Hugging Face's logo

shuohsuan
/
svla_1m_5

Robotics
LeRobot
Safetensors
smolvla
Model card Files
xet
Community
shuohsuan commited on
Commit
05771d9
·
verified ·
1 Parent(s): e7e9ef0

Upload folder using huggingface_hub

Browse files
Files changed (3) hide show
  1. config.json +88 -0
  2. model.safetensors +3 -0
  3. train_config.json +207 -0
config.json ADDED
@@ -0,0 +1,88 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "smolvla",
3
+ "n_obs_steps": 1,
4
+ "normalization_mapping": {
5
+ "VISUAL": "IDENTITY",
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.side": {
17
+ "type": "VISUAL",
18
+ "shape": [
19
+ 3,
20
+ 480,
21
+ 640
22
+ ]
23
+ },
24
+ "observation.images.wrist": {
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
+ "push_to_hub": true,
44
+ "repo_id": "shuohsuan/svla_1m_5",
45
+ "private": null,
46
+ "tags": null,
47
+ "license": null,
48
+ "chunk_size": 50,
49
+ "n_action_steps": 50,
50
+ "max_state_dim": 32,
51
+ "max_action_dim": 32,
52
+ "resize_imgs_with_padding": [
53
+ 512,
54
+ 512
55
+ ],
56
+ "empty_cameras": 0,
57
+ "adapt_to_pi_aloha": false,
58
+ "use_delta_joint_actions_aloha": false,
59
+ "tokenizer_max_length": 48,
60
+ "num_steps": 10,
61
+ "use_cache": true,
62
+ "freeze_vision_encoder": true,
63
+ "train_expert_only": true,
64
+ "train_state_proj": true,
65
+ "optimizer_lr": 0.0001,
66
+ "optimizer_betas": [
67
+ 0.9,
68
+ 0.95
69
+ ],
70
+ "optimizer_eps": 1e-08,
71
+ "optimizer_weight_decay": 1e-10,
72
+ "optimizer_grad_clip_norm": 10.0,
73
+ "scheduler_warmup_steps": 1000,
74
+ "scheduler_decay_steps": 30000,
75
+ "scheduler_decay_lr": 2.5e-06,
76
+ "vlm_model_name": "HuggingFaceTB/SmolVLM2-500M-Video-Instruct",
77
+ "load_vlm_weights": true,
78
+ "add_image_special_tokens": false,
79
+ "attention_mode": "cross_attn",
80
+ "prefix_length": 0,
81
+ "pad_language_to": "max_length",
82
+ "num_expert_layers": 0,
83
+ "num_vlm_layers": 16,
84
+ "self_attn_every_n_layers": 2,
85
+ "expert_width_multiplier": 0.75,
86
+ "min_period": 0.004,
87
+ "max_period": 4.0
88
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fdcf4e73eaa90b139f79886193177b24092723b195a6910213e2be7392d9af64
3
+ size 906713296
train_config.json ADDED
@@ -0,0 +1,207 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": [
4
+ "shuohsuan/data1m",
5
+ "shuohsuan/data2",
6
+ "shuohsuan/data3",
7
+ "shuohsuan/data4",
8
+ "shuohsuan/data5"
9
+ ],
10
+ "repo_idx": 0,
11
+ "root": null,
12
+ "episodes": null,
13
+ "image_transforms": {
14
+ "enable": false,
15
+ "max_num_transforms": 3,
16
+ "random_order": false,
17
+ "tfs": {
18
+ "brightness": {
19
+ "weight": 1.0,
20
+ "type": "ColorJitter",
21
+ "kwargs": {
22
+ "brightness": [
23
+ 0.8,
24
+ 1.2
25
+ ]
26
+ }
27
+ },
28
+ "contrast": {
29
+ "weight": 1.0,
30
+ "type": "ColorJitter",
31
+ "kwargs": {
32
+ "contrast": [
33
+ 0.8,
34
+ 1.2
35
+ ]
36
+ }
37
+ },
38
+ "saturation": {
39
+ "weight": 1.0,
40
+ "type": "ColorJitter",
41
+ "kwargs": {
42
+ "saturation": [
43
+ 0.5,
44
+ 1.5
45
+ ]
46
+ }
47
+ },
48
+ "hue": {
49
+ "weight": 1.0,
50
+ "type": "ColorJitter",
51
+ "kwargs": {
52
+ "hue": [
53
+ -0.05,
54
+ 0.05
55
+ ]
56
+ }
57
+ },
58
+ "sharpness": {
59
+ "weight": 1.0,
60
+ "type": "SharpnessJitter",
61
+ "kwargs": {
62
+ "sharpness": [
63
+ 0.5,
64
+ 1.5
65
+ ]
66
+ }
67
+ }
68
+ }
69
+ },
70
+ "revision": null,
71
+ "use_imagenet_stats": true,
72
+ "video_backend": "torchcodec"
73
+ },
74
+ "env": null,
75
+ "policy": {
76
+ "type": "smolvla",
77
+ "n_obs_steps": 1,
78
+ "normalization_mapping": {
79
+ "VISUAL": "IDENTITY",
80
+ "STATE": "MEAN_STD",
81
+ "ACTION": "MEAN_STD"
82
+ },
83
+ "input_features": {
84
+ "observation.state": {
85
+ "type": "STATE",
86
+ "shape": [
87
+ 6
88
+ ]
89
+ },
90
+ "observation.images.side": {
91
+ "type": "VISUAL",
92
+ "shape": [
93
+ 3,
94
+ 480,
95
+ 640
96
+ ]
97
+ },
98
+ "observation.images.wrist": {
99
+ "type": "VISUAL",
100
+ "shape": [
101
+ 3,
102
+ 480,
103
+ 640
104
+ ]
105
+ }
106
+ },
107
+ "output_features": {
108
+ "action": {
109
+ "type": "ACTION",
110
+ "shape": [
111
+ 6
112
+ ]
113
+ }
114
+ },
115
+ "device": "cuda",
116
+ "use_amp": false,
117
+ "push_to_hub": true,
118
+ "repo_id": "shuohsuan/svla_1m_5",
119
+ "private": null,
120
+ "tags": null,
121
+ "license": null,
122
+ "chunk_size": 50,
123
+ "n_action_steps": 50,
124
+ "max_state_dim": 32,
125
+ "max_action_dim": 32,
126
+ "resize_imgs_with_padding": [
127
+ 512,
128
+ 512
129
+ ],
130
+ "empty_cameras": 0,
131
+ "adapt_to_pi_aloha": false,
132
+ "use_delta_joint_actions_aloha": false,
133
+ "tokenizer_max_length": 48,
134
+ "num_steps": 10,
135
+ "use_cache": true,
136
+ "freeze_vision_encoder": true,
137
+ "train_expert_only": true,
138
+ "train_state_proj": true,
139
+ "optimizer_lr": 0.0001,
140
+ "optimizer_betas": [
141
+ 0.9,
142
+ 0.95
143
+ ],
144
+ "optimizer_eps": 1e-08,
145
+ "optimizer_weight_decay": 1e-10,
146
+ "optimizer_grad_clip_norm": 10.0,
147
+ "scheduler_warmup_steps": 1000,
148
+ "scheduler_decay_steps": 30000,
149
+ "scheduler_decay_lr": 2.5e-06,
150
+ "vlm_model_name": "HuggingFaceTB/SmolVLM2-500M-Video-Instruct",
151
+ "load_vlm_weights": true,
152
+ "add_image_special_tokens": false,
153
+ "attention_mode": "cross_attn",
154
+ "prefix_length": 0,
155
+ "pad_language_to": "max_length",
156
+ "num_expert_layers": 0,
157
+ "num_vlm_layers": 16,
158
+ "self_attn_every_n_layers": 2,
159
+ "expert_width_multiplier": 0.75,
160
+ "min_period": 0.004,
161
+ "max_period": 4.0
162
+ },
163
+ "output_dir": "/mnt/storage/shuohsuan/train/lerobot/svla_1m_5",
164
+ "job_name": "base",
165
+ "resume": false,
166
+ "seed": 1000,
167
+ "num_workers": 4,
168
+ "batch_size": 64,
169
+ "steps": 100000,
170
+ "eval_freq": 20000,
171
+ "log_freq": 200,
172
+ "save_checkpoint": true,
173
+ "save_freq": 20000,
174
+ "use_policy_training_preset": true,
175
+ "optimizer": {
176
+ "type": "adamw",
177
+ "lr": 0.0001,
178
+ "weight_decay": 1e-10,
179
+ "grad_clip_norm": 10.0,
180
+ "betas": [
181
+ 0.9,
182
+ 0.95
183
+ ],
184
+ "eps": 1e-08
185
+ },
186
+ "scheduler": {
187
+ "type": "cosine_decay_with_warmup",
188
+ "num_warmup_steps": 1000,
189
+ "num_decay_steps": 30000,
190
+ "peak_lr": 0.0001,
191
+ "decay_lr": 2.5e-06
192
+ },
193
+ "eval": {
194
+ "n_episodes": 50,
195
+ "batch_size": 50,
196
+ "use_async_envs": false
197
+ },
198
+ "wandb": {
199
+ "enable": false,
200
+ "disable_artifact": false,
201
+ "project": "lerobot",
202
+ "entity": null,
203
+ "notes": null,
204
+ "run_id": null,
205
+ "mode": null
206
+ }
207
+ }