TTaekwan commited on
Commit
ecb06cf
·
verified ·
1 Parent(s): fcae19e

openarm_bimanual_base: final checkpoint (checkpoint-40000)

Browse files
openarm_bimanual_base/config.json ADDED
@@ -0,0 +1,73 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action_dim": 32,
3
+ "action_head_cfg": {
4
+ "action_dim": 32,
5
+ "action_horizon": 16,
6
+ "add_pos_embed": true,
7
+ "backbone_embedding_dim": 2048,
8
+ "diffusion_model_cfg": {
9
+ "attention_head_dim": 48,
10
+ "cross_attention_dim": 2048,
11
+ "dropout": 0.2,
12
+ "final_dropout": true,
13
+ "interleave_self_attention": true,
14
+ "norm_type": "ada_norm",
15
+ "num_attention_heads": 32,
16
+ "num_layers": 16,
17
+ "output_dim": 1024,
18
+ "positional_embeddings": null
19
+ },
20
+ "future_seq_len": 16,
21
+ "hidden_size": 1024,
22
+ "history_seq_len": 16,
23
+ "input_embedding_dim": 1536,
24
+ "max_action_dim": 32,
25
+ "max_state_dim": 64,
26
+ "model_dtype": "float32",
27
+ "n_cond_tokens": 16,
28
+ "noise_beta_alpha": 1.5,
29
+ "noise_beta_beta": 1.0,
30
+ "noise_s": 0.999,
31
+ "num_inference_timesteps": 4,
32
+ "num_target_vision_tokens": 32,
33
+ "num_timestep_buckets": 1000,
34
+ "physics_dropout_prob": 0.0,
35
+ "prediction_mode": "base",
36
+ "tune_diffusion_model": true,
37
+ "tune_projector": true,
38
+ "use_key_value_gate": false,
39
+ "use_logit_gate": false,
40
+ "use_vlln": true,
41
+ "vl_self_attention_cfg": {
42
+ "attention_head_dim": 64,
43
+ "dropout": 0.2,
44
+ "final_dropout": true,
45
+ "num_attention_heads": 32,
46
+ "num_layers": 4,
47
+ "positional_embeddings": null
48
+ }
49
+ },
50
+ "action_horizon": 16,
51
+ "architectures": [
52
+ "GR00T_N1_MoSS"
53
+ ],
54
+ "attn_implementation": null,
55
+ "backbone_cfg": {
56
+ "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
57
+ "load_bf16": false,
58
+ "project_to_dim": null,
59
+ "reproject_vision": false,
60
+ "select_layer": 12,
61
+ "tune_llm": false,
62
+ "tune_visual": true,
63
+ "use_flash_attention": true
64
+ },
65
+ "compute_dtype": "bfloat16",
66
+ "cond_encoder_type": "mlp",
67
+ "hidden_size": 2048,
68
+ "model_dtype": "float32",
69
+ "model_type": "gr00t_n1_5",
70
+ "prediction_mode": "base",
71
+ "torch_dtype": "bfloat16",
72
+ "transformers_version": "4.51.3"
73
+ }
openarm_bimanual_base/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,535 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "left_arm": {
6
+ "max": [
7
+ 0.7036316394805908,
8
+ 0.024986648932099342,
9
+ 1.0450522899627686,
10
+ 2.2344930171966553,
11
+ 0.9515907764434814,
12
+ 0.7207980751991272,
13
+ 0.4888609051704407
14
+ ],
15
+ "min": [
16
+ -0.41638055443763733,
17
+ -0.6101701259613037,
18
+ -0.4697871506214142,
19
+ 0.6715877056121826,
20
+ -0.17757686972618103,
21
+ -0.7375829815864563,
22
+ -0.37861448526382446
23
+ ],
24
+ "mean": [
25
+ 0.08194174617528915,
26
+ -0.2163074016571045,
27
+ 0.08329900354146957,
28
+ 1.5327811241149902,
29
+ 0.37673911452293396,
30
+ 0.09872952848672867,
31
+ 0.0013253180077299476
32
+ ],
33
+ "std": [
34
+ 0.24660606682300568,
35
+ 0.13784821331501007,
36
+ 0.3305414319038391,
37
+ 0.2744458317756653,
38
+ 0.24320469796657562,
39
+ 0.24266797304153442,
40
+ 0.12675808370113373
41
+ ],
42
+ "q01": [
43
+ -0.3377965986728668,
44
+ -0.5815594792366028,
45
+ -0.3572518527507782,
46
+ 0.8264744281768799,
47
+ -0.13218127191066742,
48
+ -0.5399786233901978,
49
+ -0.25158312916755676
50
+ ],
51
+ "q99": [
52
+ 0.5739223361015304,
53
+ 0.00019073777366429567,
54
+ 0.8310444951057434,
55
+ 2.0723659992218018,
56
+ 0.825696170330046,
57
+ 0.6746395230293274,
58
+ 0.37899595499038696
59
+ ]
60
+ },
61
+ "right_arm": {
62
+ "max": [
63
+ 0.6269550919532776,
64
+ 0.576218843460083,
65
+ 0.23861294984817505,
66
+ 1.962119460105896,
67
+ 0.2508201599121094,
68
+ 0.3126192092895508,
69
+ 0.536163866519928
70
+ ],
71
+ "min": [
72
+ -0.5666819214820862,
73
+ -0.12073700875043869,
74
+ -1.1671243906021118,
75
+ 0.7825970649719238,
76
+ -0.8512626886367798,
77
+ -0.7551308274269104,
78
+ -0.465209424495697
79
+ ],
80
+ "mean": [
81
+ 0.03592834249138832,
82
+ 0.19172526895999908,
83
+ -0.2652074992656708,
84
+ 1.448072075843811,
85
+ -0.21943968534469604,
86
+ -0.2309868037700653,
87
+ 0.08182784169912338
88
+ ],
89
+ "std": [
90
+ 0.17945554852485657,
91
+ 0.14554661512374878,
92
+ 0.2392909675836563,
93
+ 0.18009504675865173,
94
+ 0.2647036910057068,
95
+ 0.19989536702632904,
96
+ 0.15510717034339905
97
+ ],
98
+ "q01": [
99
+ -0.34962233901023865,
100
+ -0.07915617525577545,
101
+ -0.8924620151519775,
102
+ 0.9386205673217773,
103
+ -0.731090247631073,
104
+ -0.7040131092071533,
105
+ -0.32101166248321533
106
+ ],
107
+ "q99": [
108
+ 0.5033569931983948,
109
+ 0.5201418995857239,
110
+ 0.164988175034523,
111
+ 1.84309148788452,
112
+ 0.18749523162841797,
113
+ 0.22793163359165192,
114
+ 0.3896772861480713
115
+ ]
116
+ },
117
+ "left_hand": {
118
+ "max": [
119
+ 1.288053035736084,
120
+ 0.4398229718208313,
121
+ 1.1606439352035522,
122
+ 1.3159782886505127,
123
+ 1.2967796325683594,
124
+ 1.275835633277893
125
+ ],
126
+ "min": [
127
+ 0.012217304669320583,
128
+ 0.006981316953897476,
129
+ -0.10471975803375244,
130
+ -0.036651913076639175,
131
+ -0.07330382615327835,
132
+ -0.07155849784612656
133
+ ],
134
+ "mean": [
135
+ 0.7000418305397034,
136
+ 0.26935580372810364,
137
+ 0.48099249601364136,
138
+ 0.5676980018615723,
139
+ 0.6116954684257507,
140
+ 0.4122258126735687
141
+ ],
142
+ "std": [
143
+ 0.19492755830287933,
144
+ 0.10537338256835938,
145
+ 0.3656688630580902,
146
+ 0.3835751712322235,
147
+ 0.41250866651535034,
148
+ 0.43397054076194763
149
+ ],
150
+ "q01": [
151
+ 0.03490658476948738,
152
+ 0.012217304669320583,
153
+ -0.0890117883682251,
154
+ -0.02094395086169243,
155
+ -0.06283185631036758,
156
+ -0.07155849784612656
157
+ ],
158
+ "q99": [
159
+ 1.0838494300842285,
160
+ 0.3787364363670349,
161
+ 1.0995573997497559,
162
+ 1.2705997228622437,
163
+ 1.2653636932373047,
164
+ 1.1536626815795898
165
+ ]
166
+ },
167
+ "right_hand": {
168
+ "max": [
169
+ 1.2252211570739746,
170
+ 0.46076691150665283,
171
+ 0.6248278617858887,
172
+ 0.9215338230133057,
173
+ 0.9267698526382446,
174
+ 1.487020492553711
175
+ ],
176
+ "min": [
177
+ -0.09250245243310928,
178
+ 0.010471975430846214,
179
+ -0.10995574295520782,
180
+ -0.0593411959707737,
181
+ -0.005235987715423107,
182
+ -0.07679449021816254
183
+ ],
184
+ "mean": [
185
+ 0.4737979471683502,
186
+ 0.14215154945850372,
187
+ 0.23598714172840118,
188
+ 0.20759284496307373,
189
+ 0.3103739023208618,
190
+ 0.09935002028942108
191
+ ],
192
+ "std": [
193
+ 0.29928818345069885,
194
+ 0.10472391545772552,
195
+ 0.087922103703022,
196
+ 0.12231124192476273,
197
+ 0.10847391188144684,
198
+ 0.09348537027835846
199
+ ],
200
+ "q01": [
201
+ -0.0872664600610733,
202
+ 0.015707964077591896,
203
+ -0.003490658476948738,
204
+ -0.0593411959707737,
205
+ 0.013962633907794952,
206
+ 0.003490658476948738
207
+ ],
208
+ "q99": [
209
+ 1.1082841157913208,
210
+ 0.4415683150291443,
211
+ 0.5044001340866089,
212
+ 0.6771877408027649,
213
+ 0.6248278617858887,
214
+ 0.43109631538391113
215
+ ]
216
+ }
217
+ },
218
+ "action": {
219
+ "left_arm": {
220
+ "max": [
221
+ 0.7045770287513733,
222
+ 0.028094541281461716,
223
+ 1.0507768392562866,
224
+ 2.250253438949585,
225
+ 0.9682290554046631,
226
+ 0.7014127373695374,
227
+ 0.4716990888118744
228
+ ],
229
+ "min": [
230
+ -0.43056878447532654,
231
+ -0.6200153827667236,
232
+ -0.4747670590877533,
233
+ 0.6625833511352539,
234
+ -0.1853245049715042,
235
+ -0.7244224548339844,
236
+ -0.42146122455596924
237
+ ],
238
+ "mean": [
239
+ 0.07833307236433029,
240
+ -0.2198743224143982,
241
+ 0.08574533462524414,
242
+ 1.5401817560195923,
243
+ 0.3842606544494629,
244
+ 0.10566387325525284,
245
+ -0.021024441346526146
246
+ ],
247
+ "std": [
248
+ 0.24943922460079193,
249
+ 0.1376669853925705,
250
+ 0.3327249586582184,
251
+ 0.27775320410728455,
252
+ 0.2413933128118515,
253
+ 0.24254348874092102,
254
+ 0.13175863027572632
255
+ ],
256
+ "q01": [
257
+ -0.34696683287620544,
258
+ -0.5828163623809814,
259
+ -0.3604778242111206,
260
+ 0.8071610331535339,
261
+ -0.11552838236093521,
262
+ -0.5238467645645142,
263
+ -0.2887718975543976
264
+ ],
265
+ "q99": [
266
+ 0.5730381608009338,
267
+ 0.00017311290488576512,
268
+ 0.8386963200569137,
269
+ 2.0838165283203125,
270
+ 0.8336222171783447,
271
+ 0.6990887498855544,
272
+ 0.37697577476501465
273
+ ]
274
+ },
275
+ "right_arm": {
276
+ "max": [
277
+ 0.6380397081375122,
278
+ 0.5828163623809814,
279
+ 0.24275246262550354,
280
+ 1.97490394115448,
281
+ 0.24975170195102692,
282
+ 0.3976845145225525,
283
+ 0.5737088322639465
284
+ ],
285
+ "min": [
286
+ -0.5676692128181458,
287
+ -0.12453299760818481,
288
+ -1.2306361198425293,
289
+ 0.7699620127677917,
290
+ -0.8719717264175415,
291
+ -0.7006457448005676,
292
+ -0.48166996240615845
293
+ ],
294
+ "mean": [
295
+ 0.04062885046005249,
296
+ 0.19403599202632904,
297
+ -0.27041366696357727,
298
+ 1.4504785537719727,
299
+ -0.22287163138389587,
300
+ -0.23296882212162018,
301
+ 0.10938776284456253
302
+ ],
303
+ "std": [
304
+ 0.17571501433849335,
305
+ 0.1470479965209961,
306
+ 0.24391970038414001,
307
+ 0.18444180488586426,
308
+ 0.26275455951690674,
309
+ 0.2114143967628479,
310
+ 0.16702422499656677
311
+ ],
312
+ "q01": [
313
+ -0.3463924825191498,
314
+ -0.08906721323728561,
315
+ -0.925373911857605,
316
+ 0.9314212012290954,
317
+ -0.7312290072441101,
318
+ -0.6998787522315979,
319
+ -0.3298058807849884
320
+ ],
321
+ "q99": [
322
+ 0.47505423426628113,
323
+ 0.5249086022377014,
324
+ 0.16566991806030273,
325
+ 1.8506438326835617,
326
+ 0.19068576812743984,
327
+ 0.26384469866752625,
328
+ 0.4153252840042114
329
+ ]
330
+ },
331
+ "left_hand": {
332
+ "max": [
333
+ 1.318833351135254,
334
+ 0.4246000051498413,
335
+ 1.478600025177002,
336
+ 1.478600025177002,
337
+ 1.4163508415222168,
338
+ 1.478600025177002
339
+ ],
340
+ "min": [
341
+ 0.0,
342
+ 0.0,
343
+ 0.0,
344
+ 0.0,
345
+ 0.0,
346
+ 0.0
347
+ ],
348
+ "mean": [
349
+ 0.6668961644172668,
350
+ 0.3263242840766907,
351
+ 0.7457484602928162,
352
+ 0.7091995477676392,
353
+ 0.7322353720664978,
354
+ 0.4643964469432831
355
+ ],
356
+ "std": [
357
+ 0.19216059148311615,
358
+ 0.1396765261888504,
359
+ 0.5262469053268433,
360
+ 0.4917907416820526,
361
+ 0.46064653992652893,
362
+ 0.4285631775856018
363
+ ],
364
+ "q01": [
365
+ 0.040405966341495514,
366
+ 0.02666635997593403,
367
+ 0.0,
368
+ 0.0,
369
+ 0.0,
370
+ 0.0
371
+ ],
372
+ "q99": [
373
+ 1.0555340361595145,
374
+ 0.4246000051498413,
375
+ 1.47752046585083,
376
+ 1.478600025177002,
377
+ 1.3745312690734863,
378
+ 1.258659472465515
379
+ ]
380
+ },
381
+ "right_hand": {
382
+ "max": [
383
+ 1.530051827430725,
384
+ 0.3528084456920624,
385
+ 1.0272839069366455,
386
+ 1.2636243104934692,
387
+ 0.8926088213920593,
388
+ 1.478600025177002
389
+ ],
390
+ "min": [
391
+ 0.0,
392
+ 0.0,
393
+ 0.0,
394
+ 0.0,
395
+ 0.0,
396
+ 0.0
397
+ ],
398
+ "mean": [
399
+ 0.6204968690872192,
400
+ 0.11970113962888718,
401
+ 0.2611398696899414,
402
+ 0.23086832463741302,
403
+ 0.32911351323127747,
404
+ 0.10907960683107376
405
+ ],
406
+ "std": [
407
+ 0.2979033887386322,
408
+ 0.07613719254732132,
409
+ 0.0970793217420578,
410
+ 0.13380549848079681,
411
+ 0.12169312685728073,
412
+ 0.1095845103263855
413
+ ],
414
+ "q01": [
415
+ 0.0,
416
+ 0.02186192385852337,
417
+ 0.01579903155565262,
418
+ 0.02701820623129606,
419
+ 0.03618954606354237,
420
+ 0.01747006196528673
421
+ ],
422
+ "q99": [
423
+ 1.311769962310791,
424
+ 0.31265057563781723,
425
+ 0.6353671920299523,
426
+ 0.7955591857433311,
427
+ 0.7028631508350371,
428
+ 0.4612578833103176
429
+ ]
430
+ }
431
+ },
432
+ "tactile": {},
433
+ "rlforce": {},
434
+ "rltorque": {},
435
+ "rldepth": {},
436
+ "torque": {},
437
+ "f1_force": {}
438
+ },
439
+ "modalities": {
440
+ "video": {
441
+ "ego_left": {
442
+ "resolution": [
443
+ 1280,
444
+ 720
445
+ ],
446
+ "channels": 3,
447
+ "fps": 30.0
448
+ },
449
+ "ego_right": {
450
+ "resolution": [
451
+ 1280,
452
+ 720
453
+ ],
454
+ "channels": 3,
455
+ "fps": 30.0
456
+ }
457
+ },
458
+ "state": {
459
+ "left_arm": {
460
+ "absolute": true,
461
+ "rotation_type": null,
462
+ "shape": [
463
+ 7
464
+ ],
465
+ "continuous": true
466
+ },
467
+ "right_arm": {
468
+ "absolute": true,
469
+ "rotation_type": null,
470
+ "shape": [
471
+ 7
472
+ ],
473
+ "continuous": true
474
+ },
475
+ "left_hand": {
476
+ "absolute": true,
477
+ "rotation_type": null,
478
+ "shape": [
479
+ 6
480
+ ],
481
+ "continuous": true
482
+ },
483
+ "right_hand": {
484
+ "absolute": true,
485
+ "rotation_type": null,
486
+ "shape": [
487
+ 6
488
+ ],
489
+ "continuous": true
490
+ }
491
+ },
492
+ "action": {
493
+ "left_arm": {
494
+ "absolute": true,
495
+ "rotation_type": null,
496
+ "shape": [
497
+ 7
498
+ ],
499
+ "continuous": true
500
+ },
501
+ "right_arm": {
502
+ "absolute": true,
503
+ "rotation_type": null,
504
+ "shape": [
505
+ 7
506
+ ],
507
+ "continuous": true
508
+ },
509
+ "left_hand": {
510
+ "absolute": true,
511
+ "rotation_type": null,
512
+ "shape": [
513
+ 6
514
+ ],
515
+ "continuous": true
516
+ },
517
+ "right_hand": {
518
+ "absolute": true,
519
+ "rotation_type": null,
520
+ "shape": [
521
+ 6
522
+ ],
523
+ "continuous": true
524
+ }
525
+ },
526
+ "tactile": {},
527
+ "rlforce": {},
528
+ "rltorque": {},
529
+ "rldepth": {},
530
+ "torque": {},
531
+ "f1_force": {}
532
+ },
533
+ "embodiment_tag": "new_embodiment"
534
+ }
535
+ }
openarm_bimanual_base/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c168a8d591b2200aa2c1b57ef70150bc25eaeea71af665a70f64232c9a1f0406
3
+ size 4999367056
openarm_bimanual_base/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:14476c3705f8a976d774cef609765dfc71b699a15c346772a9f105079d679b23
3
+ size 2586705344
openarm_bimanual_base/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
openarm_bimanual_base/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ac245d3087c5948e3fbf118fa4af76bfea7fccbbc8cdd78f56c6b36928eae5f4
3
+ size 8550719959
openarm_bimanual_base/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2d04563a96eb1fd54fbc8e6b0e3370e44e2bb10dca83d07fc2d4930785f7aed7
3
+ size 14917
openarm_bimanual_base/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3e2fcf6d7f04bd60e969c0da40ef166337b6d3576ba261fc91345273a695387d
3
+ size 14917
openarm_bimanual_base/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:566e63d72ec3169fd51f5dc061a30bf9c5195d001f715224cdd5f6178206f3e5
3
+ size 1465
openarm_bimanual_base/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff