add network def
Browse filesSigned-off-by: Can-Zhao <[email protected]>
- configs/config_network_ddpm.json +154 -0
- configs/config_network_rflow.json +150 -0
configs/config_network_ddpm.json
ADDED
|
@@ -0,0 +1,154 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"spatial_dims": 3,
|
| 3 |
+
"image_channels": 1,
|
| 4 |
+
"latent_channels": 4,
|
| 5 |
+
"include_body_region": true,
|
| 6 |
+
"mask_generation_latent_shape": [
|
| 7 |
+
4,
|
| 8 |
+
64,
|
| 9 |
+
64,
|
| 10 |
+
64
|
| 11 |
+
],
|
| 12 |
+
"autoencoder_def": {
|
| 13 |
+
"_target_": "monai.apps.generation.maisi.networks.autoencoderkl_maisi.AutoencoderKlMaisi",
|
| 14 |
+
"spatial_dims": "@spatial_dims",
|
| 15 |
+
"in_channels": "@image_channels",
|
| 16 |
+
"out_channels": "@image_channels",
|
| 17 |
+
"latent_channels": "@latent_channels",
|
| 18 |
+
"num_channels": [
|
| 19 |
+
64,
|
| 20 |
+
128,
|
| 21 |
+
256
|
| 22 |
+
],
|
| 23 |
+
"num_res_blocks": [2,2,2],
|
| 24 |
+
"norm_num_groups": 32,
|
| 25 |
+
"norm_eps": 1e-06,
|
| 26 |
+
"attention_levels": [
|
| 27 |
+
false,
|
| 28 |
+
false,
|
| 29 |
+
false
|
| 30 |
+
],
|
| 31 |
+
"with_encoder_nonlocal_attn": false,
|
| 32 |
+
"with_decoder_nonlocal_attn": false,
|
| 33 |
+
"use_checkpointing": false,
|
| 34 |
+
"use_convtranspose": false,
|
| 35 |
+
"norm_float16": true,
|
| 36 |
+
"num_splits": 8,
|
| 37 |
+
"dim_split": 1
|
| 38 |
+
},
|
| 39 |
+
"diffusion_unet_def": {
|
| 40 |
+
"_target_": "monai.apps.generation.maisi.networks.diffusion_model_unet_maisi.DiffusionModelUNetMaisi",
|
| 41 |
+
"spatial_dims": "@spatial_dims",
|
| 42 |
+
"in_channels": "@latent_channels",
|
| 43 |
+
"out_channels": "@latent_channels",
|
| 44 |
+
"num_channels": [
|
| 45 |
+
64,
|
| 46 |
+
128,
|
| 47 |
+
256,
|
| 48 |
+
512
|
| 49 |
+
],
|
| 50 |
+
"attention_levels": [
|
| 51 |
+
false,
|
| 52 |
+
false,
|
| 53 |
+
true,
|
| 54 |
+
true
|
| 55 |
+
],
|
| 56 |
+
"num_head_channels": [
|
| 57 |
+
0,
|
| 58 |
+
0,
|
| 59 |
+
32,
|
| 60 |
+
32
|
| 61 |
+
],
|
| 62 |
+
"num_res_blocks": 2,
|
| 63 |
+
"use_flash_attention": true,
|
| 64 |
+
"include_top_region_index_input": "@include_body_region",
|
| 65 |
+
"include_bottom_region_index_input": "@include_body_region",
|
| 66 |
+
"include_spacing_input": true
|
| 67 |
+
},
|
| 68 |
+
"controlnet_def": {
|
| 69 |
+
"_target_": "monai.apps.generation.maisi.networks.controlnet_maisi.ControlNetMaisi",
|
| 70 |
+
"spatial_dims": "@spatial_dims",
|
| 71 |
+
"in_channels": "@latent_channels",
|
| 72 |
+
"num_channels": [
|
| 73 |
+
64,
|
| 74 |
+
128,
|
| 75 |
+
256,
|
| 76 |
+
512
|
| 77 |
+
],
|
| 78 |
+
"attention_levels": [
|
| 79 |
+
false,
|
| 80 |
+
false,
|
| 81 |
+
true,
|
| 82 |
+
true
|
| 83 |
+
],
|
| 84 |
+
"num_head_channels": [
|
| 85 |
+
0,
|
| 86 |
+
0,
|
| 87 |
+
32,
|
| 88 |
+
32
|
| 89 |
+
],
|
| 90 |
+
"num_res_blocks": 2,
|
| 91 |
+
"use_flash_attention": true,
|
| 92 |
+
"conditioning_embedding_in_channels": 8,
|
| 93 |
+
"conditioning_embedding_num_channels": [8, 32, 64]
|
| 94 |
+
},
|
| 95 |
+
"mask_generation_autoencoder_def": {
|
| 96 |
+
"_target_": "monai.apps.generation.maisi.networks.autoencoderkl_maisi.AutoencoderKlMaisi",
|
| 97 |
+
"spatial_dims": "@spatial_dims",
|
| 98 |
+
"in_channels": 8,
|
| 99 |
+
"out_channels": 125,
|
| 100 |
+
"latent_channels": "@latent_channels",
|
| 101 |
+
"num_channels": [
|
| 102 |
+
32,
|
| 103 |
+
64,
|
| 104 |
+
128
|
| 105 |
+
],
|
| 106 |
+
"num_res_blocks": [1, 2, 2],
|
| 107 |
+
"norm_num_groups": 32,
|
| 108 |
+
"norm_eps": 1e-06,
|
| 109 |
+
"attention_levels": [
|
| 110 |
+
false,
|
| 111 |
+
false,
|
| 112 |
+
false
|
| 113 |
+
],
|
| 114 |
+
"with_encoder_nonlocal_attn": false,
|
| 115 |
+
"with_decoder_nonlocal_attn": false,
|
| 116 |
+
"use_flash_attention": false,
|
| 117 |
+
"use_checkpointing": true,
|
| 118 |
+
"use_convtranspose": true,
|
| 119 |
+
"norm_float16": true,
|
| 120 |
+
"num_splits": 8,
|
| 121 |
+
"dim_split": 1
|
| 122 |
+
},
|
| 123 |
+
"mask_generation_diffusion_def": {
|
| 124 |
+
"_target_": "monai.networks.nets.diffusion_model_unet.DiffusionModelUNet",
|
| 125 |
+
"spatial_dims": "@spatial_dims",
|
| 126 |
+
"in_channels": "@latent_channels",
|
| 127 |
+
"out_channels": "@latent_channels",
|
| 128 |
+
"channels":[64, 128, 256, 512],
|
| 129 |
+
"attention_levels":[false, false, true, true],
|
| 130 |
+
"num_head_channels":[0, 0, 32, 32],
|
| 131 |
+
"num_res_blocks": 2,
|
| 132 |
+
"use_flash_attention": true,
|
| 133 |
+
"with_conditioning": true,
|
| 134 |
+
"upcast_attention": true,
|
| 135 |
+
"cross_attention_dim": 10
|
| 136 |
+
},
|
| 137 |
+
"mask_generation_scale_factor": 1.0055984258651733,
|
| 138 |
+
"noise_scheduler": {
|
| 139 |
+
"_target_": "monai.networks.schedulers.ddpm.DDPMScheduler",
|
| 140 |
+
"num_train_timesteps": 1000,
|
| 141 |
+
"beta_start": 0.0015,
|
| 142 |
+
"beta_end": 0.0195,
|
| 143 |
+
"schedule": "scaled_linear_beta",
|
| 144 |
+
"clip_sample": false
|
| 145 |
+
},
|
| 146 |
+
"mask_generation_noise_scheduler": {
|
| 147 |
+
"_target_": "monai.networks.schedulers.ddpm.DDPMScheduler",
|
| 148 |
+
"num_train_timesteps": 1000,
|
| 149 |
+
"beta_start": 0.0015,
|
| 150 |
+
"beta_end": 0.0195,
|
| 151 |
+
"schedule": "scaled_linear_beta",
|
| 152 |
+
"clip_sample": false
|
| 153 |
+
}
|
| 154 |
+
}
|
configs/config_network_rflow.json
ADDED
|
@@ -0,0 +1,150 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"spatial_dims": 3,
|
| 3 |
+
"image_channels": 1,
|
| 4 |
+
"latent_channels": 4,
|
| 5 |
+
"include_body_region": false,
|
| 6 |
+
"mask_generation_latent_shape": [
|
| 7 |
+
4,
|
| 8 |
+
64,
|
| 9 |
+
64,
|
| 10 |
+
64
|
| 11 |
+
],
|
| 12 |
+
"autoencoder_def": {
|
| 13 |
+
"_target_": "monai.apps.generation.maisi.networks.autoencoderkl_maisi.AutoencoderKlMaisi",
|
| 14 |
+
"spatial_dims": "@spatial_dims",
|
| 15 |
+
"in_channels": "@image_channels",
|
| 16 |
+
"out_channels": "@image_channels",
|
| 17 |
+
"latent_channels": "@latent_channels",
|
| 18 |
+
"num_channels": [
|
| 19 |
+
64,
|
| 20 |
+
128,
|
| 21 |
+
256
|
| 22 |
+
],
|
| 23 |
+
"num_res_blocks": [2,2,2],
|
| 24 |
+
"norm_num_groups": 32,
|
| 25 |
+
"norm_eps": 1e-06,
|
| 26 |
+
"attention_levels": [
|
| 27 |
+
false,
|
| 28 |
+
false,
|
| 29 |
+
false
|
| 30 |
+
],
|
| 31 |
+
"with_encoder_nonlocal_attn": false,
|
| 32 |
+
"with_decoder_nonlocal_attn": false,
|
| 33 |
+
"use_checkpointing": false,
|
| 34 |
+
"use_convtranspose": false,
|
| 35 |
+
"norm_float16": true,
|
| 36 |
+
"num_splits": 4,
|
| 37 |
+
"dim_split": 1
|
| 38 |
+
},
|
| 39 |
+
"diffusion_unet_def": {
|
| 40 |
+
"_target_": "monai.apps.generation.maisi.networks.diffusion_model_unet_maisi.DiffusionModelUNetMaisi",
|
| 41 |
+
"spatial_dims": "@spatial_dims",
|
| 42 |
+
"in_channels": "@latent_channels",
|
| 43 |
+
"out_channels": "@latent_channels",
|
| 44 |
+
"num_channels": [64, 128, 256, 512],
|
| 45 |
+
"attention_levels": [
|
| 46 |
+
false,
|
| 47 |
+
false,
|
| 48 |
+
true,
|
| 49 |
+
true
|
| 50 |
+
],
|
| 51 |
+
"num_head_channels": [
|
| 52 |
+
0,
|
| 53 |
+
0,
|
| 54 |
+
32,
|
| 55 |
+
32
|
| 56 |
+
],
|
| 57 |
+
"num_res_blocks": 2,
|
| 58 |
+
"use_flash_attention": true,
|
| 59 |
+
"include_top_region_index_input": "@include_body_region",
|
| 60 |
+
"include_bottom_region_index_input": "@include_body_region",
|
| 61 |
+
"include_spacing_input": true,
|
| 62 |
+
"num_class_embeds": 128,
|
| 63 |
+
"resblock_updown": true,
|
| 64 |
+
"include_fc": true
|
| 65 |
+
},
|
| 66 |
+
"controlnet_def": {
|
| 67 |
+
"_target_": "monai.apps.generation.maisi.networks.controlnet_maisi.ControlNetMaisi",
|
| 68 |
+
"spatial_dims": "@spatial_dims",
|
| 69 |
+
"in_channels": "@latent_channels",
|
| 70 |
+
"num_channels": [64, 128, 256, 512],
|
| 71 |
+
"attention_levels": [
|
| 72 |
+
false,
|
| 73 |
+
false,
|
| 74 |
+
true,
|
| 75 |
+
true
|
| 76 |
+
],
|
| 77 |
+
"num_head_channels": [
|
| 78 |
+
0,
|
| 79 |
+
0,
|
| 80 |
+
32,
|
| 81 |
+
32
|
| 82 |
+
],
|
| 83 |
+
"num_res_blocks": 2,
|
| 84 |
+
"use_flash_attention": true,
|
| 85 |
+
"conditioning_embedding_in_channels": 8,
|
| 86 |
+
"conditioning_embedding_num_channels": [8, 32, 64],
|
| 87 |
+
"num_class_embeds": 128,
|
| 88 |
+
"resblock_updown": true,
|
| 89 |
+
"include_fc": true
|
| 90 |
+
},
|
| 91 |
+
"mask_generation_autoencoder_def": {
|
| 92 |
+
"_target_": "monai.apps.generation.maisi.networks.autoencoderkl_maisi.AutoencoderKlMaisi",
|
| 93 |
+
"spatial_dims": "@spatial_dims",
|
| 94 |
+
"in_channels": 8,
|
| 95 |
+
"out_channels": 125,
|
| 96 |
+
"latent_channels": "@latent_channels",
|
| 97 |
+
"num_channels": [
|
| 98 |
+
32,
|
| 99 |
+
64,
|
| 100 |
+
128
|
| 101 |
+
],
|
| 102 |
+
"num_res_blocks": [1, 2, 2],
|
| 103 |
+
"norm_num_groups": 32,
|
| 104 |
+
"norm_eps": 1e-06,
|
| 105 |
+
"attention_levels": [
|
| 106 |
+
false,
|
| 107 |
+
false,
|
| 108 |
+
false
|
| 109 |
+
],
|
| 110 |
+
"with_encoder_nonlocal_attn": false,
|
| 111 |
+
"with_decoder_nonlocal_attn": false,
|
| 112 |
+
"use_flash_attention": false,
|
| 113 |
+
"use_checkpointing": true,
|
| 114 |
+
"use_convtranspose": true,
|
| 115 |
+
"norm_float16": true,
|
| 116 |
+
"num_splits": 8,
|
| 117 |
+
"dim_split": 1
|
| 118 |
+
},
|
| 119 |
+
"mask_generation_diffusion_def": {
|
| 120 |
+
"_target_": "monai.networks.nets.diffusion_model_unet.DiffusionModelUNet",
|
| 121 |
+
"spatial_dims": "@spatial_dims",
|
| 122 |
+
"in_channels": "@latent_channels",
|
| 123 |
+
"out_channels": "@latent_channels",
|
| 124 |
+
"channels":[64, 128, 256, 512],
|
| 125 |
+
"attention_levels":[false, false, true, true],
|
| 126 |
+
"num_head_channels":[0, 0, 32, 32],
|
| 127 |
+
"num_res_blocks": 2,
|
| 128 |
+
"use_flash_attention": true,
|
| 129 |
+
"with_conditioning": true,
|
| 130 |
+
"upcast_attention": true,
|
| 131 |
+
"cross_attention_dim": 10
|
| 132 |
+
},
|
| 133 |
+
"mask_generation_scale_factor": 1.0055984258651733,
|
| 134 |
+
"noise_scheduler": {
|
| 135 |
+
"_target_": "monai.networks.schedulers.rectified_flow.RFlowScheduler",
|
| 136 |
+
"num_train_timesteps": 1000,
|
| 137 |
+
"use_discrete_timesteps": false,
|
| 138 |
+
"use_timestep_transform": true,
|
| 139 |
+
"sample_method": "uniform",
|
| 140 |
+
"scale":1.4
|
| 141 |
+
},
|
| 142 |
+
"mask_generation_noise_scheduler": {
|
| 143 |
+
"_target_": "monai.networks.schedulers.ddpm.DDPMScheduler",
|
| 144 |
+
"num_train_timesteps": 1000,
|
| 145 |
+
"beta_start": 0.0015,
|
| 146 |
+
"beta_end": 0.0195,
|
| 147 |
+
"schedule": "scaled_linear_beta",
|
| 148 |
+
"clip_sample": false
|
| 149 |
+
}
|
| 150 |
+
}
|