hanshanxue commited on
Commit
ac2ad97
·
verified ·
1 Parent(s): 56ac0d4

Upload folder using huggingface_hub

Browse files
worldstereo-memory-dmd/.ipynb_checkpoints/config-checkpoint.json ADDED
@@ -0,0 +1,163 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "base_model": "Wan-AI/Wan2.1-I2V-14B-480P-Diffusers",
3
+ "load_uni3c": false,
4
+ "seed": 1024,
5
+ "nframe": 21,
6
+ "num_workers": 8,
7
+ "scale_map": [
8
+ [
9
+ 480,
10
+ 832
11
+ ],
12
+ [
13
+ 480,
14
+ 768
15
+ ],
16
+ [
17
+ 512,
18
+ 784
19
+ ],
20
+ [
21
+ 512,
22
+ 720
23
+ ],
24
+ [
25
+ 576,
26
+ 704
27
+ ],
28
+ [
29
+ 576,
30
+ 640
31
+ ],
32
+ [
33
+ 640,
34
+ 640
35
+ ],
36
+ [
37
+ 640,
38
+ 576
39
+ ],
40
+ [
41
+ 704,
42
+ 576
43
+ ],
44
+ [
45
+ 720,
46
+ 512
47
+ ],
48
+ [
49
+ 784,
50
+ 512
51
+ ],
52
+ [
53
+ 768,
54
+ 480
55
+ ],
56
+ [
57
+ 832,
58
+ 480
59
+ ]
60
+ ],
61
+ "grad_checkpoint": true,
62
+ "freeze_backbone": false,
63
+ "ema_eval": true,
64
+ "render_radius": 0.008,
65
+ "points_per_pixel": 8,
66
+ "background_color": [
67
+ 0,
68
+ 0,
69
+ 0
70
+ ],
71
+ "sobel_threshold": 0.35,
72
+ "center_scale": 1.0,
73
+ "start_elevation": 5.0,
74
+ "pcd_aug_cfg": {
75
+ "crop_ratio": 0.1,
76
+ "edge_mask_ratio": 0.5,
77
+ "pcd_degrade": true,
78
+ "fly_edge_ratio": 0.1,
79
+ "fly_dege_size": 3,
80
+ "global_distortion_ratio": 0.0,
81
+ "global_distortion_size": 21,
82
+ "local_distortion_ratio": 0.0,
83
+ "local_distortion_size": 27,
84
+ "local_distortion_region": 4
85
+ },
86
+ "ref_aug_cfg": {
87
+ "motion_blur_aug": true,
88
+ "motion_blur_aug_ratio": 0.1,
89
+ "motion_blur_kernel_size": 7,
90
+ "jitter_aug": true,
91
+ "jitter_brightness": 0.15,
92
+ "jitter_contrast": 0.15,
93
+ "jitter_saturation": 0.15
94
+ },
95
+ "camera_embedding": true,
96
+ "is_w2c": true,
97
+ "controlnet_cfg": {
98
+ "conv_out_dim": 5120,
99
+ "time_embed_dim": 5120,
100
+ "dim": 1024,
101
+ "ffn_dim": 4096,
102
+ "num_heads": 16,
103
+ "num_layers": 20,
104
+ "add_channels": 7,
105
+ "mid_channels": 256,
106
+ "mask_downsample": 1,
107
+ "update_ref": true,
108
+ "ref_embedding": false,
109
+ "main_camera_embedding": true,
110
+ "camera_embedding_dim": 7,
111
+ "freeze_attn2": true,
112
+ "freeze_ffn": false
113
+ },
114
+ "dmd_lora_rank": 256,
115
+ "dmd_lora": false,
116
+ "freeze_controlnet": false,
117
+ "dmd_steps": 4,
118
+ "dmd_start_steps": 250,
119
+ "dmd_end_steps": 1000,
120
+ "gen_shift": 5.0,
121
+ "gen_update_ratio": 5,
122
+ "real_guidance_scale": 5.0,
123
+ "negative_prompt": "色调艳丽,过曝,静态,细节模糊不清,字幕,风格,作品,画作,画面,整体发灰,最差质量,低质量,JPEG压缩残留,丑陋的,残缺的,多余的手指,画得不好的手部,画得不好的脸部,畸形的,毁容的,形态畸形的肢体,手指融合,静止不动的画面,杂乱的背景,三条腿,背景人很多,倒着走,阳光,明亮",
124
+ "generator_fsdp": "fsdp_hybrid",
125
+ "real_score_fsdp": "fsdp_hybrid",
126
+ "fake_score_fsdp": "fsdp_hybrid",
127
+ "other_fsdp": "fsdp_hybrid",
128
+ "cpu_offload": false,
129
+ "lr": 2e-06,
130
+ "lr_critic": 4e-07,
131
+ "grad_clip": 1.0,
132
+ "weight_decay": 0.0,
133
+ "adam_eps": 1e-08,
134
+ "beta1": 0.0,
135
+ "beta2": 0.999,
136
+ "lr_scheduler": "constant",
137
+ "resume_epoch": -1,
138
+ "render_dropout": 0.05,
139
+ "camera_dropout": 0.08,
140
+ "pointmap_dropout": 0.0,
141
+ "referece_dropout": 0.0,
142
+ "latent_cond_mode": "first_frame_only",
143
+ "use_ema": true,
144
+ "ema_decay": 0.999,
145
+ "ema_warmup_steps": 20,
146
+ "loss": {
147
+ "num_train_timesteps": 1000,
148
+ "shift": 5.0,
149
+ "use_discrete_timesteps": false,
150
+ "sample_method": "logit-normal",
151
+ "use_timestep_transform": true,
152
+ "loc": 0.0,
153
+ "scale": 1.0,
154
+ "t_min": 0.02,
155
+ "t_max": 0.98
156
+ },
157
+ "total_steps": 3001,
158
+ "warmup_steps": 0,
159
+ "log_every": 1,
160
+ "eval_every": 500,
161
+ "ckpt_every": 250,
162
+ "enable_prope": false
163
+ }
worldstereo-memory-dmd/config.json ADDED
@@ -0,0 +1,163 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "base_model": "Wan-AI/Wan2.1-I2V-14B-480P-Diffusers",
3
+ "load_uni3c": false,
4
+ "seed": 1024,
5
+ "nframe": 21,
6
+ "num_workers": 8,
7
+ "scale_map": [
8
+ [
9
+ 480,
10
+ 832
11
+ ],
12
+ [
13
+ 480,
14
+ 768
15
+ ],
16
+ [
17
+ 512,
18
+ 784
19
+ ],
20
+ [
21
+ 512,
22
+ 720
23
+ ],
24
+ [
25
+ 576,
26
+ 704
27
+ ],
28
+ [
29
+ 576,
30
+ 640
31
+ ],
32
+ [
33
+ 640,
34
+ 640
35
+ ],
36
+ [
37
+ 640,
38
+ 576
39
+ ],
40
+ [
41
+ 704,
42
+ 576
43
+ ],
44
+ [
45
+ 720,
46
+ 512
47
+ ],
48
+ [
49
+ 784,
50
+ 512
51
+ ],
52
+ [
53
+ 768,
54
+ 480
55
+ ],
56
+ [
57
+ 832,
58
+ 480
59
+ ]
60
+ ],
61
+ "grad_checkpoint": true,
62
+ "freeze_backbone": false,
63
+ "ema_eval": true,
64
+ "render_radius": 0.008,
65
+ "points_per_pixel": 8,
66
+ "background_color": [
67
+ 0,
68
+ 0,
69
+ 0
70
+ ],
71
+ "sobel_threshold": 0.35,
72
+ "center_scale": 1.0,
73
+ "start_elevation": 5.0,
74
+ "pcd_aug_cfg": {
75
+ "crop_ratio": 0.1,
76
+ "edge_mask_ratio": 0.5,
77
+ "pcd_degrade": true,
78
+ "fly_edge_ratio": 0.1,
79
+ "fly_dege_size": 3,
80
+ "global_distortion_ratio": 0.0,
81
+ "global_distortion_size": 21,
82
+ "local_distortion_ratio": 0.0,
83
+ "local_distortion_size": 27,
84
+ "local_distortion_region": 4
85
+ },
86
+ "ref_aug_cfg": {
87
+ "motion_blur_aug": true,
88
+ "motion_blur_aug_ratio": 0.1,
89
+ "motion_blur_kernel_size": 7,
90
+ "jitter_aug": true,
91
+ "jitter_brightness": 0.15,
92
+ "jitter_contrast": 0.15,
93
+ "jitter_saturation": 0.15
94
+ },
95
+ "camera_embedding": true,
96
+ "is_w2c": true,
97
+ "controlnet_cfg": {
98
+ "conv_out_dim": 5120,
99
+ "time_embed_dim": 5120,
100
+ "dim": 1024,
101
+ "ffn_dim": 4096,
102
+ "num_heads": 16,
103
+ "num_layers": 20,
104
+ "add_channels": 7,
105
+ "mid_channels": 256,
106
+ "mask_downsample": 1,
107
+ "update_ref": true,
108
+ "ref_embedding": false,
109
+ "main_camera_embedding": true,
110
+ "camera_embedding_dim": 7,
111
+ "freeze_attn2": true,
112
+ "freeze_ffn": false
113
+ },
114
+ "dmd_lora_rank": 256,
115
+ "dmd_lora": false,
116
+ "freeze_controlnet": false,
117
+ "dmd_steps": 4,
118
+ "dmd_start_steps": 250,
119
+ "dmd_end_steps": 1000,
120
+ "gen_shift": 5.0,
121
+ "gen_update_ratio": 5,
122
+ "real_guidance_scale": 5.0,
123
+ "negative_prompt": "色调艳丽,过曝,静态,细节模糊不清,字幕,风格,作品,画作,画面,整体发灰,最差质量,低质量,JPEG压缩残留,丑陋的,残缺的,多余的手指,画得不好的手部,画得不好的脸部,畸形的,毁容的,形态畸形的肢体,手指融合,静止不动的画面,杂乱的背景,三条腿,背景人很多,倒着走,阳光,明亮",
124
+ "generator_fsdp": "fsdp_hybrid",
125
+ "real_score_fsdp": "fsdp_hybrid",
126
+ "fake_score_fsdp": "fsdp_hybrid",
127
+ "other_fsdp": "fsdp_hybrid",
128
+ "cpu_offload": false,
129
+ "lr": 2e-06,
130
+ "lr_critic": 4e-07,
131
+ "grad_clip": 1.0,
132
+ "weight_decay": 0.0,
133
+ "adam_eps": 1e-08,
134
+ "beta1": 0.0,
135
+ "beta2": 0.999,
136
+ "lr_scheduler": "constant",
137
+ "resume_epoch": -1,
138
+ "render_dropout": 0.05,
139
+ "camera_dropout": 0.08,
140
+ "pointmap_dropout": 0.0,
141
+ "referece_dropout": 0.0,
142
+ "latent_cond_mode": "first_frame_only",
143
+ "use_ema": true,
144
+ "ema_decay": 0.999,
145
+ "ema_warmup_steps": 20,
146
+ "loss": {
147
+ "num_train_timesteps": 1000,
148
+ "shift": 5.0,
149
+ "use_discrete_timesteps": false,
150
+ "sample_method": "logit-normal",
151
+ "use_timestep_transform": true,
152
+ "loc": 0.0,
153
+ "scale": 1.0,
154
+ "t_min": 0.02,
155
+ "t_max": 0.98
156
+ },
157
+ "total_steps": 3001,
158
+ "warmup_steps": 0,
159
+ "log_every": 1,
160
+ "eval_every": 500,
161
+ "ckpt_every": 250,
162
+ "enable_prope": false
163
+ }
worldstereo-memory-dmd/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a7f6e8e8dcec9ef81c47fb408d61c78f8c7dc27b4f53f0ee228ce583765d35e1
3
+ size 34863748704