conv_1107)
%onnx::Conv_1131 = Identity(%onnx::Conv_1107)
%onnx::Conv_1128 = Identity(%onnx::Conv_1107)
%onnx::Conv_1125 = Identity(%onnx::Conv_1107)
%onnx::Conv_1122 = Identity(%onnx::Conv_1107)
%onnx::Conv_1119 = Identity(%onnx::Conv_1107)
%onnx::Conv_1116 = Identity(%onnx::Conv_1107)
%onnx::Conv_1113 = Identity(%onnx::Conv_1107)
%onnx::Conv_1110 = Identity(%onnx::Conv_1107)
%onnx::Conv_1104 = Identity(%onnx::Conv_1044)
%onnx::Conv_1101 = Identity(%onnx::Conv_1044)
%onnx::Conv_1098 = Identity(%onnx::Conv_1044)
%onnx::Conv_1095 = Identity(%onnx::Conv_1044)
%onnx::Conv_1092 = Identity(%onnx::Conv_1044)
%onnx::Conv_1089 = Identity(%onnx::Conv_1044)
%onnx::Conv_1086 = Identity(%onnx::Conv_1044)
%onnx::Conv_1083 = Identity(%onnx::Conv_1044)
%onnx::Conv_1080 = Identity(%onnx::Conv_1044)
%onnx::Conv_1077 = Identity(%onnx::Conv_1044)
%onnx::Conv_1074 = Identity(%onnx::Conv_1044)
%onnx::Conv_1071 = Identity(%onnx::Conv_1044)
%onnx::Conv_1068 = Identity(%onnx::Conv_1044)
%onnx::Conv_1065 = Identity(%onnx::Conv_1044)
%onnx::Conv_1062 = Identity(%onnx::Conv_1044)
%onnx::Conv_1059 = Identity(%onnx::Conv_1044)
%onnx::Conv_1056 = Identity(%onnx::Conv_1044)
%onnx::Conv_1053 = Identity(%onnx::Conv_1044)
%onnx::Conv_1050 = Identity(%onnx::Conv_1044)
%onnx::Conv_1047 = Identity(%onnx::Conv_1044)
%onnx::Conv_1041 = Identity(%onnx::Conv_978)
%onnx::Conv_1038 = Identity(%onnx::Conv_978)
%onnx::Conv_1035 = Identity(%onnx::Conv_978)
%onnx::Conv_1032 = Identity(%onnx::Conv_978)
%onnx::Conv_1029 = Identity(%onnx::Conv_978)
%onnx::Conv_1026 = Identity(%onnx::Conv_978)
%onnx::Conv_1023 = Identity(%onnx::Conv_978)
%onnx::Conv_1020 = Identity(%onnx::Conv_978)
%onnx::Conv_1017 = Identity(%onnx::Conv_978)
%onnx::Conv_1014 = Identity(%onnx::Conv_978)
%onnx::Conv_1011 = Identity(%onnx::Conv_978)
%onnx::Conv_1008 = Identity(%onnx::Conv_978)
%onnx::Conv_1005 = Identity(%onnx::Conv_978)
%onnx::Conv_1002 = Identity(%onnx::Conv_978)
%onnx::Conv_999 = Identity(%onnx::Conv_978)
%onnx::Conv_996 = Identity(%onnx::Conv_978)
%onnx::Conv_993 = Identity(%onnx::Conv_978)
%onnx::Conv_990 = Identity(%onnx::Conv_978)
%onnx::Conv_987 = Identity(%onnx::Conv_978)
%onnx::Conv_984 = Identity(%onnx::Conv_978)
%onnx::Conv_981 = Identity(%onnx::Conv_978)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_977, %onnx::Conv_978)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/Add_2_output_0, %/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/Add_5_output_0 = Add(%/layers.1/Add_4_output_0, %/layers.1/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.1/Add_6_output_0 = Add(%/layers.1/Add_5_output_0, %/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_6_output_0)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/Add_2_output_0, %/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/Add_5_output_0 = Add(%/layers.2/Add_4_output_0, %/layers.2/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.2/Add_6_output_0 = Add(%/layers.2/Add_5_output_0, %/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_6_output_0)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_1025, %onnx::Conv_1026)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1028, %onnx::Conv_1029)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_1031, %onnx::Conv_1032)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1034, %onnx::Conv_1035)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/Add_2_output_0, %/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1037, %onnx::Conv_1038)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/Add_5_output_0 = Add(%/layers.3/Add_4_output_0, %/layers.3/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.3/Add_6_output_0 = Add(%/layers.3/Add_5_output_0, %/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_6_output_0)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_1040, %onnx::Conv_1041)
%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1043, %onnx::Conv_1044)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_1046, %onnx::Conv_1047)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1049, %onnx::Conv_1050)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_1052, %onnx::Conv_1053)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1055, %onnx::Conv_1056)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/Add_2_output_0, %/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1058, %onnx::Conv_1059)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/Add_5_output_0 = Add(%/layers.5/Add_4_output_0, %/layers.5/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.5/Add_6_output_0 = Add(%/layers.5/Add_5_output_0, %/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_6_output_0)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_1061, %onnx::Conv_1062)
%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1064, %onnx::Conv_1065)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_1067, %onnx::Conv_1068)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1070, %onnx::Conv_1071)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_1073, %onnx::Conv_1074)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1076, %onnx::Conv_1077)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/Add_2_output_0, %/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1079, %onnx::Conv_1080)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/Add_5_output_0 = Add(%/layers.6/Add_4_output_0, %/layers.6/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.6/Add_6_output_0 = Add(%/layers.6/Add_5_output_0, %/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_6_output_0)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_1082, %onnx::Conv_1083)
%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1085, %onnx::Conv_1086)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_1088, %onnx::Conv_1089)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1091, %onnx::Conv_1092)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_1094, %onnx::Conv_1095)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1097, %onnx::Conv_1098)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/Add_2_output_0, %/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1100, %onnx::Conv_1101)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/Add_5_output_0 = Add(%/layers.7/Add_4_output_0, %/layers.7/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.7/Add_6_output_0 = Add(%/layers.7/Add_5_output_0, %/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_6_output_0)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_1103, %onnx::Conv_1104)
%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1106, %onnx::Conv_1107)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_1109, %onnx::Conv_1110)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1112, %onnx::Conv_1113)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_1115, %onnx::Conv_1116)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1118, %onnx::Conv_1119)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/Add_2_output_0, %/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1121, %onnx::Conv_1122)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/Add_5_output_0 = Add(%/layers.9/Add_4_output_0, %/layers.9/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.9/Add_6_output_0 = Add(%/layers.9/Add_5_output_0, %/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_6_output_0)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_1124, %onnx::Conv_1125)
%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1127, %onnx::Conv_1128)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_1130, %onnx::Conv_1131)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1133, %onnx::Conv_1134)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_1136, %onnx::Conv_1137)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1139, %onnx::Conv_1140)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/Add_2_output_0, %/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1142, %onnx::Conv_1143)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/Add_5_output_0 = Add(%/layers.10/Add_4_output_0, %/layers.10/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.10/Add_6_output_0 = Add(%/layers.10/Add_5_output_0, %/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_6_output_0)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_1145, %onnx::Conv_1146)
%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1148, %onnx::Conv_1149)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_1151, %onnx::Conv_1152)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1154, %onnx::Conv_1155)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_1157, %onnx::Conv_1158)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1160, %onnx::Conv_1161)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/Add_2_output_0, %/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1163, %onnx::Conv_1164)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/Add_5_output_0 = Add(%/layers.11/Add_4_output_0, %/layers.11/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.11/Add_6_output_0 = Add(%/layers.11/Add_5_output_0, %/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_6_output_0)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.4/maxpool/MaxPool_output_0, %onnx::Conv_1166, %onnx::Conv_1167)
%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/vertex_op.5/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%975 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %975
}
|
val_accuracy
| 91.636616
| 4,442,302,464
| 14,873,994
|
{'zcp_epe_nas': 119.65389797108637, 'zcp_fisher': 3.459697484970092, 'zcp_flops': 71076839424.0, 'zcp_grad_norm': 34.08123779296875, 'zcp_grasp': -0.9551239013671871, 'zcp_jacov': -16.049263260224624, 'zcp_l2_norm': 1422.3529052734375, 'zcp_nwot': 235.86391144346578, 'zcp_params': 14873994.0, 'zcp_plain': 0.027060952037572004, 'zcp_snip': 302.2171325683594, 'zcp_synflow': 100.13752339086888, 'zcp_zen': 115.2118148803711, 'zcp_val_accuracy': 0.9247796535491941}
| |
NASBench101_101409
|
NASBench101
|
101409
|
3d5da559a80c5673339d4a08c18fe084
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_986[FLOAT, 128x3x3x3]
%onnx::Conv_987[FLOAT, 128]
%onnx::Conv_989[FLOAT, 64x128x1x1]
%onnx::Conv_990[FLOAT, 64]
%onnx::Conv_992[FLOAT, 64x64x3x3]
%onnx::Conv_995[FLOAT, 64x64x1x1]
%onnx::Conv_998[FLOAT, 64x128x1x1]
%onnx::Conv_1001[FLOAT, 64x64x1x1]
%onnx::Conv_1004[FLOAT, 64x128x1x1]
%onnx::Conv_1007[FLOAT, 64x64x3x3]
%onnx::Conv_1010[FLOAT, 64x128x1x1]
%onnx::Conv_1013[FLOAT, 64x64x3x3]
%onnx::Conv_1016[FLOAT, 64x64x1x1]
%onnx::Conv_1019[FLOAT, 64x128x1x1]
%onnx::Conv_1022[FLOAT, 64x64x1x1]
%onnx::Conv_1025[FLOAT, 64x128x1x1]
%onnx::Conv_1028[FLOAT, 64x64x3x3]
%onnx::Conv_1031[FLOAT, 64x128x1x1]
%onnx::Conv_1034[FLOAT, 64x64x3x3]
%onnx::Conv_1037[FLOAT, 64x64x1x1]
%onnx::Conv_1040[FLOAT, 64x128x1x1]
%onnx::Conv_1043[FLOAT, 64x64x1x1]
%onnx::Conv_1046[FLOAT, 64x128x1x1]
%onnx::Conv_1049[FLOAT, 64x64x3x3]
%onnx::Conv_1052[FLOAT, 128x128x1x1]
%onnx::Conv_1055[FLOAT, 128x128x3x3]
%onnx::Conv_1058[FLOAT, 128x128x1x1]
%onnx::Conv_1061[FLOAT, 128x128x1x1]
%onnx::Conv_1064[FLOAT, 128x128x1x1]
%onnx::Conv_1067[FLOAT, 128x128x1x1]
%onnx::Conv_1070[FLOAT, 128x128x3x3]
%onnx::Conv_1073[FLOAT, 128x256x1x1]
%onnx::Conv_1076[FLOAT, 128x128x3x3]
%onnx::Conv_1079[FLOAT, 128x128x1x1]
%onnx::Conv_1082[FLOAT, 128x256x1x1]
%onnx::Conv_1085[FLOAT, 128x128x1x1]
%onnx::Conv_1088[FLOAT, 128x256x1x1]
%onnx::Conv_1091[FLOAT, 128x128x3x3]
%onnx::Conv_1094[FLOAT, 128x256x1x1]
%onnx::Conv_1097[FLOAT, 128x128x3x3]
%onnx::Conv_1100[FLOAT, 128x128x1x1]
%onnx::Conv_1103[FLOAT, 128x256x1x1]
%onnx::Conv_1106[FLOAT, 128x128x1x1]
%onnx::Conv_1109[FLOAT, 128x256x1x1]
%onnx::Conv_1112[FLOAT, 128x128x3x3]
%onnx::Conv_1115[FLOAT, 256x256x1x1]
%onnx::Conv_1116[FLOAT, 256]
%onnx::Conv_1118[FLOAT, 256x256x3x3]
%onnx::Conv_1121[FLOAT, 256x256x1x1]
%onnx::Conv_1124[FLOAT, 256x256x1x1]
%onnx::Conv_1127[FLOAT, 256x256x1x1]
%onnx::Conv_1130[FLOAT, 256x256x1x1]
%onnx::Conv_1133[FLOAT, 256x256x3x3]
%onnx::Conv_1136[FLOAT, 256x512x1x1]
%onnx::Conv_1139[FLOAT, 256x256x3x3]
%onnx::Conv_1142[FLOAT, 256x256x1x1]
%onnx::Conv_1145[FLOAT, 256x512x1x1]
%onnx::Conv_1148[FLOAT, 256x256x1x1]
%onnx::Conv_1151[FLOAT, 256x512x1x1]
%onnx::Conv_1154[FLOAT, 256x256x3x3]
%onnx::Conv_1157[FLOAT, 256x512x1x1]
%onnx::Conv_1160[FLOAT, 256x256x3x3]
%onnx::Conv_1163[FLOAT, 256x256x1x1]
%onnx::Conv_1166[FLOAT, 256x512x1x1]
%onnx::Conv_1169[FLOAT, 256x256x1x1]
%onnx::Conv_1172[FLOAT, 256x512x1x1]
%onnx::Conv_1175[FLOAT, 256x256x3x3]
) {
%onnx::Conv_1176 = Identity(%onnx::Conv_1116)
%onnx::Conv_1173 = Identity(%onnx::Conv_1116)
%onnx::Conv_1170 = Identity(%onnx::Conv_1116)
%onnx::Conv_1167 = Identity(%onnx::Conv_1116)
%onnx::Conv_1164 = Identity(%onnx::Conv_1116)
%onnx::Conv_1161 = Identity(%onnx::Conv_1116)
%onnx::Conv_1158 = Identity(%onnx::Conv_1116)
%onnx::Conv_1155 = Identity(%onnx::Conv_1116)
%onnx::Conv_1152 = Identity(%onnx::Conv_1116)
%onnx::Conv_1149 = Identity(%onnx::Conv_1116)
%onnx::Conv_1146 = Identity(%onnx::Conv_1116)
%onnx::Conv_1143 = Identity(%onnx::Conv_1116)
%onnx::Conv_1140 = Identity(%onnx::Conv_1116)
%onnx::Conv_1137 = Identity(%onnx::Conv_1116)
%onnx::Conv_1134 = Identity(%onnx::Conv_1116)
%onnx::Conv_1131 = Identity(%onnx::Conv_1116)
%onnx::Conv_1128 = Identity(%onnx::Conv_1116)
%onnx::Conv_1125 = Identity(%onnx::Conv_1116)
%onnx::Conv_1122 = Identity(%onnx::Conv_1116)
%onnx::Conv_1119 = Identity(%onnx::Conv_1116)
%onnx::Conv_1113 = Identity(%onnx::Conv_987)
%onnx::Conv_1110 = Identity(%onnx::Conv_987)
%onnx::Conv_1107 = Identity(%onnx::Conv_987)
%onnx::Conv_1104 = Identity(%onnx::Conv_987)
%onnx::Conv_1101 = Identity(%onnx::Conv_987)
%onnx::Conv_1098 = Identity(%onnx::Conv_987)
%onnx::Conv_1095 = Identity(%onnx::Conv_987)
%onnx::Conv_1092 = Identity(%onnx::Conv_987)
%onnx::Conv_1089 = Identity(%onnx::Conv_987)
%onnx::Conv_1086 = Identity(%onnx::Conv_987)
%onnx::Conv_1083 = Identity(%onnx::Conv_987)
%onnx::Conv_1080 = Identity(%onnx::Conv_987)
%onnx::Conv_1077 = Identity(%onnx::Conv_987)
%onnx::Conv_1074 = Identity(%onnx::Conv_987)
%onnx::Conv_1071 = Identity(%onnx::Conv_987)
%onnx::Conv_1068 = Identity(%onnx::Conv_987)
%onnx::Conv_1065 = Identity(%onnx::Conv_987)
%onnx::Conv_1062 = Identity(%onnx::Conv_987)
%onnx::Conv_1059 = Identity(%onnx::Conv_987)
%onnx::Conv_1056 = Identity(%onnx::Conv_987)
%onnx::Conv_1053 = Identity(%onnx::Conv_987)
%onnx::Conv_1050 = Identity(%onnx::Conv_990)
%onnx::Conv_1047 = Identity(%onnx::Conv_990)
%onnx::Conv_1044 = Identity(%onnx::Conv_990)
%onnx::Conv_1041 = Identity(%onnx::Conv_990)
%onnx::Conv_1038 = Identity(%onnx::Conv_990)
%onnx::Conv_1035 = Identity(%onnx::Conv_990)
%onnx::Conv_1032 = Identity(%onnx::Conv_990)
%onnx::Conv_1029 = Identity(%onnx::Conv_990)
%onnx::Conv_1026 = Identity(%onnx::Conv_990)
%onnx::Conv_1023 = Identity(%onnx::Conv_990)
%onnx::Conv_1020 = Identity(%onnx::Conv_990)
%onnx::Conv_1017 = Identity(%onnx::Conv_990)
%onnx::Conv_1014 = Identity(%onnx::Conv_990)
%onnx::Conv_1011 = Identity(%onnx::Conv_990)
%onnx::Conv_1008 = Identity(%onnx::Conv_990)
%onnx::Conv_1005 = Identity(%onnx::Conv_990)
%onnx::Conv_1002 = Identity(%onnx::Conv_990)
%onnx::Conv_999 = Identity(%onnx::Conv_990)
%onnx::Conv_996 = Identity(%onnx::Conv_990)
%onnx::Conv_993 = Identity(%onnx::Conv_990)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_986, %onnx::Conv_987)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/Add_2_output_0, %/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/Add_5_output_0 = Add(%/layers.1/Add_4_output_0, %/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_5_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_6_output_0 = Add(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_4_output_0)
%/layers.1/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_6_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/Add_2_output_0, %/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_1025, %onnx::Conv_1026)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/Add_5_output_0 = Add(%/layers.2/Add_4_output_0, %/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_5_output_0, %onnx::Conv_1028, %onnx::Conv_1029)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_6_output_0 = Add(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_4_output_0)
%/layers.2/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_6_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_1031, %onnx::Conv_1032)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_1034, %onnx::Conv_1035)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_1037, %onnx::Conv_1038)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_1040, %onnx::Conv_1041)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/Add_2_output_0, %/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_1043, %onnx::Conv_1044)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_1046, %onnx::Conv_1047)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/Add_5_output_0 = Add(%/layers.3/Add_4_output_0, %/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_5_output_0, %onnx::Conv_1049, %onnx::Conv_1050)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_6_output_0 = Add(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_4_output_0)
%/layers.3/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_6_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1052, %onnx::Conv_1053)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_1055, %onnx::Conv_1056)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_1058, %onnx::Conv_1059)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1061, %onnx::Conv_1062)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/Add_2_output_0, %/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_1064, %onnx::Conv_1065)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_1067, %onnx::Conv_1068)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/Add_5_output_0 = Add(%/layers.5/Add_4_output_0, %/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_5_output_0, %onnx::Conv_1070, %onnx::Conv_1071)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_6_output_0 = Add(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_4_output_0)
%/layers.5/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_6_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_1073, %onnx::Conv_1074)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_1076, %onnx::Conv_1077)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_1079, %onnx::Conv_1080)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_1082, %onnx::Conv_1083)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/Add_2_output_0, %/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_1085, %onnx::Conv_1086)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_1088, %onnx::Conv_1089)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/Add_5_output_0 = Add(%/layers.6/Add_4_output_0, %/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_5_output_0, %onnx::Conv_1091, %onnx::Conv_1092)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_6_output_0 = Add(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_4_output_0)
%/layers.6/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_6_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_1094, %onnx::Conv_1095)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_1097, %onnx::Conv_1098)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_1100, %onnx::Conv_1101)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_1103, %onnx::Conv_1104)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/Add_2_output_0, %/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_1106, %onnx::Conv_1107)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_1109, %onnx::Conv_1110)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/Add_5_output_0 = Add(%/layers.7/Add_4_output_0, %/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_5_output_0, %onnx::Conv_1112, %onnx::Conv_1113)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_6_output_0 = Add(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_4_output_0)
%/layers.7/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_6_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1115, %onnx::Conv_1116)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_1118, %onnx::Conv_1119)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_1121, %onnx::Conv_1122)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1124, %onnx::Conv_1125)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/Add_2_output_0, %/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_1127, %onnx::Conv_1128)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1130, %onnx::Conv_1131)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/Add_5_output_0 = Add(%/layers.9/Add_4_output_0, %/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_5_output_0, %onnx::Conv_1133, %onnx::Conv_1134)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_6_output_0 = Add(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_4_output_0)
%/layers.9/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_6_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_1136, %onnx::Conv_1137)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_1139, %onnx::Conv_1140)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_1142, %onnx::Conv_1143)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_1145, %onnx::Conv_1146)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/Add_2_output_0, %/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_1148, %onnx::Conv_1149)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_1151, %onnx::Conv_1152)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/Add_5_output_0 = Add(%/layers.10/Add_4_output_0, %/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_5_output_0, %onnx::Conv_1154, %onnx::Conv_1155)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_6_output_0 = Add(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_4_output_0)
%/layers.10/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_6_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1157, %onnx::Conv_1158)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_1160, %onnx::Conv_1161)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_1163, %onnx::Conv_1164)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1166, %onnx::Conv_1167)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/Add_2_output_0, %/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_1169, %onnx::Conv_1170)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1172, %onnx::Conv_1173)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/Add_5_output_0 = Add(%/layers.11/Add_4_output_0, %/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_5_output_0, %onnx::Conv_1175, %onnx::Conv_1176)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_6_output_0 = Add(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_4_output_0)
%/layers.11/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_6_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.5/maxpool/MaxPool_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%984 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %984
}
|
val_accuracy
| 92.037261
| 1,940,006,912
| 6,491,146
|
{'zcp_epe_nas': 109.62213205618751, 'zcp_fisher': 207.36378479003906, 'zcp_flops': 31040110592.0, 'zcp_grad_norm': 319.5948181152344, 'zcp_grasp': -56.544921875, 'zcp_jacov': -16.0538759152152, 'zcp_l2_norm': 1189.562255859375, 'zcp_nwot': 226.96602244622997, 'zcp_params': 6491146.0, 'zcp_plain': 0.011232508346438, 'zcp_snip': 1893.542724609375, 'zcp_synflow': 139.83667321319862, 'zcp_zen': 112.21036529541016, 'zcp_val_accuracy': 0.921674668788909}
| |
NASBench101_198299
|
NASBench101
|
198299
|
78072ff92e3f0a4172c8d38f2314c840
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_770[FLOAT, 128x3x3x3]
%onnx::Conv_771[FLOAT, 128]
%onnx::Conv_773[FLOAT, 43x128x1x1]
%onnx::Conv_774[FLOAT, 43]
%onnx::Conv_776[FLOAT, 43x128x1x1]
%onnx::Conv_779[FLOAT, 43x43x3x3]
%onnx::Conv_782[FLOAT, 42x128x1x1]
%onnx::Conv_783[FLOAT, 42]
%onnx::Conv_785[FLOAT, 42x42x1x1]
%onnx::Conv_788[FLOAT, 43x128x1x1]
%onnx::Conv_791[FLOAT, 43x128x1x1]
%onnx::Conv_794[FLOAT, 43x43x3x3]
%onnx::Conv_797[FLOAT, 42x128x1x1]
%onnx::Conv_800[FLOAT, 42x42x1x1]
%onnx::Conv_803[FLOAT, 43x128x1x1]
%onnx::Conv_806[FLOAT, 43x128x1x1]
%onnx::Conv_809[FLOAT, 43x43x3x3]
%onnx::Conv_812[FLOAT, 42x128x1x1]
%onnx::Conv_815[FLOAT, 42x42x1x1]
%onnx::Conv_818[FLOAT, 86x128x1x1]
%onnx::Conv_819[FLOAT, 86]
%onnx::Conv_821[FLOAT, 85x128x1x1]
%onnx::Conv_822[FLOAT, 85]
%onnx::Conv_824[FLOAT, 85x85x3x3]
%onnx::Conv_827[FLOAT, 85x128x1x1]
%onnx::Conv_830[FLOAT, 85x85x1x1]
%onnx::Conv_833[FLOAT, 86x256x1x1]
%onnx::Conv_836[FLOAT, 85x256x1x1]
%onnx::Conv_839[FLOAT, 85x85x3x3]
%onnx::Conv_842[FLOAT, 85x256x1x1]
%onnx::Conv_845[FLOAT, 85x85x1x1]
%onnx::Conv_848[FLOAT, 86x256x1x1]
%onnx::Conv_851[FLOAT, 85x256x1x1]
%onnx::Conv_854[FLOAT, 85x85x3x3]
%onnx::Conv_857[FLOAT, 85x256x1x1]
%onnx::Conv_860[FLOAT, 85x85x1x1]
%onnx::Conv_863[FLOAT, 171x256x1x1]
%onnx::Conv_864[FLOAT, 171]
%onnx::Conv_866[FLOAT, 171x256x1x1]
%onnx::Conv_869[FLOAT, 171x171x3x3]
%onnx::Conv_872[FLOAT, 170x256x1x1]
%onnx::Conv_873[FLOAT, 170]
%onnx::Conv_875[FLOAT, 170x170x1x1]
%onnx::Conv_878[FLOAT, 171x512x1x1]
%onnx::Conv_881[FLOAT, 171x512x1x1]
%onnx::Conv_884[FLOAT, 171x171x3x3]
%onnx::Conv_887[FLOAT, 170x512x1x1]
%onnx::Conv_890[FLOAT, 170x170x1x1]
%onnx::Conv_893[FLOAT, 171x512x1x1]
%onnx::Conv_896[FLOAT, 171x512x1x1]
%onnx::Conv_899[FLOAT, 171x171x3x3]
%onnx::Conv_902[FLOAT, 170x512x1x1]
%onnx::Conv_905[FLOAT, 170x170x1x1]
) {
%onnx::Conv_906 = Identity(%onnx::Conv_873)
%onnx::Conv_903 = Identity(%onnx::Conv_873)
%onnx::Conv_900 = Identity(%onnx::Conv_864)
%onnx::Conv_897 = Identity(%onnx::Conv_864)
%onnx::Conv_894 = Identity(%onnx::Conv_864)
%onnx::Conv_891 = Identity(%onnx::Conv_873)
%onnx::Conv_888 = Identity(%onnx::Conv_873)
%onnx::Conv_885 = Identity(%onnx::Conv_864)
%onnx::Conv_882 = Identity(%onnx::Conv_864)
%onnx::Conv_879 = Identity(%onnx::Conv_864)
%onnx::Conv_876 = Identity(%onnx::Conv_873)
%onnx::Conv_870 = Identity(%onnx::Conv_864)
%onnx::Conv_867 = Identity(%onnx::Conv_864)
%onnx::Conv_861 = Identity(%onnx::Conv_822)
%onnx::Conv_858 = Identity(%onnx::Conv_822)
%onnx::Conv_855 = Identity(%onnx::Conv_822)
%onnx::Conv_852 = Identity(%onnx::Conv_822)
%onnx::Conv_849 = Identity(%onnx::Conv_819)
%onnx::Conv_846 = Identity(%onnx::Conv_822)
%onnx::Conv_843 = Identity(%onnx::Conv_822)
%onnx::Conv_840 = Identity(%onnx::Conv_822)
%onnx::Conv_837 = Identity(%onnx::Conv_822)
%onnx::Conv_834 = Identity(%onnx::Conv_819)
%onnx::Conv_831 = Identity(%onnx::Conv_822)
%onnx::Conv_828 = Identity(%onnx::Conv_822)
%onnx::Conv_825 = Identity(%onnx::Conv_822)
%onnx::Conv_816 = Identity(%onnx::Conv_783)
%onnx::Conv_813 = Identity(%onnx::Conv_783)
%onnx::Conv_810 = Identity(%onnx::Conv_774)
%onnx::Conv_807 = Identity(%onnx::Conv_774)
%onnx::Conv_804 = Identity(%onnx::Conv_774)
%onnx::Conv_801 = Identity(%onnx::Conv_783)
%onnx::Conv_798 = Identity(%onnx::Conv_783)
%onnx::Conv_795 = Identity(%onnx::Conv_774)
%onnx::Conv_792 = Identity(%onnx::Conv_774)
%onnx::Conv_789 = Identity(%onnx::Conv_774)
%onnx::Conv_786 = Identity(%onnx::Conv_783)
%onnx::Conv_780 = Identity(%onnx::Conv_774)
%onnx::Conv_777 = Identity(%onnx::Conv_774)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_770, %onnx::Conv_771)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_773, %onnx::Conv_774)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_776, %onnx::Conv_777)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_779, %onnx::Conv_780)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.1/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.1/Slice_output_0 = Slice(%/layers.1/vertex_op.2/maxpool/MaxPool_output_0, %/layers.1/Constant_3_output_0, %/layers.1/Constant_4_output_0, %/layers.1/Constant_2_output_0, %/layers.1/Constant_5_output_0)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_782, %onnx::Conv_783)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/Slice_output_0, %/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_785, %onnx::Conv_786)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_788, %onnx::Conv_789)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.2/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.2/Slice_output_0 = Slice(%/layers.2/vertex_op.2/maxpool/MaxPool_output_0, %/layers.2/Constant_3_output_0, %/layers.2/Constant_4_output_0, %/layers.2/Constant_2_output_0, %/layers.2/Constant_5_output_0)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_797, %onnx::Conv_798)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/Slice_output_0, %/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_800, %onnx::Conv_801)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_803, %onnx::Conv_804)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_806, %onnx::Conv_807)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_809, %onnx::Conv_810)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.3/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.3/Slice_output_0 = Slice(%/layers.3/vertex_op.2/maxpool/MaxPool_output_0, %/layers.3/Constant_3_output_0, %/layers.3/Constant_4_output_0, %/layers.3/Constant_2_output_0, %/layers.3/Constant_5_output_0)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_812, %onnx::Conv_813)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/Slice_output_0, %/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_815, %onnx::Conv_816)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_818, %onnx::Conv_819)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_821, %onnx::Conv_822)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.5/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.5/Slice_output_0 = Slice(%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/Constant_3_output_0, %/layers.5/Constant_4_output_0, %/layers.5/Constant_2_output_0, %/layers.5/Constant_5_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/Slice_output_0, %/layers.5/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_824, %onnx::Conv_825)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_827, %onnx::Conv_828)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.2/maxpool/MaxPool_output_0, %/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_830, %onnx::Conv_831)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_833, %onnx::Conv_834)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_836, %onnx::Conv_837)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.6/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.6/Slice_output_0 = Slice(%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/Constant_3_output_0, %/layers.6/Constant_4_output_0, %/layers.6/Constant_2_output_0, %/layers.6/Constant_5_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/Slice_output_0, %/layers.6/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_839, %onnx::Conv_840)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_842, %onnx::Conv_843)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.2/maxpool/MaxPool_output_0, %/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_845, %onnx::Conv_846)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_848, %onnx::Conv_849)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_851, %onnx::Conv_852)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.7/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.7/Slice_output_0 = Slice(%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/Constant_3_output_0, %/layers.7/Constant_4_output_0, %/layers.7/Constant_2_output_0, %/layers.7/Constant_5_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/Slice_output_0, %/layers.7/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_854, %onnx::Conv_855)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_857, %onnx::Conv_858)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.2/maxpool/MaxPool_output_0, %/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.9/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.9/Slice_output_0 = Slice(%/layers.9/vertex_op.2/maxpool/MaxPool_output_0, %/layers.9/Constant_3_output_0, %/layers.9/Constant_4_output_0, %/layers.9/Constant_2_output_0, %/layers.9/Constant_5_output_0)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/Slice_output_0, %/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.10/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.10/Slice_output_0 = Slice(%/layers.10/vertex_op.2/maxpool/MaxPool_output_0, %/layers.10/Constant_3_output_0, %/layers.10/Constant_4_output_0, %/layers.10/Constant_2_output_0, %/layers.10/Constant_5_output_0)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/Slice_output_0, %/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_3_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_4_output_0 = Constant[value = <Tensor>]()
%/layers.11/Constant_5_output_0 = Constant[value = <Tensor>]()
%/layers.11/Slice_output_0 = Slice(%/layers.11/vertex_op.2/maxpool/MaxPool_output_0, %/layers.11/Constant_3_output_0, %/layers.11/Constant_4_output_0, %/layers.11/Constant_2_output_0, %/layers.11/Constant_5_output_0)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/Slice_output_0, %/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%768 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %768
}
|
val_accuracy
| 92.23758
| 621,945,216
| 2,034,318
|
{'zcp_epe_nas': 118.68430364788378, 'zcp_fisher': 1.16056227684021, 'zcp_flops': 9951123456.0, 'zcp_grad_norm': 21.622812271118164, 'zcp_grasp': -0.34933853149414, 'zcp_jacov': -16.0509161403367, 'zcp_l2_norm': 835.03369140625, 'zcp_nwot': 215.36088495416018, 'zcp_params': 2034318.0, 'zcp_plain': 0.0007105061085890001, 'zcp_snip': 123.41704559326172, 'zcp_synflow': 66.05862418856572, 'zcp_zen': 81.28589630126953, 'zcp_val_accuracy': 0.922576129436492}
| |
NASBench101_410613
|
NASBench101
|
410613
|
f81eb192e050b2bfefea8495051689e1
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_878[FLOAT, 128x3x3x3]
%onnx::Conv_879[FLOAT, 128]
%onnx::Conv_881[FLOAT, 128x128x1x1]
%onnx::Conv_884[FLOAT, 128x128x1x1]
%onnx::Conv_887[FLOAT, 128x128x3x3]
%onnx::Conv_890[FLOAT, 128x128x3x3]
%onnx::Conv_893[FLOAT, 128x128x1x1]
%onnx::Conv_896[FLOAT, 128x128x1x1]
%onnx::Conv_899[FLOAT, 128x128x1x1]
%onnx::Conv_902[FLOAT, 128x128x1x1]
%onnx::Conv_905[FLOAT, 128x128x3x3]
%onnx::Conv_908[FLOAT, 128x128x3x3]
%onnx::Conv_911[FLOAT, 128x128x1x1]
%onnx::Conv_914[FLOAT, 128x128x1x1]
%onnx::Conv_917[FLOAT, 128x128x1x1]
%onnx::Conv_920[FLOAT, 128x128x1x1]
%onnx::Conv_923[FLOAT, 128x128x3x3]
%onnx::Conv_926[FLOAT, 128x128x3x3]
%onnx::Conv_929[FLOAT, 128x128x1x1]
%onnx::Conv_932[FLOAT, 128x128x1x1]
%onnx::Conv_935[FLOAT, 256x128x1x1]
%onnx::Conv_936[FLOAT, 256]
%onnx::Conv_938[FLOAT, 256x256x1x1]
%onnx::Conv_941[FLOAT, 256x256x3x3]
%onnx::Conv_944[FLOAT, 256x256x3x3]
%onnx::Conv_947[FLOAT, 256x128x1x1]
%onnx::Conv_950[FLOAT, 256x128x1x1]
%onnx::Conv_953[FLOAT, 256x256x1x1]
%onnx::Conv_956[FLOAT, 256x256x1x1]
%onnx::Conv_959[FLOAT, 256x256x3x3]
%onnx::Conv_962[FLOAT, 256x256x3x3]
%onnx::Conv_965[FLOAT, 256x256x1x1]
%onnx::Conv_968[FLOAT, 256x256x1x1]
%onnx::Conv_971[FLOAT, 256x256x1x1]
%onnx::Conv_974[FLOAT, 256x256x1x1]
%onnx::Conv_977[FLOAT, 256x256x3x3]
%onnx::Conv_980[FLOAT, 256x256x3x3]
%onnx::Conv_983[FLOAT, 256x256x1x1]
%onnx::Conv_986[FLOAT, 256x256x1x1]
%onnx::Conv_989[FLOAT, 512x256x1x1]
%onnx::Conv_990[FLOAT, 512]
%onnx::Conv_992[FLOAT, 512x512x1x1]
%onnx::Conv_995[FLOAT, 512x512x3x3]
%onnx::Conv_998[FLOAT, 512x512x3x3]
%onnx::Conv_1001[FLOAT, 512x256x1x1]
%onnx::Conv_1004[FLOAT, 512x256x1x1]
%onnx::Conv_1007[FLOAT, 512x512x1x1]
%onnx::Conv_1010[FLOAT, 512x512x1x1]
%onnx::Conv_1013[FLOAT, 512x512x3x3]
%onnx::Conv_1016[FLOAT, 512x512x3x3]
%onnx::Conv_1019[FLOAT, 512x512x1x1]
%onnx::Conv_1022[FLOAT, 512x512x1x1]
%onnx::Conv_1025[FLOAT, 512x512x1x1]
%onnx::Conv_1028[FLOAT, 512x512x1x1]
%onnx::Conv_1031[FLOAT, 512x512x3x3]
%onnx::Conv_1034[FLOAT, 512x512x3x3]
%onnx::Conv_1037[FLOAT, 512x512x1x1]
%onnx::Conv_1040[FLOAT, 512x512x1x1]
) {
%onnx::Conv_1041 = Identity(%onnx::Conv_990)
%onnx::Conv_1038 = Identity(%onnx::Conv_990)
%onnx::Conv_1035 = Identity(%onnx::Conv_990)
%onnx::Conv_1032 = Identity(%onnx::Conv_990)
%onnx::Conv_1029 = Identity(%onnx::Conv_990)
%onnx::Conv_1026 = Identity(%onnx::Conv_990)
%onnx::Conv_1023 = Identity(%onnx::Conv_990)
%onnx::Conv_1020 = Identity(%onnx::Conv_990)
%onnx::Conv_1017 = Identity(%onnx::Conv_990)
%onnx::Conv_1014 = Identity(%onnx::Conv_990)
%onnx::Conv_1011 = Identity(%onnx::Conv_990)
%onnx::Conv_1008 = Identity(%onnx::Conv_990)
%onnx::Conv_1005 = Identity(%onnx::Conv_990)
%onnx::Conv_1002 = Identity(%onnx::Conv_990)
%onnx::Conv_999 = Identity(%onnx::Conv_990)
%onnx::Conv_996 = Identity(%onnx::Conv_990)
%onnx::Conv_993 = Identity(%onnx::Conv_990)
%onnx::Conv_987 = Identity(%onnx::Conv_936)
%onnx::Conv_984 = Identity(%onnx::Conv_936)
%onnx::Conv_981 = Identity(%onnx::Conv_936)
%onnx::Conv_978 = Identity(%onnx::Conv_936)
%onnx::Conv_975 = Identity(%onnx::Conv_936)
%onnx::Conv_972 = Identity(%onnx::Conv_936)
%onnx::Conv_969 = Identity(%onnx::Conv_936)
%onnx::Conv_966 = Identity(%onnx::Conv_936)
%onnx::Conv_963 = Identity(%onnx::Conv_936)
%onnx::Conv_960 = Identity(%onnx::Conv_936)
%onnx::Conv_957 = Identity(%onnx::Conv_936)
%onnx::Conv_954 = Identity(%onnx::Conv_936)
%onnx::Conv_951 = Identity(%onnx::Conv_936)
%onnx::Conv_948 = Identity(%onnx::Conv_936)
%onnx::Conv_945 = Identity(%onnx::Conv_936)
%onnx::Conv_942 = Identity(%onnx::Conv_936)
%onnx::Conv_939 = Identity(%onnx::Conv_936)
%onnx::Conv_933 = Identity(%onnx::Conv_879)
%onnx::Conv_930 = Identity(%onnx::Conv_879)
%onnx::Conv_927 = Identity(%onnx::Conv_879)
%onnx::Conv_924 = Identity(%onnx::Conv_879)
%onnx::Conv_921 = Identity(%onnx::Conv_879)
%onnx::Conv_918 = Identity(%onnx::Conv_879)
%onnx::Conv_915 = Identity(%onnx::Conv_879)
%onnx::Conv_912 = Identity(%onnx::Conv_879)
%onnx::Conv_909 = Identity(%onnx::Conv_879)
%onnx::Conv_906 = Identity(%onnx::Conv_879)
%onnx::Conv_903 = Identity(%onnx::Conv_879)
%onnx::Conv_900 = Identity(%onnx::Conv_879)
%onnx::Conv_897 = Identity(%onnx::Conv_879)
%onnx::Conv_894 = Identity(%onnx::Conv_879)
%onnx::Conv_891 = Identity(%onnx::Conv_879)
%onnx::Conv_888 = Identity(%onnx::Conv_879)
%onnx::Conv_885 = Identity(%onnx::Conv_879)
%onnx::Conv_882 = Identity(%onnx::Conv_879)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_878, %onnx::Conv_879)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/Add_3_output_0, %/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/Add_5_output_0 = Add(%/layers.1/Add_4_output_0, %/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_5_output_0)
%/layers.1/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_6_output_0 = Add(%/layers.1/vertex_op.5/maxpool/MaxPool_output_0, %/layers.1/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_6_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_908, %onnx::Conv_909)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_6_output_0, %onnx::Conv_911, %onnx::Conv_912)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/Add_3_output_0, %/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/Add_5_output_0 = Add(%/layers.2/Add_4_output_0, %/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_5_output_0)
%/layers.2/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_6_output_0, %onnx::Conv_914, %onnx::Conv_915)
%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_6_output_0 = Add(%/layers.2/vertex_op.5/maxpool/MaxPool_output_0, %/layers.2/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_6_output_0, %onnx::Conv_917, %onnx::Conv_918)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_920, %onnx::Conv_921)
%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_923, %onnx::Conv_924)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_926, %onnx::Conv_927)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_6_output_0, %onnx::Conv_929, %onnx::Conv_930)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/Add_3_output_0, %/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/Add_5_output_0 = Add(%/layers.3/Add_4_output_0, %/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_5_output_0)
%/layers.3/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_6_output_0, %onnx::Conv_932, %onnx::Conv_933)
%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_6_output_0 = Add(%/layers.3/vertex_op.5/maxpool/MaxPool_output_0, %/layers.3/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Add_6_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_935, %onnx::Conv_936)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_938, %onnx::Conv_939)
%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_941, %onnx::Conv_942)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_944, %onnx::Conv_945)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_947, %onnx::Conv_948)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/Add_3_output_0, %/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/Add_5_output_0 = Add(%/layers.5/Add_4_output_0, %/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_5_output_0)
%/layers.5/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_950, %onnx::Conv_951)
%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_6_output_0 = Add(%/layers.5/vertex_op.5/maxpool/MaxPool_output_0, %/layers.5/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_6_output_0, %onnx::Conv_953, %onnx::Conv_954)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_956, %onnx::Conv_957)
%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_959, %onnx::Conv_960)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_962, %onnx::Conv_963)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_6_output_0, %onnx::Conv_965, %onnx::Conv_966)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/Add_3_output_0, %/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/Add_5_output_0 = Add(%/layers.6/Add_4_output_0, %/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_5_output_0)
%/layers.6/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_6_output_0, %onnx::Conv_968, %onnx::Conv_969)
%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_6_output_0 = Add(%/layers.6/vertex_op.5/maxpool/MaxPool_output_0, %/layers.6/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_6_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_974, %onnx::Conv_975)
%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_977, %onnx::Conv_978)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_6_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/Add_3_output_0, %/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/Add_5_output_0 = Add(%/layers.7/Add_4_output_0, %/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_5_output_0)
%/layers.7/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_6_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_6_output_0 = Add(%/layers.7/vertex_op.5/maxpool/MaxPool_output_0, %/layers.7/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Add_6_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/Add_3_output_0, %/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/Add_5_output_0 = Add(%/layers.9/Add_4_output_0, %/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_5_output_0)
%/layers.9/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_6_output_0 = Add(%/layers.9/vertex_op.5/maxpool/MaxPool_output_0, %/layers.9/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_6_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_6_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/Add_3_output_0, %/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/Add_5_output_0 = Add(%/layers.10/Add_4_output_0, %/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_5_output_0)
%/layers.10/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_6_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_6_output_0 = Add(%/layers.10/vertex_op.5/maxpool/MaxPool_output_0, %/layers.10/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_6_output_0, %onnx::Conv_1025, %onnx::Conv_1026)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_1028, %onnx::Conv_1029)
%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_1031, %onnx::Conv_1032)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_1034, %onnx::Conv_1035)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_6_output_0, %onnx::Conv_1037, %onnx::Conv_1038)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/Add_3_output_0, %/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/Add_5_output_0 = Add(%/layers.11/Add_4_output_0, %/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_5_output_0)
%/layers.11/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_6_output_0, %onnx::Conv_1040, %onnx::Conv_1041)
%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_6_output_0 = Add(%/layers.11/vertex_op.5/maxpool/MaxPool_output_0, %/layers.11/input_op.6/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Add_6_output_0)
%876 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %876
}
|
val_accuracy
| 90.304488
| 6,584,281,088
| 22,257,802
|
{'zcp_epe_nas': 93.37492791698335, 'zcp_fisher': 2966.58203125, 'zcp_flops': 105348497408.0, 'zcp_grad_norm': 944.9176635742188, 'zcp_grasp': -4188.0390625, 'zcp_jacov': -16.049656299931435, 'zcp_l2_norm': 1225.9312744140625, 'zcp_nwot': 234.92642624794757, 'zcp_params': 22257802.0, 'zcp_plain': 0.2058657258749, 'zcp_snip': 7714.8369140625, 'zcp_synflow': 129.9164711733704, 'zcp_zen': 122.6285171508789, 'zcp_val_accuracy': 0.901141822338104}
| |
NASBench101_34173
|
NASBench101
|
34173
|
14b1a531b8eefa99a8561da4fea29180
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_572[FLOAT, 128x3x3x3]
%onnx::Conv_573[FLOAT, 128]
%onnx::Conv_575[FLOAT, 64x128x1x1]
%onnx::Conv_576[FLOAT, 64]
%onnx::Conv_578[FLOAT, 64x64x3x3]
%onnx::Conv_581[FLOAT, 64x128x1x1]
%onnx::Conv_584[FLOAT, 64x128x1x1]
%onnx::Conv_587[FLOAT, 64x64x3x3]
%onnx::Conv_590[FLOAT, 64x128x1x1]
%onnx::Conv_593[FLOAT, 64x128x1x1]
%onnx::Conv_596[FLOAT, 64x64x3x3]
%onnx::Conv_599[FLOAT, 64x128x1x1]
%onnx::Conv_602[FLOAT, 128x128x1x1]
%onnx::Conv_605[FLOAT, 128x128x3x3]
%onnx::Conv_608[FLOAT, 128x128x1x1]
%onnx::Conv_611[FLOAT, 128x256x1x1]
%onnx::Conv_614[FLOAT, 128x128x3x3]
%onnx::Conv_617[FLOAT, 128x256x1x1]
%onnx::Conv_620[FLOAT, 128x256x1x1]
%onnx::Conv_623[FLOAT, 128x128x3x3]
%onnx::Conv_626[FLOAT, 128x256x1x1]
%onnx::Conv_629[FLOAT, 256x256x1x1]
%onnx::Conv_630[FLOAT, 256]
%onnx::Conv_632[FLOAT, 256x256x3x3]
%onnx::Conv_635[FLOAT, 256x256x1x1]
%onnx::Conv_638[FLOAT, 256x512x1x1]
%onnx::Conv_641[FLOAT, 256x256x3x3]
%onnx::Conv_644[FLOAT, 256x512x1x1]
%onnx::Conv_647[FLOAT, 256x512x1x1]
%onnx::Conv_650[FLOAT, 256x256x3x3]
%onnx::Conv_653[FLOAT, 256x512x1x1]
) {
%onnx::Conv_654 = Identity(%onnx::Conv_630)
%onnx::Conv_651 = Identity(%onnx::Conv_630)
%onnx::Conv_648 = Identity(%onnx::Conv_630)
%onnx::Conv_645 = Identity(%onnx::Conv_630)
%onnx::Conv_642 = Identity(%onnx::Conv_630)
%onnx::Conv_639 = Identity(%onnx::Conv_630)
%onnx::Conv_636 = Identity(%onnx::Conv_630)
%onnx::Conv_633 = Identity(%onnx::Conv_630)
%onnx::Conv_627 = Identity(%onnx::Conv_573)
%onnx::Conv_624 = Identity(%onnx::Conv_573)
%onnx::Conv_621 = Identity(%onnx::Conv_573)
%onnx::Conv_618 = Identity(%onnx::Conv_573)
%onnx::Conv_615 = Identity(%onnx::Conv_573)
%onnx::Conv_612 = Identity(%onnx::Conv_573)
%onnx::Conv_609 = Identity(%onnx::Conv_573)
%onnx::Conv_606 = Identity(%onnx::Conv_573)
%onnx::Conv_603 = Identity(%onnx::Conv_573)
%onnx::Conv_600 = Identity(%onnx::Conv_576)
%onnx::Conv_597 = Identity(%onnx::Conv_576)
%onnx::Conv_594 = Identity(%onnx::Conv_576)
%onnx::Conv_591 = Identity(%onnx::Conv_576)
%onnx::Conv_588 = Identity(%onnx::Conv_576)
%onnx::Conv_585 = Identity(%onnx::Conv_576)
%onnx::Conv_582 = Identity(%onnx::Conv_576)
%onnx::Conv_579 = Identity(%onnx::Conv_576)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_572, %onnx::Conv_573)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_575, %onnx::Conv_576)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_578, %onnx::Conv_579)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_581, %onnx::Conv_582)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/Add_1_output_0, %/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_2_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/Add_3_output_0, %/layers.1/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.1/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_4_output_0)
%/layers.1/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.4/maxpool/MaxPool_output_0, %/layers.1/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_584, %onnx::Conv_585)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_587, %onnx::Conv_588)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_590, %onnx::Conv_591)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/Add_1_output_0, %/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_2_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/Add_3_output_0, %/layers.2/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.2/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_4_output_0)
%/layers.2/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.4/maxpool/MaxPool_output_0, %/layers.2/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_593, %onnx::Conv_594)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_596, %onnx::Conv_597)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_599, %onnx::Conv_600)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/Add_1_output_0, %/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_2_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/Add_3_output_0, %/layers.3/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.3/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_4_output_0)
%/layers.3/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.4/maxpool/MaxPool_output_0, %/layers.3/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_602, %onnx::Conv_603)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_605, %onnx::Conv_606)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_608, %onnx::Conv_609)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/Add_1_output_0, %/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_2_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/Add_3_output_0, %/layers.5/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.5/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_4_output_0)
%/layers.5/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.4/maxpool/MaxPool_output_0, %/layers.5/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_611, %onnx::Conv_612)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_614, %onnx::Conv_615)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_617, %onnx::Conv_618)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/Add_1_output_0, %/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_2_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/Add_3_output_0, %/layers.6/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.6/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_4_output_0)
%/layers.6/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.4/maxpool/MaxPool_output_0, %/layers.6/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_620, %onnx::Conv_621)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_623, %onnx::Conv_624)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_626, %onnx::Conv_627)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/Add_1_output_0, %/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_2_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/Add_3_output_0, %/layers.7/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.7/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_4_output_0)
%/layers.7/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.4/maxpool/MaxPool_output_0, %/layers.7/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_629, %onnx::Conv_630)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_632, %onnx::Conv_633)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_635, %onnx::Conv_636)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/Add_1_output_0, %/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_2_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/Add_3_output_0, %/layers.9/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.9/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_4_output_0)
%/layers.9/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.4/maxpool/MaxPool_output_0, %/layers.9/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_638, %onnx::Conv_639)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_641, %onnx::Conv_642)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_644, %onnx::Conv_645)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/Add_1_output_0, %/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_2_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/Add_3_output_0, %/layers.10/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.10/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_4_output_0)
%/layers.10/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.4/maxpool/MaxPool_output_0, %/layers.10/vertex_op.4/maxpool/MaxPool_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_647, %onnx::Conv_648)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_650, %onnx::Conv_651)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_653, %onnx::Conv_654)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/Add_1_output_0, %/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.2/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_2_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/Add_3_output_0, %/layers.11/vertex_op.2/maxpool/MaxPool_output_0)
%/layers.11/vertex_op.3/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_4_output_0)
%/layers.11/vertex_op.4/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/vertex_op.3/maxpool/MaxPool_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.4/maxpool/MaxPool_output_0, %/layers.11/vertex_op.4/maxpool/MaxPool_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%570 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %570
}
|
val_accuracy
| 84.845752
| 964,306,944
| 3,207,690
|
{'zcp_epe_nas': 91.17469046264713, 'zcp_fisher': 174.70362854003906, 'zcp_flops': 15428911104.0, 'zcp_grad_norm': 187.89195251464844, 'zcp_grasp': -271.04736328125, 'zcp_jacov': -16.05358042141953, 'zcp_l2_norm': 544.5073852539062, 'zcp_nwot': 214.18788554345954, 'zcp_params': 3207690.0, 'zcp_plain': 0.487712174654006, 'zcp_snip': 1231.91943359375, 'zcp_synflow': 73.53999990042048, 'zcp_zen': 66.50757598876953, 'zcp_val_accuracy': 0.91015625}
| |
NASBench101_221117
|
NASBench101
|
221117
|
86002858ede01c97a7707ec24d6946e8
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_887[FLOAT, 128x3x3x3]
%onnx::Conv_888[FLOAT, 128]
%onnx::Conv_890[FLOAT, 64x128x1x1]
%onnx::Conv_891[FLOAT, 64]
%onnx::Conv_893[FLOAT, 64x64x3x3]
%onnx::Conv_896[FLOAT, 64x64x3x3]
%onnx::Conv_899[FLOAT, 64x64x3x3]
%onnx::Conv_902[FLOAT, 64x128x1x1]
%onnx::Conv_905[FLOAT, 64x64x3x3]
%onnx::Conv_908[FLOAT, 64x128x1x1]
%onnx::Conv_911[FLOAT, 64x64x3x3]
%onnx::Conv_914[FLOAT, 64x64x3x3]
%onnx::Conv_917[FLOAT, 64x64x3x3]
%onnx::Conv_920[FLOAT, 64x128x1x1]
%onnx::Conv_923[FLOAT, 64x64x3x3]
%onnx::Conv_926[FLOAT, 64x128x1x1]
%onnx::Conv_929[FLOAT, 64x64x3x3]
%onnx::Conv_932[FLOAT, 64x64x3x3]
%onnx::Conv_935[FLOAT, 64x64x3x3]
%onnx::Conv_938[FLOAT, 64x128x1x1]
%onnx::Conv_941[FLOAT, 64x64x3x3]
%onnx::Conv_944[FLOAT, 128x128x1x1]
%onnx::Conv_947[FLOAT, 128x128x3x3]
%onnx::Conv_950[FLOAT, 128x128x3x3]
%onnx::Conv_953[FLOAT, 128x128x3x3]
%onnx::Conv_956[FLOAT, 128x128x1x1]
%onnx::Conv_959[FLOAT, 128x128x3x3]
%onnx::Conv_962[FLOAT, 128x256x1x1]
%onnx::Conv_965[FLOAT, 128x128x3x3]
%onnx::Conv_968[FLOAT, 128x128x3x3]
%onnx::Conv_971[FLOAT, 128x128x3x3]
%onnx::Conv_974[FLOAT, 128x256x1x1]
%onnx::Conv_977[FLOAT, 128x128x3x3]
%onnx::Conv_980[FLOAT, 128x256x1x1]
%onnx::Conv_983[FLOAT, 128x128x3x3]
%onnx::Conv_986[FLOAT, 128x128x3x3]
%onnx::Conv_989[FLOAT, 128x128x3x3]
%onnx::Conv_992[FLOAT, 128x256x1x1]
%onnx::Conv_995[FLOAT, 128x128x3x3]
%onnx::Conv_998[FLOAT, 256x256x1x1]
%onnx::Conv_999[FLOAT, 256]
%onnx::Conv_1001[FLOAT, 256x256x3x3]
%onnx::Conv_1004[FLOAT, 256x256x3x3]
%onnx::Conv_1007[FLOAT, 256x256x3x3]
%onnx::Conv_1010[FLOAT, 256x256x1x1]
%onnx::Conv_1013[FLOAT, 256x256x3x3]
%onnx::Conv_1016[FLOAT, 256x512x1x1]
%onnx::Conv_1019[FLOAT, 256x256x3x3]
%onnx::Conv_1022[FLOAT, 256x256x3x3]
%onnx::Conv_1025[FLOAT, 256x256x3x3]
%onnx::Conv_1028[FLOAT, 256x512x1x1]
%onnx::Conv_1031[FLOAT, 256x256x3x3]
%onnx::Conv_1034[FLOAT, 256x512x1x1]
%onnx::Conv_1037[FLOAT, 256x256x3x3]
%onnx::Conv_1040[FLOAT, 256x256x3x3]
%onnx::Conv_1043[FLOAT, 256x256x3x3]
%onnx::Conv_1046[FLOAT, 256x512x1x1]
%onnx::Conv_1049[FLOAT, 256x256x3x3]
) {
%onnx::Conv_1050 = Identity(%onnx::Conv_999)
%onnx::Conv_1047 = Identity(%onnx::Conv_999)
%onnx::Conv_1044 = Identity(%onnx::Conv_999)
%onnx::Conv_1041 = Identity(%onnx::Conv_999)
%onnx::Conv_1038 = Identity(%onnx::Conv_999)
%onnx::Conv_1035 = Identity(%onnx::Conv_999)
%onnx::Conv_1032 = Identity(%onnx::Conv_999)
%onnx::Conv_1029 = Identity(%onnx::Conv_999)
%onnx::Conv_1026 = Identity(%onnx::Conv_999)
%onnx::Conv_1023 = Identity(%onnx::Conv_999)
%onnx::Conv_1020 = Identity(%onnx::Conv_999)
%onnx::Conv_1017 = Identity(%onnx::Conv_999)
%onnx::Conv_1014 = Identity(%onnx::Conv_999)
%onnx::Conv_1011 = Identity(%onnx::Conv_999)
%onnx::Conv_1008 = Identity(%onnx::Conv_999)
%onnx::Conv_1005 = Identity(%onnx::Conv_999)
%onnx::Conv_1002 = Identity(%onnx::Conv_999)
%onnx::Conv_996 = Identity(%onnx::Conv_888)
%onnx::Conv_993 = Identity(%onnx::Conv_888)
%onnx::Conv_990 = Identity(%onnx::Conv_888)
%onnx::Conv_987 = Identity(%onnx::Conv_888)
%onnx::Conv_984 = Identity(%onnx::Conv_888)
%onnx::Conv_981 = Identity(%onnx::Conv_888)
%onnx::Conv_978 = Identity(%onnx::Conv_888)
%onnx::Conv_975 = Identity(%onnx::Conv_888)
%onnx::Conv_972 = Identity(%onnx::Conv_888)
%onnx::Conv_969 = Identity(%onnx::Conv_888)
%onnx::Conv_966 = Identity(%onnx::Conv_888)
%onnx::Conv_963 = Identity(%onnx::Conv_888)
%onnx::Conv_960 = Identity(%onnx::Conv_888)
%onnx::Conv_957 = Identity(%onnx::Conv_888)
%onnx::Conv_954 = Identity(%onnx::Conv_888)
%onnx::Conv_951 = Identity(%onnx::Conv_888)
%onnx::Conv_948 = Identity(%onnx::Conv_888)
%onnx::Conv_945 = Identity(%onnx::Conv_888)
%onnx::Conv_942 = Identity(%onnx::Conv_891)
%onnx::Conv_939 = Identity(%onnx::Conv_891)
%onnx::Conv_936 = Identity(%onnx::Conv_891)
%onnx::Conv_933 = Identity(%onnx::Conv_891)
%onnx::Conv_930 = Identity(%onnx::Conv_891)
%onnx::Conv_927 = Identity(%onnx::Conv_891)
%onnx::Conv_924 = Identity(%onnx::Conv_891)
%onnx::Conv_921 = Identity(%onnx::Conv_891)
%onnx::Conv_918 = Identity(%onnx::Conv_891)
%onnx::Conv_915 = Identity(%onnx::Conv_891)
%onnx::Conv_912 = Identity(%onnx::Conv_891)
%onnx::Conv_909 = Identity(%onnx::Conv_891)
%onnx::Conv_906 = Identity(%onnx::Conv_891)
%onnx::Conv_903 = Identity(%onnx::Conv_891)
%onnx::Conv_900 = Identity(%onnx::Conv_891)
%onnx::Conv_897 = Identity(%onnx::Conv_891)
%onnx::Conv_894 = Identity(%onnx::Conv_891)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_887, %onnx::Conv_888)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/Add_3_output_0 = Add(%/layers.1/Add_2_output_0, %/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_899, %onnx::Conv_900)
%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_902, %onnx::Conv_903)
%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_3_output_0)
%/layers.1/Add_5_output_0 = Add(%/layers.1/Add_4_output_0, %/layers.1/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_5_output_0, %onnx::Conv_905, %onnx::Conv_906)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_6_output_0 = Add(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_4_output_0)
%/layers.1/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_6_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_908, %onnx::Conv_909)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0, %onnx::Conv_911, %onnx::Conv_912)
%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_914, %onnx::Conv_915)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/Add_3_output_0 = Add(%/layers.2/Add_2_output_0, %/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_917, %onnx::Conv_918)
%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_920, %onnx::Conv_921)
%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_3_output_0)
%/layers.2/Add_5_output_0 = Add(%/layers.2/Add_4_output_0, %/layers.2/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_5_output_0, %onnx::Conv_923, %onnx::Conv_924)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_6_output_0 = Add(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_4_output_0)
%/layers.2/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_6_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_926, %onnx::Conv_927)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0, %onnx::Conv_929, %onnx::Conv_930)
%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_932, %onnx::Conv_933)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/Add_3_output_0 = Add(%/layers.3/Add_2_output_0, %/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_935, %onnx::Conv_936)
%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_938, %onnx::Conv_939)
%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_3_output_0)
%/layers.3/Add_5_output_0 = Add(%/layers.3/Add_4_output_0, %/layers.3/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_5_output_0, %onnx::Conv_941, %onnx::Conv_942)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_6_output_0 = Add(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_4_output_0)
%/layers.3/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_6_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_944, %onnx::Conv_945)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0, %onnx::Conv_947, %onnx::Conv_948)
%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_950, %onnx::Conv_951)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/Add_3_output_0 = Add(%/layers.5/Add_2_output_0, %/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_953, %onnx::Conv_954)
%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_956, %onnx::Conv_957)
%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_3_output_0)
%/layers.5/Add_5_output_0 = Add(%/layers.5/Add_4_output_0, %/layers.5/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_5_output_0, %onnx::Conv_959, %onnx::Conv_960)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_6_output_0 = Add(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_4_output_0)
%/layers.5/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_6_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_962, %onnx::Conv_963)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0, %onnx::Conv_965, %onnx::Conv_966)
%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_968, %onnx::Conv_969)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/Add_3_output_0 = Add(%/layers.6/Add_2_output_0, %/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_971, %onnx::Conv_972)
%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_974, %onnx::Conv_975)
%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_3_output_0)
%/layers.6/Add_5_output_0 = Add(%/layers.6/Add_4_output_0, %/layers.6/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_5_output_0, %onnx::Conv_977, %onnx::Conv_978)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_6_output_0 = Add(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_4_output_0)
%/layers.6/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_6_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_980, %onnx::Conv_981)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0, %onnx::Conv_983, %onnx::Conv_984)
%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_986, %onnx::Conv_987)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/Add_3_output_0 = Add(%/layers.7/Add_2_output_0, %/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_989, %onnx::Conv_990)
%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_992, %onnx::Conv_993)
%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_3_output_0)
%/layers.7/Add_5_output_0 = Add(%/layers.7/Add_4_output_0, %/layers.7/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_5_output_0, %onnx::Conv_995, %onnx::Conv_996)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_6_output_0 = Add(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_4_output_0)
%/layers.7/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_6_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_998, %onnx::Conv_999)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0, %onnx::Conv_1001, %onnx::Conv_1002)
%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_1004, %onnx::Conv_1005)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/Add_3_output_0 = Add(%/layers.9/Add_2_output_0, %/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_1007, %onnx::Conv_1008)
%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_1010, %onnx::Conv_1011)
%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_3_output_0)
%/layers.9/Add_5_output_0 = Add(%/layers.9/Add_4_output_0, %/layers.9/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_5_output_0, %onnx::Conv_1013, %onnx::Conv_1014)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_6_output_0 = Add(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_4_output_0)
%/layers.9/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_6_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_1016, %onnx::Conv_1017)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0, %onnx::Conv_1019, %onnx::Conv_1020)
%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_1022, %onnx::Conv_1023)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/Add_3_output_0 = Add(%/layers.10/Add_2_output_0, %/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_1025, %onnx::Conv_1026)
%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_1028, %onnx::Conv_1029)
%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_3_output_0)
%/layers.10/Add_5_output_0 = Add(%/layers.10/Add_4_output_0, %/layers.10/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_5_output_0, %onnx::Conv_1031, %onnx::Conv_1032)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_6_output_0 = Add(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_4_output_0)
%/layers.10/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_6_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1034, %onnx::Conv_1035)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0, %onnx::Conv_1037, %onnx::Conv_1038)
%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_1040, %onnx::Conv_1041)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.1/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/Add_3_output_0 = Add(%/layers.11/Add_2_output_0, %/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_1043, %onnx::Conv_1044)
%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_1046, %onnx::Conv_1047)
%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_3_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.3/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_3_output_0)
%/layers.11/Add_5_output_0 = Add(%/layers.11/Add_4_output_0, %/layers.11/input_op.4/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_5_output_0, %onnx::Conv_1049, %onnx::Conv_1050)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_4_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_6_output_0 = Add(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_4_output_0)
%/layers.11/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_6_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.5/maxpool/MaxPool_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%885 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %885
}
|
val_accuracy
| 91.125804
| 3,010,996,224
| 10,183,050
|
{'zcp_epe_nas': 92.2603281441057, 'zcp_fisher': 887.4979248046875, 'zcp_flops': 48175939584.0, 'zcp_grad_norm': 560.4744873046875, 'zcp_grasp': -183.38671875, 'zcp_jacov': -16.05609192996885, 'zcp_l2_norm': 994.7036743164062, 'zcp_nwot': 224.18760086745016, 'zcp_params': 10183050.0, 'zcp_plain': 0.03324956819415, 'zcp_snip': 3345.305908203125, 'zcp_synflow': 165.89270128208645, 'zcp_zen': 117.77238464355469, 'zcp_val_accuracy': 0.9104567170143121}
| |
NASBench101_382833
|
NASBench101
|
382833
|
e77449588150e7362bfd3cdb021a446b
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_761[FLOAT, 128x3x3x3]
%onnx::Conv_762[FLOAT, 128]
%onnx::Conv_764[FLOAT, 64x128x1x1]
%onnx::Conv_765[FLOAT, 64]
%onnx::Conv_767[FLOAT, 64x64x1x1]
%onnx::Conv_770[FLOAT, 64x128x1x1]
%onnx::Conv_773[FLOAT, 64x64x1x1]
%onnx::Conv_776[FLOAT, 64x64x1x1]
%onnx::Conv_779[FLOAT, 64x128x1x1]
%onnx::Conv_782[FLOAT, 64x64x1x1]
%onnx::Conv_785[FLOAT, 64x128x1x1]
%onnx::Conv_788[FLOAT, 64x64x1x1]
%onnx::Conv_791[FLOAT, 64x64x1x1]
%onnx::Conv_794[FLOAT, 64x128x1x1]
%onnx::Conv_797[FLOAT, 64x64x1x1]
%onnx::Conv_800[FLOAT, 64x128x1x1]
%onnx::Conv_803[FLOAT, 64x64x1x1]
%onnx::Conv_806[FLOAT, 64x64x1x1]
%onnx::Conv_809[FLOAT, 128x128x1x1]
%onnx::Conv_812[FLOAT, 128x128x1x1]
%onnx::Conv_815[FLOAT, 128x128x1x1]
%onnx::Conv_818[FLOAT, 128x128x1x1]
%onnx::Conv_821[FLOAT, 128x128x1x1]
%onnx::Conv_824[FLOAT, 128x256x1x1]
%onnx::Conv_827[FLOAT, 128x128x1x1]
%onnx::Conv_830[FLOAT, 128x256x1x1]
%onnx::Conv_833[FLOAT, 128x128x1x1]
%onnx::Conv_836[FLOAT, 128x128x1x1]
%onnx::Conv_839[FLOAT, 128x256x1x1]
%onnx::Conv_842[FLOAT, 128x128x1x1]
%onnx::Conv_845[FLOAT, 128x256x1x1]
%onnx::Conv_848[FLOAT, 128x128x1x1]
%onnx::Conv_851[FLOAT, 128x128x1x1]
%onnx::Conv_854[FLOAT, 256x256x1x1]
%onnx::Conv_855[FLOAT, 256]
%onnx::Conv_857[FLOAT, 256x256x1x1]
%onnx::Conv_860[FLOAT, 256x256x1x1]
%onnx::Conv_863[FLOAT, 256x256x1x1]
%onnx::Conv_866[FLOAT, 256x256x1x1]
%onnx::Conv_869[FLOAT, 256x512x1x1]
%onnx::Conv_872[FLOAT, 256x256x1x1]
%onnx::Conv_875[FLOAT, 256x512x1x1]
%onnx::Conv_878[FLOAT, 256x256x1x1]
%onnx::Conv_881[FLOAT, 256x256x1x1]
%onnx::Conv_884[FLOAT, 256x512x1x1]
%onnx::Conv_887[FLOAT, 256x256x1x1]
%onnx::Conv_890[FLOAT, 256x512x1x1]
%onnx::Conv_893[FLOAT, 256x256x1x1]
%onnx::Conv_896[FLOAT, 256x256x1x1]
) {
%onnx::Conv_897 = Identity(%onnx::Conv_855)
%onnx::Conv_894 = Identity(%onnx::Conv_855)
%onnx::Conv_891 = Identity(%onnx::Conv_855)
%onnx::Conv_888 = Identity(%onnx::Conv_855)
%onnx::Conv_885 = Identity(%onnx::Conv_855)
%onnx::Conv_882 = Identity(%onnx::Conv_855)
%onnx::Conv_879 = Identity(%onnx::Conv_855)
%onnx::Conv_876 = Identity(%onnx::Conv_855)
%onnx::Conv_873 = Identity(%onnx::Conv_855)
%onnx::Conv_870 = Identity(%onnx::Conv_855)
%onnx::Conv_867 = Identity(%onnx::Conv_855)
%onnx::Conv_864 = Identity(%onnx::Conv_855)
%onnx::Conv_861 = Identity(%onnx::Conv_855)
%onnx::Conv_858 = Identity(%onnx::Conv_855)
%onnx::Conv_852 = Identity(%onnx::Conv_762)
%onnx::Conv_849 = Identity(%onnx::Conv_762)
%onnx::Conv_846 = Identity(%onnx::Conv_762)
%onnx::Conv_843 = Identity(%onnx::Conv_762)
%onnx::Conv_840 = Identity(%onnx::Conv_762)
%onnx::Conv_837 = Identity(%onnx::Conv_762)
%onnx::Conv_834 = Identity(%onnx::Conv_762)
%onnx::Conv_831 = Identity(%onnx::Conv_762)
%onnx::Conv_828 = Identity(%onnx::Conv_762)
%onnx::Conv_825 = Identity(%onnx::Conv_762)
%onnx::Conv_822 = Identity(%onnx::Conv_762)
%onnx::Conv_819 = Identity(%onnx::Conv_762)
%onnx::Conv_816 = Identity(%onnx::Conv_762)
%onnx::Conv_813 = Identity(%onnx::Conv_762)
%onnx::Conv_810 = Identity(%onnx::Conv_762)
%onnx::Conv_807 = Identity(%onnx::Conv_765)
%onnx::Conv_804 = Identity(%onnx::Conv_765)
%onnx::Conv_801 = Identity(%onnx::Conv_765)
%onnx::Conv_798 = Identity(%onnx::Conv_765)
%onnx::Conv_795 = Identity(%onnx::Conv_765)
%onnx::Conv_792 = Identity(%onnx::Conv_765)
%onnx::Conv_789 = Identity(%onnx::Conv_765)
%onnx::Conv_786 = Identity(%onnx::Conv_765)
%onnx::Conv_783 = Identity(%onnx::Conv_765)
%onnx::Conv_780 = Identity(%onnx::Conv_765)
%onnx::Conv_777 = Identity(%onnx::Conv_765)
%onnx::Conv_774 = Identity(%onnx::Conv_765)
%onnx::Conv_771 = Identity(%onnx::Conv_765)
%onnx::Conv_768 = Identity(%onnx::Conv_765)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_761, %onnx::Conv_762)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_764, %onnx::Conv_765)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_767, %onnx::Conv_768)
%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_770, %onnx::Conv_771)
%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_1_output_0 = Add(%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/Add_1_output_0, %/layers.1/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_773, %onnx::Conv_774)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_776, %onnx::Conv_777)
%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_4_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_779, %onnx::Conv_780)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_782, %onnx::Conv_783)
%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_785, %onnx::Conv_786)
%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_1_output_0 = Add(%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/Add_1_output_0, %/layers.2/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_788, %onnx::Conv_789)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_4_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_797, %onnx::Conv_798)
%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_800, %onnx::Conv_801)
%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_1_output_0 = Add(%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/Add_1_output_0, %/layers.3/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_803, %onnx::Conv_804)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_806, %onnx::Conv_807)
%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_4_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_809, %onnx::Conv_810)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_812, %onnx::Conv_813)
%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_815, %onnx::Conv_816)
%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_1_output_0 = Add(%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/Add_1_output_0, %/layers.5/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_818, %onnx::Conv_819)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_821, %onnx::Conv_822)
%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_4_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_824, %onnx::Conv_825)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_827, %onnx::Conv_828)
%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_830, %onnx::Conv_831)
%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_1_output_0 = Add(%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/Add_1_output_0, %/layers.6/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_833, %onnx::Conv_834)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_836, %onnx::Conv_837)
%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_4_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_839, %onnx::Conv_840)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_842, %onnx::Conv_843)
%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_845, %onnx::Conv_846)
%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_1_output_0 = Add(%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/Add_1_output_0, %/layers.7/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_848, %onnx::Conv_849)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_851, %onnx::Conv_852)
%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_4_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_854, %onnx::Conv_855)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_857, %onnx::Conv_858)
%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_1_output_0 = Add(%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/Add_1_output_0, %/layers.9/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_4_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_1_output_0 = Add(%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/Add_1_output_0, %/layers.10/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_4_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_1_output_0 = Add(%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/vertex_op.2/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/Add_1_output_0, %/layers.11/input_op.3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.4/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_4_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/vertex_op.5/maxpool/MaxPool_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%759 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %759
}
|
val_accuracy
| 88.521636
| 516,827,136
| 1,664,778
|
{'zcp_epe_nas': 74.54047025249923, 'zcp_fisher': 31.327184677124023, 'zcp_flops': 8269234176.0, 'zcp_grad_norm': 115.37684631347656, 'zcp_grasp': -16.080322265625, 'zcp_jacov': -16.05554572738727, 'zcp_l2_norm': 844.7763671875, 'zcp_nwot': 221.89740974193623, 'zcp_params': 1664778.0, 'zcp_plain': 0.073549471795558, 'zcp_snip': 592.9468994140625, 'zcp_synflow': 98.87137594139153, 'zcp_zen': 73.37715148925781, 'zcp_val_accuracy': 0.926582515239715}
| |
NASBench101_178226
|
NASBench101
|
178226
|
6be8080e335585f54c55f779ac53365b
|
graph torch_jit (
%input.1[FLOAT, 1x3x32x32]
%classifier.weight[FLOAT, 10x512]
%classifier.bias[FLOAT, 10]
%onnx::Conv_761[FLOAT, 128x3x3x3]
%onnx::Conv_762[FLOAT, 128]
%onnx::Conv_764[FLOAT, 64x128x1x1]
%onnx::Conv_765[FLOAT, 64]
%onnx::Conv_767[FLOAT, 64x128x1x1]
%onnx::Conv_770[FLOAT, 64x64x3x3]
%onnx::Conv_773[FLOAT, 64x64x1x1]
%onnx::Conv_776[FLOAT, 64x64x3x3]
%onnx::Conv_779[FLOAT, 64x128x1x1]
%onnx::Conv_782[FLOAT, 64x128x1x1]
%onnx::Conv_785[FLOAT, 64x64x3x3]
%onnx::Conv_788[FLOAT, 64x64x1x1]
%onnx::Conv_791[FLOAT, 64x64x3x3]
%onnx::Conv_794[FLOAT, 64x128x1x1]
%onnx::Conv_797[FLOAT, 64x128x1x1]
%onnx::Conv_800[FLOAT, 64x64x3x3]
%onnx::Conv_803[FLOAT, 64x64x1x1]
%onnx::Conv_806[FLOAT, 64x64x3x3]
%onnx::Conv_809[FLOAT, 128x128x1x1]
%onnx::Conv_812[FLOAT, 128x128x1x1]
%onnx::Conv_815[FLOAT, 128x128x3x3]
%onnx::Conv_818[FLOAT, 128x128x1x1]
%onnx::Conv_821[FLOAT, 128x128x3x3]
%onnx::Conv_824[FLOAT, 128x256x1x1]
%onnx::Conv_827[FLOAT, 128x256x1x1]
%onnx::Conv_830[FLOAT, 128x128x3x3]
%onnx::Conv_833[FLOAT, 128x128x1x1]
%onnx::Conv_836[FLOAT, 128x128x3x3]
%onnx::Conv_839[FLOAT, 128x256x1x1]
%onnx::Conv_842[FLOAT, 128x256x1x1]
%onnx::Conv_845[FLOAT, 128x128x3x3]
%onnx::Conv_848[FLOAT, 128x128x1x1]
%onnx::Conv_851[FLOAT, 128x128x3x3]
%onnx::Conv_854[FLOAT, 256x256x1x1]
%onnx::Conv_855[FLOAT, 256]
%onnx::Conv_857[FLOAT, 256x256x1x1]
%onnx::Conv_860[FLOAT, 256x256x3x3]
%onnx::Conv_863[FLOAT, 256x256x1x1]
%onnx::Conv_866[FLOAT, 256x256x3x3]
%onnx::Conv_869[FLOAT, 256x512x1x1]
%onnx::Conv_872[FLOAT, 256x512x1x1]
%onnx::Conv_875[FLOAT, 256x256x3x3]
%onnx::Conv_878[FLOAT, 256x256x1x1]
%onnx::Conv_881[FLOAT, 256x256x3x3]
%onnx::Conv_884[FLOAT, 256x512x1x1]
%onnx::Conv_887[FLOAT, 256x512x1x1]
%onnx::Conv_890[FLOAT, 256x256x3x3]
%onnx::Conv_893[FLOAT, 256x256x1x1]
%onnx::Conv_896[FLOAT, 256x256x3x3]
) {
%onnx::Conv_897 = Identity(%onnx::Conv_855)
%onnx::Conv_894 = Identity(%onnx::Conv_855)
%onnx::Conv_891 = Identity(%onnx::Conv_855)
%onnx::Conv_888 = Identity(%onnx::Conv_855)
%onnx::Conv_885 = Identity(%onnx::Conv_855)
%onnx::Conv_882 = Identity(%onnx::Conv_855)
%onnx::Conv_879 = Identity(%onnx::Conv_855)
%onnx::Conv_876 = Identity(%onnx::Conv_855)
%onnx::Conv_873 = Identity(%onnx::Conv_855)
%onnx::Conv_870 = Identity(%onnx::Conv_855)
%onnx::Conv_867 = Identity(%onnx::Conv_855)
%onnx::Conv_864 = Identity(%onnx::Conv_855)
%onnx::Conv_861 = Identity(%onnx::Conv_855)
%onnx::Conv_858 = Identity(%onnx::Conv_855)
%onnx::Conv_852 = Identity(%onnx::Conv_762)
%onnx::Conv_849 = Identity(%onnx::Conv_762)
%onnx::Conv_846 = Identity(%onnx::Conv_762)
%onnx::Conv_843 = Identity(%onnx::Conv_762)
%onnx::Conv_840 = Identity(%onnx::Conv_762)
%onnx::Conv_837 = Identity(%onnx::Conv_762)
%onnx::Conv_834 = Identity(%onnx::Conv_762)
%onnx::Conv_831 = Identity(%onnx::Conv_762)
%onnx::Conv_828 = Identity(%onnx::Conv_762)
%onnx::Conv_825 = Identity(%onnx::Conv_762)
%onnx::Conv_822 = Identity(%onnx::Conv_762)
%onnx::Conv_819 = Identity(%onnx::Conv_762)
%onnx::Conv_816 = Identity(%onnx::Conv_762)
%onnx::Conv_813 = Identity(%onnx::Conv_762)
%onnx::Conv_810 = Identity(%onnx::Conv_762)
%onnx::Conv_807 = Identity(%onnx::Conv_765)
%onnx::Conv_804 = Identity(%onnx::Conv_765)
%onnx::Conv_801 = Identity(%onnx::Conv_765)
%onnx::Conv_798 = Identity(%onnx::Conv_765)
%onnx::Conv_795 = Identity(%onnx::Conv_765)
%onnx::Conv_792 = Identity(%onnx::Conv_765)
%onnx::Conv_789 = Identity(%onnx::Conv_765)
%onnx::Conv_786 = Identity(%onnx::Conv_765)
%onnx::Conv_783 = Identity(%onnx::Conv_765)
%onnx::Conv_780 = Identity(%onnx::Conv_765)
%onnx::Conv_777 = Identity(%onnx::Conv_765)
%onnx::Conv_774 = Identity(%onnx::Conv_765)
%onnx::Conv_771 = Identity(%onnx::Conv_765)
%onnx::Conv_768 = Identity(%onnx::Conv_765)
%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%input.1, %onnx::Conv_761, %onnx::Conv_762)
%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.0/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_764, %onnx::Conv_765)
%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_output_0 = Add(%/layers.1/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_output_0)
%/layers.1/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_output_0)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.0/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %onnx::Conv_767, %onnx::Conv_768)
%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_1_output_0 = Add(%/layers.1/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_1_output_0)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_1_output_0, %onnx::Conv_770, %onnx::Conv_771)
%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_2_output_0 = Add(%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Add_2_output_0, %onnx::Conv_773, %onnx::Conv_774)
%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.1/Add_3_output_0 = Add(%/layers.1/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/Constant_2_output_0)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_3_output_0, %onnx::Conv_776, %onnx::Conv_777)
%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.1/Add_4_output_0 = Add(%/layers.1/vertex_op.1/maxpool/MaxPool_output_0, %/layers.1/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.1/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.1/Add_4_output_0)
%/layers.1/Concat_output_0 = Concat[axis = 1](%/layers.1/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.1/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_779, %onnx::Conv_780)
%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_output_0 = Add(%/layers.2/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_output_0)
%/layers.2/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_output_0)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.1/Concat_output_0, %onnx::Conv_782, %onnx::Conv_783)
%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_1_output_0 = Add(%/layers.2/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_1_output_0)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_1_output_0, %onnx::Conv_785, %onnx::Conv_786)
%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_2_output_0 = Add(%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Add_2_output_0, %onnx::Conv_788, %onnx::Conv_789)
%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.2/Add_3_output_0 = Add(%/layers.2/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/Constant_2_output_0)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_3_output_0, %onnx::Conv_791, %onnx::Conv_792)
%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.2/Add_4_output_0 = Add(%/layers.2/vertex_op.1/maxpool/MaxPool_output_0, %/layers.2/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.2/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.2/Add_4_output_0)
%/layers.2/Concat_output_0 = Concat[axis = 1](%/layers.2/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.2/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_794, %onnx::Conv_795)
%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_output_0 = Add(%/layers.3/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_output_0)
%/layers.3/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_output_0)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.2/Concat_output_0, %onnx::Conv_797, %onnx::Conv_798)
%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_1_output_0 = Add(%/layers.3/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_1_output_0)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_1_output_0, %onnx::Conv_800, %onnx::Conv_801)
%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_2_output_0 = Add(%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.3/Add_2_output_0, %onnx::Conv_803, %onnx::Conv_804)
%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.3/Add_3_output_0 = Add(%/layers.3/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/Constant_2_output_0)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_3_output_0, %onnx::Conv_806, %onnx::Conv_807)
%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.3/Add_4_output_0 = Add(%/layers.3/vertex_op.1/maxpool/MaxPool_output_0, %/layers.3/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.3/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.3/Add_4_output_0)
%/layers.3/Concat_output_0 = Concat[axis = 1](%/layers.3/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.3/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.4/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.3/Concat_output_0)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_809, %onnx::Conv_810)
%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_output_0 = Add(%/layers.5/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_output_0)
%/layers.5/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_output_0)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.4/MaxPool_output_0, %onnx::Conv_812, %onnx::Conv_813)
%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_1_output_0 = Add(%/layers.5/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_1_output_0)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_1_output_0, %onnx::Conv_815, %onnx::Conv_816)
%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_2_output_0 = Add(%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Add_2_output_0, %onnx::Conv_818, %onnx::Conv_819)
%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.5/Add_3_output_0 = Add(%/layers.5/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/Constant_2_output_0)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_3_output_0, %onnx::Conv_821, %onnx::Conv_822)
%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.5/Add_4_output_0 = Add(%/layers.5/vertex_op.1/maxpool/MaxPool_output_0, %/layers.5/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.5/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.5/Add_4_output_0)
%/layers.5/Concat_output_0 = Concat[axis = 1](%/layers.5/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.5/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_824, %onnx::Conv_825)
%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_output_0 = Add(%/layers.6/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_output_0)
%/layers.6/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_output_0)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.5/Concat_output_0, %onnx::Conv_827, %onnx::Conv_828)
%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_1_output_0 = Add(%/layers.6/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_1_output_0)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_1_output_0, %onnx::Conv_830, %onnx::Conv_831)
%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_2_output_0 = Add(%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Add_2_output_0, %onnx::Conv_833, %onnx::Conv_834)
%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.6/Add_3_output_0 = Add(%/layers.6/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/Constant_2_output_0)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_3_output_0, %onnx::Conv_836, %onnx::Conv_837)
%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.6/Add_4_output_0 = Add(%/layers.6/vertex_op.1/maxpool/MaxPool_output_0, %/layers.6/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.6/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.6/Add_4_output_0)
%/layers.6/Concat_output_0 = Concat[axis = 1](%/layers.6/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.6/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_839, %onnx::Conv_840)
%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_output_0 = Add(%/layers.7/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_output_0)
%/layers.7/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_output_0)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.6/Concat_output_0, %onnx::Conv_842, %onnx::Conv_843)
%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_1_output_0 = Add(%/layers.7/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_1_output_0)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_1_output_0, %onnx::Conv_845, %onnx::Conv_846)
%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_2_output_0 = Add(%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.7/Add_2_output_0, %onnx::Conv_848, %onnx::Conv_849)
%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.7/Add_3_output_0 = Add(%/layers.7/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/Constant_2_output_0)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_3_output_0, %onnx::Conv_851, %onnx::Conv_852)
%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.7/Add_4_output_0 = Add(%/layers.7/vertex_op.1/maxpool/MaxPool_output_0, %/layers.7/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.7/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.7/Add_4_output_0)
%/layers.7/Concat_output_0 = Concat[axis = 1](%/layers.7/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.7/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.8/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [2, 2], pads = [0, 0, 0, 0], strides = [2, 2]](%/layers.7/Concat_output_0)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_854, %onnx::Conv_855)
%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_output_0 = Add(%/layers.9/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_output_0)
%/layers.9/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_output_0)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.8/MaxPool_output_0, %onnx::Conv_857, %onnx::Conv_858)
%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_1_output_0 = Add(%/layers.9/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_1_output_0)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_1_output_0, %onnx::Conv_860, %onnx::Conv_861)
%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_2_output_0 = Add(%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Add_2_output_0, %onnx::Conv_863, %onnx::Conv_864)
%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.9/Add_3_output_0 = Add(%/layers.9/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/Constant_2_output_0)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_3_output_0, %onnx::Conv_866, %onnx::Conv_867)
%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.9/Add_4_output_0 = Add(%/layers.9/vertex_op.1/maxpool/MaxPool_output_0, %/layers.9/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.9/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.9/Add_4_output_0)
%/layers.9/Concat_output_0 = Concat[axis = 1](%/layers.9/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.9/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_869, %onnx::Conv_870)
%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_output_0 = Add(%/layers.10/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_output_0)
%/layers.10/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_output_0)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.9/Concat_output_0, %onnx::Conv_872, %onnx::Conv_873)
%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_1_output_0 = Add(%/layers.10/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_1_output_0)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_1_output_0, %onnx::Conv_875, %onnx::Conv_876)
%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_2_output_0 = Add(%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Add_2_output_0, %onnx::Conv_878, %onnx::Conv_879)
%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.10/Add_3_output_0 = Add(%/layers.10/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/Constant_2_output_0)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_3_output_0, %onnx::Conv_881, %onnx::Conv_882)
%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.10/Add_4_output_0 = Add(%/layers.10/vertex_op.1/maxpool/MaxPool_output_0, %/layers.10/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.10/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.10/Add_4_output_0)
%/layers.10/Concat_output_0 = Concat[axis = 1](%/layers.10/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.10/vertex_op.5/maxpool/MaxPool_output_0)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_884, %onnx::Conv_885)
%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_output_0 = Add(%/layers.11/input_op.1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_output_0)
%/layers.11/vertex_op.1/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_output_0)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.10/Concat_output_0, %onnx::Conv_887, %onnx::Conv_888)
%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_1_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_1_output_0 = Add(%/layers.11/input_op.2/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_1_output_0)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_1_output_0, %onnx::Conv_890, %onnx::Conv_891)
%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_2_output_0 = Add(%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [1, 1], pads = [0, 0, 0, 0], strides = [1, 1]](%/layers.11/Add_2_output_0, %onnx::Conv_893, %onnx::Conv_894)
%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Constant_2_output_0 = Constant[value = <Scalar Tensor []>]()
%/layers.11/Add_3_output_0 = Add(%/layers.11/vertex_op.2/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/Constant_2_output_0)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0 = Conv[dilations = [1, 1], group = 1, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_3_output_0, %onnx::Conv_896, %onnx::Conv_897)
%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0 = Relu(%/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.0/Conv_output_0)
%/layers.11/Add_4_output_0 = Add(%/layers.11/vertex_op.1/maxpool/MaxPool_output_0, %/layers.11/vertex_op.4/conv3x3/conv_bn_relu/conv_bn_relu.2/Relu_output_0)
%/layers.11/vertex_op.5/maxpool/MaxPool_output_0 = MaxPool[ceil_mode = 0, kernel_shape = [3, 3], pads = [1, 1, 1, 1], strides = [1, 1]](%/layers.11/Add_4_output_0)
%/layers.11/Concat_output_0 = Concat[axis = 1](%/layers.11/vertex_op.3/conv1x1/conv_bn_relu/conv_bn_relu.2/Relu_output_0, %/layers.11/vertex_op.5/maxpool/MaxPool_output_0)
%/ReduceMean_output_0 = ReduceMean[axes = [2, 3], keepdims = 0](%/layers.11/Concat_output_0)
%759 = Gemm[alpha = 1, beta = 1, transB = 1](%/ReduceMean_output_0, %classifier.weight, %classifier.bias)
return %759
}
|
val_accuracy
| 91.085738
| 1,724,786,688
| 5,793,546
|
{'zcp_epe_nas': 77.070152901591, 'zcp_fisher': 41.36446762084961, 'zcp_flops': 27596587008.0, 'zcp_grad_norm': 128.0829315185547, 'zcp_grasp': -5.4698486328125, 'zcp_jacov': -16.057643917005986, 'zcp_l2_norm': 843.2216796875, 'zcp_nwot': 221.17726726298528, 'zcp_params': 5793546.0, 'zcp_plain': 0.17727263271808602, 'zcp_snip': 806.476318359375, 'zcp_synflow': 96.60755727743371, 'zcp_zen': 90.16925811767578, 'zcp_val_accuracy': 0.9260817170143121}
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.