Upload folder using huggingface_hub
Browse files
2d_tokenizer_imagenet512/config.yaml
ADDED
|
@@ -0,0 +1,126 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
arch:
|
| 2 |
+
ema: null
|
| 3 |
+
hparams:
|
| 4 |
+
loss_type: mse
|
| 5 |
+
restart_unused_codes: true
|
| 6 |
+
use_padding_idx: false
|
| 7 |
+
masked_dropout: 0.0
|
| 8 |
+
bottleneck_type: rq
|
| 9 |
+
embed_dim: 256
|
| 10 |
+
n_embed: 16384
|
| 11 |
+
latent_shape:
|
| 12 |
+
- 32
|
| 13 |
+
- 32
|
| 14 |
+
- 256
|
| 15 |
+
code_shape:
|
| 16 |
+
- 32
|
| 17 |
+
- 32
|
| 18 |
+
- 2
|
| 19 |
+
shared_codebook: true
|
| 20 |
+
decay: 0.99
|
| 21 |
+
latent_loss_weight: 0.25
|
| 22 |
+
checkpointing: true
|
| 23 |
+
type: rq-vae
|
| 24 |
+
code_hier: 1
|
| 25 |
+
ddconfig:
|
| 26 |
+
double_z: false
|
| 27 |
+
z_channels: 256
|
| 28 |
+
resolution: 256
|
| 29 |
+
in_channels: 3
|
| 30 |
+
out_ch: 3
|
| 31 |
+
ch: 128
|
| 32 |
+
ch_mult:
|
| 33 |
+
- 1
|
| 34 |
+
- 1
|
| 35 |
+
- 2
|
| 36 |
+
- 2
|
| 37 |
+
- 4
|
| 38 |
+
num_res_blocks: 2
|
| 39 |
+
attn_resolutions:
|
| 40 |
+
- 8
|
| 41 |
+
dropout: 0.0
|
| 42 |
+
dataset:
|
| 43 |
+
transform:
|
| 44 |
+
type: null
|
| 45 |
+
type: imagenet
|
| 46 |
+
transforms:
|
| 47 |
+
type: imagenet512x512
|
| 48 |
+
optimizer:
|
| 49 |
+
type: adam
|
| 50 |
+
max_gn: null
|
| 51 |
+
warmup:
|
| 52 |
+
mode: fix
|
| 53 |
+
start_from_zero: true
|
| 54 |
+
epoch: 0.5
|
| 55 |
+
multiplier: 1
|
| 56 |
+
buffer_epoch: 0
|
| 57 |
+
min_lr: 4.0e-05
|
| 58 |
+
init_lr: 4.0e-05
|
| 59 |
+
weight_decay: 0.0
|
| 60 |
+
betas:
|
| 61 |
+
- 0.5
|
| 62 |
+
- 0.9
|
| 63 |
+
grad_accm_steps: 1
|
| 64 |
+
experiment:
|
| 65 |
+
test_freq: 1
|
| 66 |
+
amp: false
|
| 67 |
+
batch_size: 20
|
| 68 |
+
epochs: 150
|
| 69 |
+
save_ckpt_freq: 5
|
| 70 |
+
total_batch_size: 160
|
| 71 |
+
gan:
|
| 72 |
+
disc:
|
| 73 |
+
optimizer:
|
| 74 |
+
type: adam
|
| 75 |
+
max_gn: null
|
| 76 |
+
warmup:
|
| 77 |
+
mode: fix
|
| 78 |
+
start_from_zero: true
|
| 79 |
+
epoch: 0.5
|
| 80 |
+
multiplier: 1
|
| 81 |
+
buffer_epoch: 0
|
| 82 |
+
min_lr: 4.0e-05
|
| 83 |
+
init_lr: 4.0e-05
|
| 84 |
+
weight_decay: 0.0
|
| 85 |
+
betas:
|
| 86 |
+
- 0.5
|
| 87 |
+
- 0.9
|
| 88 |
+
arch:
|
| 89 |
+
in_channels: 3
|
| 90 |
+
num_layers: 2
|
| 91 |
+
use_actnorm: false
|
| 92 |
+
ndf: 64
|
| 93 |
+
spectral_norm: false
|
| 94 |
+
loss:
|
| 95 |
+
disc_loss: hinge
|
| 96 |
+
gen_loss: vanilla
|
| 97 |
+
disc_weight: 0.75
|
| 98 |
+
perceptual_weight: 1.0
|
| 99 |
+
disc_start: 0
|
| 100 |
+
seed: 0
|
| 101 |
+
runtime:
|
| 102 |
+
args:
|
| 103 |
+
model_config: /cpfs01/user/cl424408/rq-vae-transformer-main/configs/imagenet256/stage1/in512-rqvae-32x32x2-arxiv-withaux.yaml
|
| 104 |
+
result_path: /cpfs01/user/cl424408/rq-vae-transformer-main/trained_models/in512_32x32x2_imagenet_shared_bsz20*8
|
| 105 |
+
load_path: ''
|
| 106 |
+
test_batch_size: 200
|
| 107 |
+
test_epoch: -1
|
| 108 |
+
postfix: ''
|
| 109 |
+
seed: 0
|
| 110 |
+
world_size: 8
|
| 111 |
+
local_rank: 0
|
| 112 |
+
node_rank: -1
|
| 113 |
+
dist_backend: nccl
|
| 114 |
+
timeout: 86400
|
| 115 |
+
eval: false
|
| 116 |
+
resume: false
|
| 117 |
+
rank: 0
|
| 118 |
+
extra_config: {}
|
| 119 |
+
distenv:
|
| 120 |
+
world_size: 8
|
| 121 |
+
world_rank: 0
|
| 122 |
+
local_rank: 0
|
| 123 |
+
num_gpus: 1
|
| 124 |
+
master: true
|
| 125 |
+
device_name: NVIDIA A100-SXM4-80GB
|
| 126 |
+
result_path: /cpfs01/user/cl424408/rq-vae-transformer-main/trained_models/in512_32x32x2_imagenet_shared_bsz20*8/in512-rqvae-32x32x2-arxiv-withaux/17092024_101256
|
2d_tokenizer_imagenet512/model.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ebc62c9a5a87cd31039898d2290c35b87dd83ceacd1cd3b540164f0132455ada
|
| 3 |
+
size 836123089
|