Hugging Face's logo

TensorAuto
/
moka_pot_libero_sft

Robotics
Safetensors
PyTorch
opentau
vla
Reinforcemnet Learning
pi0
libero
manipulation
flow-matching
Model card Files
xet
 Community
autox-lerobot-ci commited on
Commit
0a02499
·
1 Parent(s): 402ff6b

Uploading checkpoints

Browse files
Files changed (3) hide show
  1. config.json +83 -0
  2. model.safetensors +3 -0
  3. train_config.json +334 -0
config.json ADDED
@@ -0,0 +1,83 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "pi0",
3
+ "n_obs_steps": 1,
4
+ "normalization_mapping": {
5
+ "VISUAL": "IDENTITY",
6
+ "STATE": "MEAN_STD",
7
+ "ACTION": "MEAN_STD"
8
+ },
9
+ "input_features": {
10
+ "state": {
11
+ "type": "STATE",
12
+ "shape": [
13
+ 32
14
+ ]
15
+ },
16
+ "camera0": {
17
+ "type": "VISUAL",
18
+ "shape": [
19
+ 3,
20
+ 224,
21
+ 224
22
+ ]
23
+ },
24
+ "camera1": {
25
+ "type": "VISUAL",
26
+ "shape": [
27
+ 3,
28
+ 224,
29
+ 224
30
+ ]
31
+ }
32
+ },
33
+ "output_features": {
34
+ "actions": {
35
+ "type": "ACTION",
36
+ "shape": [
37
+ 32
38
+ ]
39
+ }
40
+ },
41
+ "device": null,
42
+ "use_amp": false,
43
+ "cloud_vlm_latency_mean": 0.0,
44
+ "cloud_vlm_latency_std": 0.0,
45
+ "cloud_vlm_latency_lower": 0.0,
46
+ "cloud_vlm_latency_upper": 0.0,
47
+ "action_decoder_latency_mean": 0.0,
48
+ "action_decoder_latency_std": 0.0,
49
+ "action_decoder_latency_lower": 0.0,
50
+ "action_decoder_latency_upper": 0.0,
51
+ "chunk_size": 10,
52
+ "n_action_steps": 10,
53
+ "safety_buffer": 0,
54
+ "max_state_dim": 32,
55
+ "max_action_dim": 32,
56
+ "resize_imgs_with_padding": [
57
+ 224,
58
+ 224
59
+ ],
60
+ "empty_cameras": 0,
61
+ "tokenizer_max_length": 100,
62
+ "proj_width": 1024,
63
+ "dropout": 0.1,
64
+ "num_steps": 10,
65
+ "advantage_threshold": 0.0,
66
+ "advantage": "on",
67
+ "init_strategy": "no_init",
68
+ "use_cache": true,
69
+ "attention_implementation": "eager",
70
+ "freeze_vision_encoder": false,
71
+ "train_expert_only": false,
72
+ "train_state_proj": true,
73
+ "optimizer_lr": 2.5e-05,
74
+ "optimizer_betas": [
75
+ 0.9,
76
+ 0.95
77
+ ],
78
+ "optimizer_eps": 1e-08,
79
+ "optimizer_weight_decay": 1e-10,
80
+ "scheduler_warmup_steps": 1000,
81
+ "scheduler_decay_steps": 30000,
82
+ "scheduler_decay_lr": 2.5e-06
83
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:94eeda515028a1640551dcf53f5546bcaa58dac7093ce1153a80c2325bbdac02
3
+ size 16112213576
train_config.json ADDED
@@ -0,0 +1,334 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset_mixture": {
3
+ "datasets": [
4
+ {
5
+ "repo_id": "physical-intelligence/libero",
6
+ "grounding": null,
7
+ "root": null,
8
+ "episodes": [
9
+ 10,
10
+ 20,
11
+ 23,
12
+ 46,
13
+ 51,
14
+ 54,
15
+ 57,
16
+ 67,
17
+ 70,
18
+ 73,
19
+ 86,
20
+ 100,
21
+ 106,
22
+ 115,
23
+ 143,
24
+ 149,
25
+ 179,
26
+ 187,
27
+ 194,
28
+ 204,
29
+ 214,
30
+ 270,
31
+ 283,
32
+ 288,
33
+ 306,
34
+ 314,
35
+ 316,
36
+ 341,
37
+ 376
38
+ ],
39
+ "image_transforms": {
40
+ "enable": false,
41
+ "max_num_transforms": 3,
42
+ "random_order": false,
43
+ "tfs": {
44
+ "brightness": {
45
+ "weight": 1.0,
46
+ "type": "ColorJitter",
47
+ "kwargs": {
48
+ "brightness": [
49
+ 0.8,
50
+ 1.2
51
+ ]
52
+ }
53
+ },
54
+ "contrast": {
55
+ "weight": 1.0,
56
+ "type": "ColorJitter",
57
+ "kwargs": {
58
+ "contrast": [
59
+ 0.8,
60
+ 1.2
61
+ ]
62
+ }
63
+ },
64
+ "saturation": {
65
+ "weight": 1.0,
66
+ "type": "ColorJitter",
67
+ "kwargs": {
68
+ "saturation": [
69
+ 0.5,
70
+ 1.5
71
+ ]
72
+ }
73
+ },
74
+ "hue": {
75
+ "weight": 1.0,
76
+ "type": "ColorJitter",
77
+ "kwargs": {
78
+ "hue": [
79
+ -0.05,
80
+ 0.05
81
+ ]
82
+ }
83
+ },
84
+ "sharpness": {
85
+ "weight": 1.0,
86
+ "type": "SharpnessJitter",
87
+ "kwargs": {
88
+ "sharpness": [
89
+ 0.5,
90
+ 1.5
91
+ ]
92
+ }
93
+ }
94
+ }
95
+ },
96
+ "revision": null,
97
+ "use_imagenet_stats": true,
98
+ "video_backend": "pyav",
99
+ "stats": null,
100
+ "data_features_name_mapping": null,
101
+ "val_split_ratio": 0.05
102
+ }
103
+ ],
104
+ "weights": [
105
+ 1.0
106
+ ],
107
+ "action_freq": 10.0,
108
+ "image_resample_strategy": "nearest",
109
+ "vector_resample_strategy": "nearest",
110
+ "val_split_ratio": 0.05
111
+ },
112
+ "policy": {
113
+ "type": "pi0",
114
+ "n_obs_steps": 1,
115
+ "normalization_mapping": {
116
+ "VISUAL": "IDENTITY",
117
+ "STATE": "MEAN_STD",
118
+ "ACTION": "MEAN_STD"
119
+ },
120
+ "input_features": {
121
+ "state": {
122
+ "type": "STATE",
123
+ "shape": [
124
+ 32
125
+ ]
126
+ },
127
+ "camera0": {
128
+ "type": "VISUAL",
129
+ "shape": [
130
+ 3,
131
+ 224,
132
+ 224
133
+ ]
134
+ },
135
+ "camera1": {
136
+ "type": "VISUAL",
137
+ "shape": [
138
+ 3,
139
+ 224,
140
+ 224
141
+ ]
142
+ }
143
+ },
144
+ "output_features": {
145
+ "actions": {
146
+ "type": "ACTION",
147
+ "shape": [
148
+ 32
149
+ ]
150
+ }
151
+ },
152
+ "device": null,
153
+ "use_amp": false,
154
+ "pretrained_path": "outputs/train/2026-02-27/03-29-24_pi0/checkpoints/009000",
155
+ "cloud_vlm_latency_mean": 0.0,
156
+ "cloud_vlm_latency_std": 0.0,
157
+ "cloud_vlm_latency_lower": 0.0,
158
+ "cloud_vlm_latency_upper": 0.0,
159
+ "action_decoder_latency_mean": 0.0,
160
+ "action_decoder_latency_std": 0.0,
161
+ "action_decoder_latency_lower": 0.0,
162
+ "action_decoder_latency_upper": 0.0,
163
+ "chunk_size": 10,
164
+ "n_action_steps": 10,
165
+ "safety_buffer": 0,
166
+ "max_state_dim": 32,
167
+ "max_action_dim": 32,
168
+ "resize_imgs_with_padding": [
169
+ 224,
170
+ 224
171
+ ],
172
+ "empty_cameras": 0,
173
+ "tokenizer_max_length": 100,
174
+ "proj_width": 1024,
175
+ "dropout": 0.1,
176
+ "num_steps": 10,
177
+ "advantage_threshold": 0.0,
178
+ "advantage": "on",
179
+ "init_strategy": "no_init",
180
+ "use_cache": true,
181
+ "attention_implementation": "eager",
182
+ "freeze_vision_encoder": false,
183
+ "train_expert_only": false,
184
+ "train_state_proj": true,
185
+ "optimizer_lr": 2.5e-05,
186
+ "optimizer_betas": [
187
+ 0.9,
188
+ 0.95
189
+ ],
190
+ "optimizer_eps": 1e-08,
191
+ "optimizer_weight_decay": 1e-10,
192
+ "scheduler_warmup_steps": 1000,
193
+ "scheduler_decay_steps": 30000,
194
+ "scheduler_decay_lr": 2.5e-06
195
+ },
196
+ "output_dir": "outputs/train/2026-02-27/03-29-24_pi0",
197
+ "job_name": "pi0",
198
+ "resume": false,
199
+ "seed": 1000,
200
+ "resolution": [
201
+ 224,
202
+ 224
203
+ ],
204
+ "num_cams": 2,
205
+ "max_state_dim": 32,
206
+ "max_action_dim": 32,
207
+ "action_chunk": 10,
208
+ "loss_weighting": {
209
+ "MSE": 1.0,
210
+ "CE": 1.0
211
+ },
212
+ "num_workers": 16,
213
+ "batch_size": 20,
214
+ "gradient_accumulation_steps": 1,
215
+ "dataloader_batch_size": 20,
216
+ "prefetch_factor": 8,
217
+ "steps": 100000,
218
+ "log_freq": 10,
219
+ "save_checkpoint": true,
220
+ "save_freq": 1000,
221
+ "use_policy_training_preset": true,
222
+ "optimizer": {
223
+ "type": "adamw",
224
+ "lr": 2.5e-05,
225
+ "weight_decay": 1e-10,
226
+ "grad_clip_norm": 10.0,
227
+ "betas": [
228
+ 0.9,
229
+ 0.95
230
+ ],
231
+ "eps": 1e-08
232
+ },
233
+ "scheduler": {
234
+ "type": "cosine_decay_with_warmup",
235
+ "num_warmup_steps": 1000,
236
+ "num_decay_steps": 30000,
237
+ "peak_lr": 2.5e-05,
238
+ "decay_lr": 2.5e-06
239
+ },
240
+ "wandb": {
241
+ "enable": true,
242
+ "entity": "wyautox-autox",
243
+ "project": "pi0_star",
244
+ "run_id": "vodcmf09",
245
+ "name": null,
246
+ "notes": "PI0* training run",
247
+ "tags": [],
248
+ "group": null,
249
+ "job_type": null,
250
+ "mode": null,
251
+ "allow_resume": true,
252
+ "disable_artifact": false
253
+ },
254
+ "debug": false,
255
+ "trace_nans": false,
256
+ "env": {
257
+ "type": "libero",
258
+ "import_name": null,
259
+ "make_id": null,
260
+ "task": "libero_10",
261
+ "fps": 10,
262
+ "features": {
263
+ "action": {
264
+ "type": "ACTION",
265
+ "shape": [
266
+ 7
267
+ ]
268
+ },
269
+ "agent_pos": {
270
+ "type": "STATE",
271
+ "shape": [
272
+ 8
273
+ ]
274
+ },
275
+ "pixels/agentview_image": {
276
+ "type": "VISUAL",
277
+ "shape": [
278
+ 360,
279
+ 360,
280
+ 3
281
+ ]
282
+ },
283
+ "pixels/robot0_eye_in_hand_image": {
284
+ "type": "VISUAL",
285
+ "shape": [
286
+ 360,
287
+ 360,
288
+ 3
289
+ ]
290
+ }
291
+ },
292
+ "features_map": {
293
+ "action": "actions",
294
+ "agent_pos": "observation.state",
295
+ "pixels/agentview_image": "observation.images.image",
296
+ "pixels/robot0_eye_in_hand_image": "observation.images.image2"
297
+ },
298
+ "max_parallel_tasks": 1,
299
+ "disable_env_checker": true,
300
+ "task_ids": [
301
+ 8,
302
+ 8,
303
+ 8,
304
+ 8,
305
+ 8,
306
+ 8,
307
+ 8,
308
+ 8
309
+ ],
310
+ "episode_length": 520,
311
+ "obs_type": "pixels_agent_pos",
312
+ "render_mode": "rgb_array",
313
+ "camera_name": "agentview_image,robot0_eye_in_hand_image",
314
+ "init_states": true,
315
+ "camera_name_mapping": null
316
+ },
317
+ "eval": {
318
+ "n_episodes": 16,
319
+ "batch_size": 16,
320
+ "use_async_envs": true,
321
+ "max_episodes_rendered": 16,
322
+ "grid_size": null,
323
+ "recording_root": null
324
+ },
325
+ "eval_freq": 1000,
326
+ "val_freq": 0,
327
+ "last_checkpoint_only": false,
328
+ "server": {
329
+ "port": 50051,
330
+ "max_workers": 4,
331
+ "max_send_message_length_mb": 100,
332
+ "max_receive_message_length_mb": 100
333
+ }
334
+ }