hanshanxue commited on
Commit
34c6c1f
·
verified ·
1 Parent(s): cd2690b

Upload folder using huggingface_hub

Browse files
worldstereo-camera/.ipynb_checkpoints/config-checkpoint.json ADDED
@@ -0,0 +1,113 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ "freeze_attn2": true,
64
+ "freeze_ffn": true,
65
+ "latent_cond_mode": "first_frame_only",
66
+ "render_radius": 0.008,
67
+ "points_per_pixel": 8,
68
+ "background_color": [
69
+ 0,
70
+ 0,
71
+ 0
72
+ ],
73
+ "sobel_threshold": 0.35,
74
+ "camera_embedding": true,
75
+ "is_w2c": true,
76
+ "dense_render": false,
77
+ "controlnet_cfg": {
78
+ "conv_out_dim": 5120,
79
+ "time_embed_dim": 5120,
80
+ "dim": 1024,
81
+ "ffn_dim": 4096,
82
+ "num_heads": 16,
83
+ "num_layers": 20,
84
+ "add_channels": 7,
85
+ "mid_channels": 256,
86
+ "mask_downsample": 1
87
+ },
88
+ "lr": 1e-05,
89
+ "grad_clip": 1.0,
90
+ "weight_decay": 0.0001,
91
+ "adam_eps": 1e-08,
92
+ "lr_scheduler": "constant",
93
+ "use_ema": true,
94
+ "ema_decay": 0.999,
95
+ "ema_warmup_steps": 20,
96
+ "render_dropout": 0.05,
97
+ "camera_dropout": 0.08,
98
+ "loss": {
99
+ "num_train_timesteps": 1000,
100
+ "shift": 5.0,
101
+ "use_discrete_timesteps": false,
102
+ "sample_method": "logit-normal",
103
+ "use_timestep_transform": true,
104
+ "loc": 0.0,
105
+ "scale": 1.0
106
+ },
107
+ "total_steps": 8001,
108
+ "warmup_steps": 400,
109
+ "log_every": 10,
110
+ "eval_every": 500,
111
+ "ckpt_every": 500,
112
+ "enable_prope": false
113
+ }
worldstereo-camera/config.json ADDED
@@ -0,0 +1,113 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ "freeze_attn2": true,
64
+ "freeze_ffn": true,
65
+ "latent_cond_mode": "first_frame_only",
66
+ "render_radius": 0.008,
67
+ "points_per_pixel": 8,
68
+ "background_color": [
69
+ 0,
70
+ 0,
71
+ 0
72
+ ],
73
+ "sobel_threshold": 0.35,
74
+ "camera_embedding": true,
75
+ "is_w2c": true,
76
+ "dense_render": false,
77
+ "controlnet_cfg": {
78
+ "conv_out_dim": 5120,
79
+ "time_embed_dim": 5120,
80
+ "dim": 1024,
81
+ "ffn_dim": 4096,
82
+ "num_heads": 16,
83
+ "num_layers": 20,
84
+ "add_channels": 7,
85
+ "mid_channels": 256,
86
+ "mask_downsample": 1
87
+ },
88
+ "lr": 1e-05,
89
+ "grad_clip": 1.0,
90
+ "weight_decay": 0.0001,
91
+ "adam_eps": 1e-08,
92
+ "lr_scheduler": "constant",
93
+ "use_ema": true,
94
+ "ema_decay": 0.999,
95
+ "ema_warmup_steps": 20,
96
+ "render_dropout": 0.05,
97
+ "camera_dropout": 0.08,
98
+ "loss": {
99
+ "num_train_timesteps": 1000,
100
+ "shift": 5.0,
101
+ "use_discrete_timesteps": false,
102
+ "sample_method": "logit-normal",
103
+ "use_timestep_transform": true,
104
+ "loc": 0.0,
105
+ "scale": 1.0
106
+ },
107
+ "total_steps": 8001,
108
+ "warmup_steps": 400,
109
+ "log_every": 10,
110
+ "eval_every": 500,
111
+ "ckpt_every": 500,
112
+ "enable_prope": false
113
+ }
worldstereo-camera/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d6884dab396c32771d040a339a113e5da26df645d6834198467e13a420cd4152
3
+ size 10868938144