Hugging Face's logo

nvidia
/
NV-Generate-MR

Model card Files
xet
Community
Can-Zhao commited on
Commit
007a29c
·
1 Parent(s): b94cb27

upload models

Browse files

Signed-off-by: Can-Zhao <canz@nvidia.com>

Files changed (4) hide show
  1. configs/.ipynb_checkpoints/config_network_rflow-checkpoint.json +150 -0
  2. configs/config_network_rflow.json +150 -0
  3. models/autoencoder_v2.pt +3 -0
  4. models/diff_unet_3d_rflow-mr.pt +3 -0
configs/.ipynb_checkpoints/config_network_rflow-checkpoint.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
+ }
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
+ }
models/autoencoder_v2.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b5ed556dc64872cae11ebe67cc33e84fbd05ebdf7e35e40c74d956404e7c1ef0
3
+ size 83831932
models/diff_unet_3d_rflow-mr.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cbf3ee62ec807ebd63e53fe3cd6fcf8ac9249b427dbeee9e4d205e2931e8ca03
3
+ size 2166664552