Hugging Face's logo

jchun
/
act_so100_unfold

Safetensors
Model card Files
xet
 Community
jchun commited on
Commit
7ef2f3c
·
verified ·
1 Parent(s): bdffe0e

Upload folder using huggingface_hub

Browse files
Files changed (3) hide show
  1. config.json +64 -0
  2. model.safetensors +3 -0
  3. train_config.json +195 -0
config.json ADDED
@@ -0,0 +1,64 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "act",
3
+ "n_obs_steps": 1,
4
+ "normalization_mapping": {
5
+ "VISUAL": "MEAN_STD",
6
+ "STATE": "MEAN_STD",
7
+ "ACTION": "MEAN_STD"
8
+ },
9
+ "input_features": {
10
+ "observation.state": {
11
+ "type": "STATE",
12
+ "shape": [
13
+ 6
14
+ ]
15
+ },
16
+ "observation.images.main": {
17
+ "type": "VISUAL",
18
+ "shape": [
19
+ 3,
20
+ 480,
21
+ 640
22
+ ]
23
+ },
24
+ "observation.images.webcam": {
25
+ "type": "VISUAL",
26
+ "shape": [
27
+ 3,
28
+ 480,
29
+ 640
30
+ ]
31
+ }
32
+ },
33
+ "output_features": {
34
+ "action": {
35
+ "type": "ACTION",
36
+ "shape": [
37
+ 6
38
+ ]
39
+ }
40
+ },
41
+ "device": "cuda",
42
+ "use_amp": false,
43
+ "chunk_size": 100,
44
+ "n_action_steps": 100,
45
+ "vision_backbone": "resnet18",
46
+ "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
47
+ "replace_final_stride_with_dilation": false,
48
+ "pre_norm": false,
49
+ "dim_model": 512,
50
+ "n_heads": 8,
51
+ "dim_feedforward": 3200,
52
+ "feedforward_activation": "relu",
53
+ "n_encoder_layers": 4,
54
+ "n_decoder_layers": 1,
55
+ "use_vae": true,
56
+ "latent_dim": 32,
57
+ "n_vae_encoder_layers": 4,
58
+ "temporal_ensemble_coeff": null,
59
+ "dropout": 0.1,
60
+ "kl_weight": 10.0,
61
+ "optimizer_lr": 1e-05,
62
+ "optimizer_weight_decay": 0.0001,
63
+ "optimizer_lr_backbone": 1e-05
64
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:786c6a75f4df1855de4f2ee41e3f0920eebc7c2faf088c3a03481df82ff27aba
3
+ size 206701072
train_config.json ADDED
@@ -0,0 +1,195 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": "jchun/so100_rm_unfold20250719_180301",
4
+ "root": null,
5
+ "episodes": null,
6
+ "image_transforms": {
7
+ "enable": false,
8
+ "max_num_transforms": 3,
9
+ "random_order": false,
10
+ "tfs": {
11
+ "crop_scale": {
12
+ "weight": 1.0,
13
+ "type": "RandomResizedCrop",
14
+ "kwargs": {
15
+ "size": [
16
+ 224,
17
+ 224
18
+ ],
19
+ "scale": [
20
+ 0.8,
21
+ 1.0
22
+ ],
23
+ "ratio": [
24
+ 0.75,
25
+ 1.33
26
+ ]
27
+ }
28
+ },
29
+ "rotate": {
30
+ "weight": 1.0,
31
+ "type": "RandomRotation",
32
+ "kwargs": {
33
+ "degrees": 30
34
+ }
35
+ },
36
+ "brightness": {
37
+ "weight": 1.0,
38
+ "type": "ColorJitter",
39
+ "kwargs": {
40
+ "brightness": [
41
+ 0.8,
42
+ 1.2
43
+ ]
44
+ }
45
+ },
46
+ "contrast": {
47
+ "weight": 1.0,
48
+ "type": "ColorJitter",
49
+ "kwargs": {
50
+ "contrast": [
51
+ 0.8,
52
+ 1.2
53
+ ]
54
+ }
55
+ },
56
+ "saturation": {
57
+ "weight": 1.0,
58
+ "type": "ColorJitter",
59
+ "kwargs": {
60
+ "saturation": [
61
+ 0.5,
62
+ 1.5
63
+ ]
64
+ }
65
+ },
66
+ "hue": {
67
+ "weight": 1.0,
68
+ "type": "ColorJitter",
69
+ "kwargs": {
70
+ "hue": [
71
+ -0.05,
72
+ 0.05
73
+ ]
74
+ }
75
+ },
76
+ "sharpness": {
77
+ "weight": 1.0,
78
+ "type": "SharpnessJitter",
79
+ "kwargs": {
80
+ "sharpness": [
81
+ 0.5,
82
+ 1.5
83
+ ]
84
+ }
85
+ }
86
+ }
87
+ },
88
+ "revision": null,
89
+ "use_imagenet_stats": true,
90
+ "video_backend": "torchcodec"
91
+ },
92
+ "env": null,
93
+ "policy": {
94
+ "type": "act",
95
+ "n_obs_steps": 1,
96
+ "normalization_mapping": {
97
+ "VISUAL": "MEAN_STD",
98
+ "STATE": "MEAN_STD",
99
+ "ACTION": "MEAN_STD"
100
+ },
101
+ "input_features": {
102
+ "observation.state": {
103
+ "type": "STATE",
104
+ "shape": [
105
+ 6
106
+ ]
107
+ },
108
+ "observation.images.main": {
109
+ "type": "VISUAL",
110
+ "shape": [
111
+ 3,
112
+ 480,
113
+ 640
114
+ ]
115
+ },
116
+ "observation.images.webcam": {
117
+ "type": "VISUAL",
118
+ "shape": [
119
+ 3,
120
+ 480,
121
+ 640
122
+ ]
123
+ }
124
+ },
125
+ "output_features": {
126
+ "action": {
127
+ "type": "ACTION",
128
+ "shape": [
129
+ 6
130
+ ]
131
+ }
132
+ },
133
+ "device": "cuda",
134
+ "use_amp": false,
135
+ "chunk_size": 100,
136
+ "n_action_steps": 100,
137
+ "vision_backbone": "resnet18",
138
+ "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
139
+ "replace_final_stride_with_dilation": false,
140
+ "pre_norm": false,
141
+ "dim_model": 512,
142
+ "n_heads": 8,
143
+ "dim_feedforward": 3200,
144
+ "feedforward_activation": "relu",
145
+ "n_encoder_layers": 4,
146
+ "n_decoder_layers": 1,
147
+ "use_vae": true,
148
+ "latent_dim": 32,
149
+ "n_vae_encoder_layers": 4,
150
+ "temporal_ensemble_coeff": null,
151
+ "dropout": 0.1,
152
+ "kl_weight": 10.0,
153
+ "optimizer_lr": 1e-05,
154
+ "optimizer_weight_decay": 0.0001,
155
+ "optimizer_lr_backbone": 1e-05
156
+ },
157
+ "output_dir": "outputs/train/act_so100_unfold",
158
+ "job_name": "act_so100_unfold",
159
+ "resume": false,
160
+ "seed": 1000,
161
+ "num_workers": 4,
162
+ "batch_size": 8,
163
+ "steps": 100000,
164
+ "eval_freq": 20000,
165
+ "log_freq": 200,
166
+ "save_checkpoint": true,
167
+ "save_freq": 10000,
168
+ "use_policy_training_preset": true,
169
+ "optimizer": {
170
+ "type": "adamw",
171
+ "lr": 1e-05,
172
+ "weight_decay": 0.0001,
173
+ "grad_clip_norm": 10.0,
174
+ "betas": [
175
+ 0.9,
176
+ 0.999
177
+ ],
178
+ "eps": 1e-08
179
+ },
180
+ "scheduler": null,
181
+ "eval": {
182
+ "n_episodes": 50,
183
+ "batch_size": 50,
184
+ "use_async_envs": false
185
+ },
186
+ "wandb": {
187
+ "enable": true,
188
+ "disable_artifact": false,
189
+ "project": "lerobot",
190
+ "entity": null,
191
+ "notes": null,
192
+ "run_id": null,
193
+ "mode": null
194
+ }
195
+ }