Hugging Face's logo

TensorAuto
/
moka_pot_RECAP_R0

Robotics
Safetensors
PyTorch
opentau
vla
pi0
libero
Reinforcement Learning
manipulation
flow-matching
Model card Files
xet
 Community
autox-lerobot-ci commited on
Commit
92b11cc
·
1 Parent(s): cfd6c15

Uploading Checkpoints

Browse files
Files changed (3) hide show
  1. config.json +83 -0
  2. model.safetensors +3 -0
  3. train_config.json +894 -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": 0.049553,
66
+ "advantage": "use",
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:aad4b1cc4372f9324bd560cf0be7150f7d6ba77542847eff1a89938cde0d0560
3
+ size 16112213576
train_config.json ADDED
@@ -0,0 +1,894 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset_mixture": {
3
+ "datasets": [
4
+ {
5
+ "repo_id": "physical-intelligence/libero",
6
+ "grounding": null,
7
+ "root": "/fss/bot/libero",
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
+ "repo_id": "physical-intelligence/libero",
105
+ "grounding": null,
106
+ "root": "/fss/bot/libero_rollout_0_rank0/rank0",
107
+ "episodes": null,
108
+ "image_transforms": {
109
+ "enable": false,
110
+ "max_num_transforms": 3,
111
+ "random_order": false,
112
+ "tfs": {
113
+ "brightness": {
114
+ "weight": 1.0,
115
+ "type": "ColorJitter",
116
+ "kwargs": {
117
+ "brightness": [
118
+ 0.8,
119
+ 1.2
120
+ ]
121
+ }
122
+ },
123
+ "contrast": {
124
+ "weight": 1.0,
125
+ "type": "ColorJitter",
126
+ "kwargs": {
127
+ "contrast": [
128
+ 0.8,
129
+ 1.2
130
+ ]
131
+ }
132
+ },
133
+ "saturation": {
134
+ "weight": 1.0,
135
+ "type": "ColorJitter",
136
+ "kwargs": {
137
+ "saturation": [
138
+ 0.5,
139
+ 1.5
140
+ ]
141
+ }
142
+ },
143
+ "hue": {
144
+ "weight": 1.0,
145
+ "type": "ColorJitter",
146
+ "kwargs": {
147
+ "hue": [
148
+ -0.05,
149
+ 0.05
150
+ ]
151
+ }
152
+ },
153
+ "sharpness": {
154
+ "weight": 1.0,
155
+ "type": "SharpnessJitter",
156
+ "kwargs": {
157
+ "sharpness": [
158
+ 0.5,
159
+ 1.5
160
+ ]
161
+ }
162
+ }
163
+ }
164
+ },
165
+ "revision": null,
166
+ "use_imagenet_stats": true,
167
+ "video_backend": "pyav",
168
+ "stats": null,
169
+ "data_features_name_mapping": null,
170
+ "val_split_ratio": 0.05
171
+ },
172
+ {
173
+ "repo_id": "physical-intelligence/libero",
174
+ "grounding": null,
175
+ "root": "/fss/bot/libero_rollout_0_rank1/rank0",
176
+ "episodes": null,
177
+ "image_transforms": {
178
+ "enable": false,
179
+ "max_num_transforms": 3,
180
+ "random_order": false,
181
+ "tfs": {
182
+ "brightness": {
183
+ "weight": 1.0,
184
+ "type": "ColorJitter",
185
+ "kwargs": {
186
+ "brightness": [
187
+ 0.8,
188
+ 1.2
189
+ ]
190
+ }
191
+ },
192
+ "contrast": {
193
+ "weight": 1.0,
194
+ "type": "ColorJitter",
195
+ "kwargs": {
196
+ "contrast": [
197
+ 0.8,
198
+ 1.2
199
+ ]
200
+ }
201
+ },
202
+ "saturation": {
203
+ "weight": 1.0,
204
+ "type": "ColorJitter",
205
+ "kwargs": {
206
+ "saturation": [
207
+ 0.5,
208
+ 1.5
209
+ ]
210
+ }
211
+ },
212
+ "hue": {
213
+ "weight": 1.0,
214
+ "type": "ColorJitter",
215
+ "kwargs": {
216
+ "hue": [
217
+ -0.05,
218
+ 0.05
219
+ ]
220
+ }
221
+ },
222
+ "sharpness": {
223
+ "weight": 1.0,
224
+ "type": "SharpnessJitter",
225
+ "kwargs": {
226
+ "sharpness": [
227
+ 0.5,
228
+ 1.5
229
+ ]
230
+ }
231
+ }
232
+ }
233
+ },
234
+ "revision": null,
235
+ "use_imagenet_stats": true,
236
+ "video_backend": "pyav",
237
+ "stats": null,
238
+ "data_features_name_mapping": null,
239
+ "val_split_ratio": 0.05
240
+ },
241
+ {
242
+ "repo_id": "physical-intelligence/libero",
243
+ "grounding": null,
244
+ "root": "/fss/bot/libero_rollout_0_rank2/rank0",
245
+ "episodes": null,
246
+ "image_transforms": {
247
+ "enable": false,
248
+ "max_num_transforms": 3,
249
+ "random_order": false,
250
+ "tfs": {
251
+ "brightness": {
252
+ "weight": 1.0,
253
+ "type": "ColorJitter",
254
+ "kwargs": {
255
+ "brightness": [
256
+ 0.8,
257
+ 1.2
258
+ ]
259
+ }
260
+ },
261
+ "contrast": {
262
+ "weight": 1.0,
263
+ "type": "ColorJitter",
264
+ "kwargs": {
265
+ "contrast": [
266
+ 0.8,
267
+ 1.2
268
+ ]
269
+ }
270
+ },
271
+ "saturation": {
272
+ "weight": 1.0,
273
+ "type": "ColorJitter",
274
+ "kwargs": {
275
+ "saturation": [
276
+ 0.5,
277
+ 1.5
278
+ ]
279
+ }
280
+ },
281
+ "hue": {
282
+ "weight": 1.0,
283
+ "type": "ColorJitter",
284
+ "kwargs": {
285
+ "hue": [
286
+ -0.05,
287
+ 0.05
288
+ ]
289
+ }
290
+ },
291
+ "sharpness": {
292
+ "weight": 1.0,
293
+ "type": "SharpnessJitter",
294
+ "kwargs": {
295
+ "sharpness": [
296
+ 0.5,
297
+ 1.5
298
+ ]
299
+ }
300
+ }
301
+ }
302
+ },
303
+ "revision": null,
304
+ "use_imagenet_stats": true,
305
+ "video_backend": "pyav",
306
+ "stats": null,
307
+ "data_features_name_mapping": null,
308
+ "val_split_ratio": 0.05
309
+ },
310
+ {
311
+ "repo_id": "physical-intelligence/libero",
312
+ "grounding": null,
313
+ "root": "/fss/bot/libero_rollout_0_rank3/rank0",
314
+ "episodes": null,
315
+ "image_transforms": {
316
+ "enable": false,
317
+ "max_num_transforms": 3,
318
+ "random_order": false,
319
+ "tfs": {
320
+ "brightness": {
321
+ "weight": 1.0,
322
+ "type": "ColorJitter",
323
+ "kwargs": {
324
+ "brightness": [
325
+ 0.8,
326
+ 1.2
327
+ ]
328
+ }
329
+ },
330
+ "contrast": {
331
+ "weight": 1.0,
332
+ "type": "ColorJitter",
333
+ "kwargs": {
334
+ "contrast": [
335
+ 0.8,
336
+ 1.2
337
+ ]
338
+ }
339
+ },
340
+ "saturation": {
341
+ "weight": 1.0,
342
+ "type": "ColorJitter",
343
+ "kwargs": {
344
+ "saturation": [
345
+ 0.5,
346
+ 1.5
347
+ ]
348
+ }
349
+ },
350
+ "hue": {
351
+ "weight": 1.0,
352
+ "type": "ColorJitter",
353
+ "kwargs": {
354
+ "hue": [
355
+ -0.05,
356
+ 0.05
357
+ ]
358
+ }
359
+ },
360
+ "sharpness": {
361
+ "weight": 1.0,
362
+ "type": "SharpnessJitter",
363
+ "kwargs": {
364
+ "sharpness": [
365
+ 0.5,
366
+ 1.5
367
+ ]
368
+ }
369
+ }
370
+ }
371
+ },
372
+ "revision": null,
373
+ "use_imagenet_stats": true,
374
+ "video_backend": "pyav",
375
+ "stats": null,
376
+ "data_features_name_mapping": null,
377
+ "val_split_ratio": 0.05
378
+ },
379
+ {
380
+ "repo_id": "physical-intelligence/libero",
381
+ "grounding": null,
382
+ "root": "/fss/bot/libero_rollout_0_rank4/rank0",
383
+ "episodes": null,
384
+ "image_transforms": {
385
+ "enable": false,
386
+ "max_num_transforms": 3,
387
+ "random_order": false,
388
+ "tfs": {
389
+ "brightness": {
390
+ "weight": 1.0,
391
+ "type": "ColorJitter",
392
+ "kwargs": {
393
+ "brightness": [
394
+ 0.8,
395
+ 1.2
396
+ ]
397
+ }
398
+ },
399
+ "contrast": {
400
+ "weight": 1.0,
401
+ "type": "ColorJitter",
402
+ "kwargs": {
403
+ "contrast": [
404
+ 0.8,
405
+ 1.2
406
+ ]
407
+ }
408
+ },
409
+ "saturation": {
410
+ "weight": 1.0,
411
+ "type": "ColorJitter",
412
+ "kwargs": {
413
+ "saturation": [
414
+ 0.5,
415
+ 1.5
416
+ ]
417
+ }
418
+ },
419
+ "hue": {
420
+ "weight": 1.0,
421
+ "type": "ColorJitter",
422
+ "kwargs": {
423
+ "hue": [
424
+ -0.05,
425
+ 0.05
426
+ ]
427
+ }
428
+ },
429
+ "sharpness": {
430
+ "weight": 1.0,
431
+ "type": "SharpnessJitter",
432
+ "kwargs": {
433
+ "sharpness": [
434
+ 0.5,
435
+ 1.5
436
+ ]
437
+ }
438
+ }
439
+ }
440
+ },
441
+ "revision": null,
442
+ "use_imagenet_stats": true,
443
+ "video_backend": "pyav",
444
+ "stats": null,
445
+ "data_features_name_mapping": null,
446
+ "val_split_ratio": 0.05
447
+ },
448
+ {
449
+ "repo_id": "physical-intelligence/libero",
450
+ "grounding": null,
451
+ "root": "/fss/bot/libero_rollout_0_rank5/rank0",
452
+ "episodes": null,
453
+ "image_transforms": {
454
+ "enable": false,
455
+ "max_num_transforms": 3,
456
+ "random_order": false,
457
+ "tfs": {
458
+ "brightness": {
459
+ "weight": 1.0,
460
+ "type": "ColorJitter",
461
+ "kwargs": {
462
+ "brightness": [
463
+ 0.8,
464
+ 1.2
465
+ ]
466
+ }
467
+ },
468
+ "contrast": {
469
+ "weight": 1.0,
470
+ "type": "ColorJitter",
471
+ "kwargs": {
472
+ "contrast": [
473
+ 0.8,
474
+ 1.2
475
+ ]
476
+ }
477
+ },
478
+ "saturation": {
479
+ "weight": 1.0,
480
+ "type": "ColorJitter",
481
+ "kwargs": {
482
+ "saturation": [
483
+ 0.5,
484
+ 1.5
485
+ ]
486
+ }
487
+ },
488
+ "hue": {
489
+ "weight": 1.0,
490
+ "type": "ColorJitter",
491
+ "kwargs": {
492
+ "hue": [
493
+ -0.05,
494
+ 0.05
495
+ ]
496
+ }
497
+ },
498
+ "sharpness": {
499
+ "weight": 1.0,
500
+ "type": "SharpnessJitter",
501
+ "kwargs": {
502
+ "sharpness": [
503
+ 0.5,
504
+ 1.5
505
+ ]
506
+ }
507
+ }
508
+ }
509
+ },
510
+ "revision": null,
511
+ "use_imagenet_stats": true,
512
+ "video_backend": "pyav",
513
+ "stats": null,
514
+ "data_features_name_mapping": null,
515
+ "val_split_ratio": 0.05
516
+ },
517
+ {
518
+ "repo_id": "physical-intelligence/libero",
519
+ "grounding": null,
520
+ "root": "/fss/bot/libero_rollout_0_rank6/rank0",
521
+ "episodes": null,
522
+ "image_transforms": {
523
+ "enable": false,
524
+ "max_num_transforms": 3,
525
+ "random_order": false,
526
+ "tfs": {
527
+ "brightness": {
528
+ "weight": 1.0,
529
+ "type": "ColorJitter",
530
+ "kwargs": {
531
+ "brightness": [
532
+ 0.8,
533
+ 1.2
534
+ ]
535
+ }
536
+ },
537
+ "contrast": {
538
+ "weight": 1.0,
539
+ "type": "ColorJitter",
540
+ "kwargs": {
541
+ "contrast": [
542
+ 0.8,
543
+ 1.2
544
+ ]
545
+ }
546
+ },
547
+ "saturation": {
548
+ "weight": 1.0,
549
+ "type": "ColorJitter",
550
+ "kwargs": {
551
+ "saturation": [
552
+ 0.5,
553
+ 1.5
554
+ ]
555
+ }
556
+ },
557
+ "hue": {
558
+ "weight": 1.0,
559
+ "type": "ColorJitter",
560
+ "kwargs": {
561
+ "hue": [
562
+ -0.05,
563
+ 0.05
564
+ ]
565
+ }
566
+ },
567
+ "sharpness": {
568
+ "weight": 1.0,
569
+ "type": "SharpnessJitter",
570
+ "kwargs": {
571
+ "sharpness": [
572
+ 0.5,
573
+ 1.5
574
+ ]
575
+ }
576
+ }
577
+ }
578
+ },
579
+ "revision": null,
580
+ "use_imagenet_stats": true,
581
+ "video_backend": "pyav",
582
+ "stats": null,
583
+ "data_features_name_mapping": null,
584
+ "val_split_ratio": 0.05
585
+ },
586
+ {
587
+ "repo_id": "physical-intelligence/libero",
588
+ "grounding": null,
589
+ "root": "/fss/bot/libero_rollout_0_rank7/rank0",
590
+ "episodes": null,
591
+ "image_transforms": {
592
+ "enable": false,
593
+ "max_num_transforms": 3,
594
+ "random_order": false,
595
+ "tfs": {
596
+ "brightness": {
597
+ "weight": 1.0,
598
+ "type": "ColorJitter",
599
+ "kwargs": {
600
+ "brightness": [
601
+ 0.8,
602
+ 1.2
603
+ ]
604
+ }
605
+ },
606
+ "contrast": {
607
+ "weight": 1.0,
608
+ "type": "ColorJitter",
609
+ "kwargs": {
610
+ "contrast": [
611
+ 0.8,
612
+ 1.2
613
+ ]
614
+ }
615
+ },
616
+ "saturation": {
617
+ "weight": 1.0,
618
+ "type": "ColorJitter",
619
+ "kwargs": {
620
+ "saturation": [
621
+ 0.5,
622
+ 1.5
623
+ ]
624
+ }
625
+ },
626
+ "hue": {
627
+ "weight": 1.0,
628
+ "type": "ColorJitter",
629
+ "kwargs": {
630
+ "hue": [
631
+ -0.05,
632
+ 0.05
633
+ ]
634
+ }
635
+ },
636
+ "sharpness": {
637
+ "weight": 1.0,
638
+ "type": "SharpnessJitter",
639
+ "kwargs": {
640
+ "sharpness": [
641
+ 0.5,
642
+ 1.5
643
+ ]
644
+ }
645
+ }
646
+ }
647
+ },
648
+ "revision": null,
649
+ "use_imagenet_stats": true,
650
+ "video_backend": "pyav",
651
+ "stats": null,
652
+ "data_features_name_mapping": null,
653
+ "val_split_ratio": 0.05
654
+ }
655
+ ],
656
+ "weights": [
657
+ 1.0,
658
+ 1.0,
659
+ 1.0,
660
+ 1.0,
661
+ 1.0,
662
+ 1.0,
663
+ 1.0,
664
+ 1.0,
665
+ 1.0
666
+ ],
667
+ "action_freq": 10.0,
668
+ "image_resample_strategy": "nearest",
669
+ "vector_resample_strategy": "nearest",
670
+ "val_split_ratio": 0.05
671
+ },
672
+ "policy": {
673
+ "type": "pi0",
674
+ "n_obs_steps": 1,
675
+ "normalization_mapping": {
676
+ "VISUAL": "IDENTITY",
677
+ "STATE": "MEAN_STD",
678
+ "ACTION": "MEAN_STD"
679
+ },
680
+ "input_features": {
681
+ "state": {
682
+ "type": "STATE",
683
+ "shape": [
684
+ 32
685
+ ]
686
+ },
687
+ "camera0": {
688
+ "type": "VISUAL",
689
+ "shape": [
690
+ 3,
691
+ 224,
692
+ 224
693
+ ]
694
+ },
695
+ "camera1": {
696
+ "type": "VISUAL",
697
+ "shape": [
698
+ 3,
699
+ 224,
700
+ 224
701
+ ]
702
+ }
703
+ },
704
+ "output_features": {
705
+ "actions": {
706
+ "type": "ACTION",
707
+ "shape": [
708
+ 32
709
+ ]
710
+ }
711
+ },
712
+ "device": null,
713
+ "use_amp": false,
714
+ "pretrained_path": "outputs/train/2026-03-13/18-17-26_pi0/checkpoints/021000",
715
+ "cloud_vlm_latency_mean": 0.0,
716
+ "cloud_vlm_latency_std": 0.0,
717
+ "cloud_vlm_latency_lower": 0.0,
718
+ "cloud_vlm_latency_upper": 0.0,
719
+ "action_decoder_latency_mean": 0.0,
720
+ "action_decoder_latency_std": 0.0,
721
+ "action_decoder_latency_lower": 0.0,
722
+ "action_decoder_latency_upper": 0.0,
723
+ "chunk_size": 10,
724
+ "n_action_steps": 10,
725
+ "safety_buffer": 0,
726
+ "max_state_dim": 32,
727
+ "max_action_dim": 32,
728
+ "resize_imgs_with_padding": [
729
+ 224,
730
+ 224
731
+ ],
732
+ "empty_cameras": 0,
733
+ "tokenizer_max_length": 100,
734
+ "proj_width": 1024,
735
+ "dropout": 0.1,
736
+ "num_steps": 10,
737
+ "advantage_threshold": 0.049553,
738
+ "advantage": "use",
739
+ "init_strategy": "no_init",
740
+ "use_cache": true,
741
+ "attention_implementation": "eager",
742
+ "freeze_vision_encoder": false,
743
+ "train_expert_only": false,
744
+ "train_state_proj": true,
745
+ "optimizer_lr": 2.5e-05,
746
+ "optimizer_betas": [
747
+ 0.9,
748
+ 0.95
749
+ ],
750
+ "optimizer_eps": 1e-08,
751
+ "optimizer_weight_decay": 1e-10,
752
+ "scheduler_warmup_steps": 1000,
753
+ "scheduler_decay_steps": 30000,
754
+ "scheduler_decay_lr": 2.5e-06
755
+ },
756
+ "output_dir": "outputs/train/2026-03-13/18-17-26_pi0",
757
+ "job_name": "pi0",
758
+ "resume": false,
759
+ "seed": 6000,
760
+ "resolution": [
761
+ 224,
762
+ 224
763
+ ],
764
+ "num_cams": 2,
765
+ "max_state_dim": 32,
766
+ "max_action_dim": 32,
767
+ "action_chunk": 10,
768
+ "loss_weighting": {
769
+ "MSE": 1.0,
770
+ "CE": 1.0
771
+ },
772
+ "num_workers": 16,
773
+ "batch_size": 20,
774
+ "gradient_accumulation_steps": 1,
775
+ "dataloader_batch_size": 20,
776
+ "prefetch_factor": 8,
777
+ "steps": 100000,
778
+ "log_freq": 10,
779
+ "save_checkpoint": true,
780
+ "save_freq": 1000,
781
+ "use_policy_training_preset": true,
782
+ "optimizer": {
783
+ "type": "adamw",
784
+ "lr": 2.5e-05,
785
+ "weight_decay": 1e-10,
786
+ "grad_clip_norm": 10.0,
787
+ "betas": [
788
+ 0.9,
789
+ 0.95
790
+ ],
791
+ "eps": 1e-08
792
+ },
793
+ "scheduler": {
794
+ "type": "cosine_decay_with_warmup",
795
+ "num_warmup_steps": 1000,
796
+ "num_decay_steps": 30000,
797
+ "peak_lr": 2.5e-05,
798
+ "decay_lr": 2.5e-06
799
+ },
800
+ "wandb": {
801
+ "enable": true,
802
+ "entity": "wyautox-autox",
803
+ "project": "pi0_star",
804
+ "run_id": "z6j1yeyb",
805
+ "name": null,
806
+ "notes": "PI0* training run",
807
+ "tags": [],
808
+ "group": null,
809
+ "job_type": null,
810
+ "mode": null,
811
+ "allow_resume": true,
812
+ "disable_artifact": false
813
+ },
814
+ "debug": false,
815
+ "trace_nans": false,
816
+ "env": {
817
+ "type": "libero",
818
+ "import_name": null,
819
+ "make_id": null,
820
+ "task": "libero_10",
821
+ "fps": 10,
822
+ "features": {
823
+ "action": {
824
+ "type": "ACTION",
825
+ "shape": [
826
+ 7
827
+ ]
828
+ },
829
+ "agent_pos": {
830
+ "type": "STATE",
831
+ "shape": [
832
+ 8
833
+ ]
834
+ },
835
+ "pixels/agentview_image": {
836
+ "type": "VISUAL",
837
+ "shape": [
838
+ 360,
839
+ 360,
840
+ 3
841
+ ]
842
+ },
843
+ "pixels/robot0_eye_in_hand_image": {
844
+ "type": "VISUAL",
845
+ "shape": [
846
+ 360,
847
+ 360,
848
+ 3
849
+ ]
850
+ }
851
+ },
852
+ "features_map": {
853
+ "action": "actions",
854
+ "agent_pos": "observation.state",
855
+ "pixels/agentview_image": "observation.images.image",
856
+ "pixels/robot0_eye_in_hand_image": "observation.images.image2"
857
+ },
858
+ "max_parallel_tasks": 1,
859
+ "disable_env_checker": true,
860
+ "task_ids": [
861
+ 8,
862
+ 8,
863
+ 8,
864
+ 8,
865
+ 8,
866
+ 8,
867
+ 8,
868
+ 8
869
+ ],
870
+ "episode_length": 520,
871
+ "obs_type": "pixels_agent_pos",
872
+ "render_mode": "rgb_array",
873
+ "camera_name": "agentview_image,robot0_eye_in_hand_image",
874
+ "init_states": true,
875
+ "camera_name_mapping": null
876
+ },
877
+ "eval": {
878
+ "n_episodes": 16,
879
+ "batch_size": 16,
880
+ "use_async_envs": true,
881
+ "max_episodes_rendered": 16,
882
+ "grid_size": null,
883
+ "recording_root": null
884
+ },
885
+ "eval_freq": 1000,
886
+ "val_freq": 0,
887
+ "last_checkpoint_only": false,
888
+ "server": {
889
+ "port": 50051,
890
+ "max_workers": 4,
891
+ "max_send_message_length_mb": 100,
892
+ "max_receive_message_length_mb": 100
893
+ }
894
+ }