Hugging Face's logo

EbotsVLA
/
IDM

Safetensors
Model card Files
xet
 Community
gayatrid-ebots commited on
Commit
c30b085
·
verified ·
1 Parent(s): a000266

Uploaded a new model: cable_harness_v3_and_v3trivialmotions_joint checkpoint-20000

Browse files
Files changed (7) hide show
  1. cable_harness_v3_and_v3trivialmotions_joint/checkpoint-20000/config.json +79 -0
  2. cable_harness_v3_and_v3trivialmotions_joint/checkpoint-20000/experiment_cfg/metadata.json +467 -0
  3. cable_harness_v3_and_v3trivialmotions_joint/checkpoint-20000/model.safetensors +3 -0
  4. cable_harness_v3_and_v3trivialmotions_joint/checkpoint-20000/optimizer.pt +3 -0
  5. cable_harness_v3_and_v3trivialmotions_joint/checkpoint-20000/rng_state.pth +3 -0
  6. cable_harness_v3_and_v3trivialmotions_joint/checkpoint-20000/scheduler.pt +3 -0
  7. cable_harness_v3_and_v3trivialmotions_joint/checkpoint-20000/trainer_state.json +0 -0
cable_harness_v3_and_v3trivialmotions_joint/checkpoint-20000/config.json ADDED
@@ -0,0 +1,79 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action_dim": 32,
3
+ "action_head_cfg": {
4
+ "_convert_": "object",
5
+ "_target_": "gr00t.model.action_head.flow_matching_action_head_idm.FlowMatchingActionHeadIDM",
6
+ "config": {
7
+ "_recursive_": false,
8
+ "_target_": "gr00t.model.action_head.flow_matching_action_head_idm.FlowMatchingActionHeadIDMConfig",
9
+ "action_dim": 32,
10
+ "action_horizon": 16,
11
+ "add_pos_embed": true,
12
+ "add_seperator_token": true,
13
+ "add_view_embed": true,
14
+ "backbone_features_projector_cfg": null,
15
+ "diffusion_model_cfg": {
16
+ "_target_": "gr00t.model.action_head.cross_attention_dit.DiT",
17
+ "attention_head_dim": 64,
18
+ "dropout": 0.2,
19
+ "final_dropout": true,
20
+ "interleave_self_attention": true,
21
+ "norm_type": "ada_norm",
22
+ "num_attention_heads": 16,
23
+ "num_layers": 8,
24
+ "output_dim": 1024,
25
+ "positional_embeddings": null
26
+ },
27
+ "hidden_size": 1024,
28
+ "max_action_dim": 32,
29
+ "max_num_views": 6,
30
+ "max_state_dim": 64,
31
+ "mm_projector_cfg": {
32
+ "_convert_": "object",
33
+ "_target_": "gr00t.model.action_head.multimodal_projector.MultimodalProjector",
34
+ "config": {
35
+ "_target_": "gr00t.model.action_head.multimodal_projector.MultimodalProjectorConfig",
36
+ "hidden_size": 1024,
37
+ "mm_hidden_size": 1024,
38
+ "mm_projector_type": "mlp_doubledownsample"
39
+ }
40
+ },
41
+ "mm_vision_select_layer": -2,
42
+ "model_dtype": "float32",
43
+ "noise_beta_alpha": 1.5,
44
+ "noise_beta_beta": 1.0,
45
+ "noise_s": 0.999,
46
+ "num_inference_timesteps": 16,
47
+ "num_timestep_buckets": 1000,
48
+ "siglip_hidden_size": 1024,
49
+ "siglip_model_cfg": {
50
+ "_convert_": "object",
51
+ "_target_": "gr00t.model.action_head.siglip.SiglipModel.from_pretrained",
52
+ "pretrained_model_name_or_path": "google/siglip2-large-patch16-256"
53
+ },
54
+ "tune_vision_tower": true,
55
+ "vl_self_attention_cfg": {
56
+ "_target_": "gr00t.model.action_head.cross_attention_dit.SelfAttentionTransformer",
57
+ "attention_head_dim": 64,
58
+ "dropout": 0.2,
59
+ "final_dropout": true,
60
+ "num_attention_heads": 16,
61
+ "num_layers": 4,
62
+ "positional_embeddings": null
63
+ }
64
+ }
65
+ },
66
+ "action_horizon": 16,
67
+ "architectures": [
68
+ "IDM"
69
+ ],
70
+ "backbone_cfg": {
71
+ "_target_": "gr00t.model.backbone.IdentityBackbone"
72
+ },
73
+ "compute_dtype": "bfloat16",
74
+ "hidden_size": 0,
75
+ "model_dtype": "float32",
76
+ "model_type": "idm",
77
+ "torch_dtype": "float32",
78
+ "transformers_version": "4.46.2"
79
+ }
cable_harness_v3_and_v3trivialmotions_joint/checkpoint-20000/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,467 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "arm_l": {
6
+ "max": [
7
+ -0.1832655966281891,
8
+ 1.34694242477417,
9
+ 1.07246732711792,
10
+ 2.390453338623047,
11
+ 3.2724924087524414,
12
+ 3.3225512504577637
13
+ ],
14
+ "min": [
15
+ -2.411808967590332,
16
+ -0.47975248098373413,
17
+ -0.8484140634536743,
18
+ -2.4320755004882812,
19
+ -0.38252368569374084,
20
+ -1.6488759517669678
21
+ ],
22
+ "mean": [
23
+ -1.2004443407058716,
24
+ 0.2307078242301941,
25
+ -0.27301621437072754,
26
+ -1.3263508081436157,
27
+ 0.847358226776123,
28
+ 1.742966651916504
29
+ ],
30
+ "std": [
31
+ 0.5402601361274719,
32
+ 0.2648753523826599,
33
+ 0.24536585807800293,
34
+ 0.6965538263320923,
35
+ 0.3979998528957367,
36
+ 0.7569601535797119
37
+ ],
38
+ "q01": [
39
+ -2.1296116304397583,
40
+ -0.19493274718523027,
41
+ -0.6763621592521667,
42
+ -2.2673933839797975,
43
+ -0.15670432329177855,
44
+ -0.4060713160037994
45
+ ],
46
+ "q99": [
47
+ -0.2636143285036086,
48
+ 1.0812226271629353,
49
+ 0.6231017553806312,
50
+ 0.32392561435699463,
51
+ 1.6002621102333088,
52
+ 2.8172222423553475
53
+ ]
54
+ },
55
+ "arm_l_finger": {
56
+ "max": [
57
+ 1.0
58
+ ],
59
+ "min": [
60
+ 0.0
61
+ ],
62
+ "mean": [
63
+ 0.5408454537391663
64
+ ],
65
+ "std": [
66
+ 0.49850842356681824
67
+ ],
68
+ "q01": [
69
+ 0.0
70
+ ],
71
+ "q99": [
72
+ 1.0
73
+ ]
74
+ },
75
+ "arm_r": {
76
+ "max": [
77
+ 2.2549219131469727,
78
+ -0.06100641191005707,
79
+ -0.10050641745328903,
80
+ -0.3156932592391968,
81
+ 1.6240766048431396,
82
+ -0.17947575449943542
83
+ ],
84
+ "min": [
85
+ 2.2549068927764893,
86
+ -0.06100641191005707,
87
+ -0.10050641745328903,
88
+ -0.31574440002441406,
89
+ 1.6240766048431396,
90
+ -0.17947575449943542
91
+ ],
92
+ "mean": [
93
+ 2.254845142364502,
94
+ -0.06103173643350601,
95
+ -0.10055934637784958,
96
+ -0.3158548176288605,
97
+ 1.624868392944336,
98
+ -0.17961634695529938
99
+ ],
100
+ "std": [
101
+ 7.68112950026989e-05,
102
+ 2.5321282009826973e-05,
103
+ 5.295541268424131e-05,
104
+ 0.00012857951514888555,
105
+ 0.0007917048060335219,
106
+ 0.00014064654533285648
107
+ ],
108
+ "q01": [
109
+ 2.2549219131469727,
110
+ -0.06100641191005707,
111
+ -0.10050641745328903,
112
+ -0.31574440002441406,
113
+ 1.6240766048431396,
114
+ -0.17947575449943542
115
+ ],
116
+ "q99": [
117
+ 2.2549219131469727,
118
+ -0.06100641191005707,
119
+ -0.10050641745328903,
120
+ -0.3156932592391968,
121
+ 1.6240766048431396,
122
+ -0.17947575449943542
123
+ ]
124
+ },
125
+ "arm_r_finger": {
126
+ "max": [
127
+ 1.0
128
+ ],
129
+ "min": [
130
+ 1.0
131
+ ],
132
+ "mean": [
133
+ 1.0
134
+ ],
135
+ "std": [
136
+ 0.0
137
+ ],
138
+ "q01": [
139
+ 1.0
140
+ ],
141
+ "q99": [
142
+ 1.0
143
+ ]
144
+ },
145
+ "stage": {
146
+ "max": [
147
+ 0.0,
148
+ 0.0,
149
+ 0.0
150
+ ],
151
+ "min": [
152
+ -9.999999747378752e-06,
153
+ -0.0004299999854993075,
154
+ 0.0
155
+ ],
156
+ "mean": [
157
+ -4.223757969157305e-06,
158
+ -0.0003598667390178889,
159
+ 0.0
160
+ ],
161
+ "std": [
162
+ 4.9396171561966185e-06,
163
+ 3.417801417526789e-05,
164
+ 0.0
165
+ ],
166
+ "q01": [
167
+ -9.999999747378752e-06,
168
+ -0.0004199999966658652,
169
+ 0.0
170
+ ],
171
+ "q99": [
172
+ 0.0,
173
+ -0.00033999999868683517,
174
+ 0.0
175
+ ]
176
+ }
177
+ },
178
+ "action": {
179
+ "arm_l": {
180
+ "max": [
181
+ -0.18317537009716034,
182
+ 1.34694242477417,
183
+ 1.07246732711792,
184
+ 2.390453338623047,
185
+ 3.2724924087524414,
186
+ 3.3225512504577637
187
+ ],
188
+ "min": [
189
+ -2.411808967590332,
190
+ -0.47975248098373413,
191
+ -0.8484140634536743,
192
+ -2.4320755004882812,
193
+ -0.38344407081604004,
194
+ -1.6488759517669678
195
+ ],
196
+ "mean": [
197
+ -1.200505018234253,
198
+ 0.23155593872070312,
199
+ -0.27247053384780884,
200
+ -1.3274281024932861,
201
+ 0.8463924527168274,
202
+ 1.7448201179504395
203
+ ],
204
+ "std": [
205
+ 0.5405924320220947,
206
+ 0.2653326094150543,
207
+ 0.24588581919670105,
208
+ 0.6955525875091553,
209
+ 0.3985750675201416,
210
+ 0.7555415630340576
211
+ ],
212
+ "q01": [
213
+ -2.129775810241699,
214
+ -0.19481402218341826,
215
+ -0.6762284028530121,
216
+ -2.2680418491363525,
217
+ -0.15798263937234877,
218
+ -0.4027722179889679
219
+ ],
220
+ "q99": [
221
+ -0.26309458076953796,
222
+ 1.0829287481307985,
223
+ 0.6233379781246187,
224
+ 0.32392561435699463,
225
+ 1.5982361912727363,
226
+ 2.8186376190185554
227
+ ]
228
+ },
229
+ "arm_l_finger": {
230
+ "max": [
231
+ 1.0
232
+ ],
233
+ "min": [
234
+ 0.0
235
+ ],
236
+ "mean": [
237
+ 0.540309488773346
238
+ ],
239
+ "std": [
240
+ 0.49822431802749634
241
+ ],
242
+ "q01": [
243
+ 0.0
244
+ ],
245
+ "q99": [
246
+ 1.0
247
+ ]
248
+ },
249
+ "arm_r": {
250
+ "max": [
251
+ 2.2549219131469727,
252
+ -0.06100641191005707,
253
+ -0.10050641745328903,
254
+ -0.3156932592391968,
255
+ 1.6240766048431396,
256
+ -0.17947575449943542
257
+ ],
258
+ "min": [
259
+ 2.2549068927764893,
260
+ -0.06100641191005707,
261
+ -0.10050641745328903,
262
+ -0.31574440002441406,
263
+ 1.6240766048431396,
264
+ -0.17947575449943542
265
+ ],
266
+ "mean": [
267
+ 2.254845142364502,
268
+ -0.06103173643350601,
269
+ -0.10055934637784958,
270
+ -0.3158548176288605,
271
+ 1.624868392944336,
272
+ -0.17961634695529938
273
+ ],
274
+ "std": [
275
+ 7.68112950026989e-05,
276
+ 2.5321282009826973e-05,
277
+ 5.295541268424131e-05,
278
+ 0.00012857951514888555,
279
+ 0.0007917048060335219,
280
+ 0.00014064654533285648
281
+ ],
282
+ "q01": [
283
+ 2.2549219131469727,
284
+ -0.06100641191005707,
285
+ -0.10050641745328903,
286
+ -0.31574440002441406,
287
+ 1.6240766048431396,
288
+ -0.17947575449943542
289
+ ],
290
+ "q99": [
291
+ 2.2549219131469727,
292
+ -0.06100641191005707,
293
+ -0.10050641745328903,
294
+ -0.3156932592391968,
295
+ 1.6240766048431396,
296
+ -0.17947575449943542
297
+ ]
298
+ },
299
+ "arm_r_finger": {
300
+ "max": [
301
+ 1.0
302
+ ],
303
+ "min": [
304
+ 1.0
305
+ ],
306
+ "mean": [
307
+ 1.0
308
+ ],
309
+ "std": [
310
+ 0.0
311
+ ],
312
+ "q01": [
313
+ 1.0
314
+ ],
315
+ "q99": [
316
+ 1.0
317
+ ]
318
+ },
319
+ "stage": {
320
+ "max": [
321
+ 0.0,
322
+ 0.0,
323
+ 0.0
324
+ ],
325
+ "min": [
326
+ -9.999999747378752e-06,
327
+ -0.0004299999854993075,
328
+ 0.0
329
+ ],
330
+ "mean": [
331
+ -4.223892119625816e-06,
332
+ -0.0003598667390178889,
333
+ 0.0
334
+ ],
335
+ "std": [
336
+ 4.939618520438671e-06,
337
+ 3.417801417526789e-05,
338
+ 0.0
339
+ ],
340
+ "q01": [
341
+ -9.999999747378752e-06,
342
+ -0.0004199999966658652,
343
+ 0.0
344
+ ],
345
+ "q99": [
346
+ 0.0,
347
+ -0.00033999999868683517,
348
+ 0.0
349
+ ]
350
+ }
351
+ }
352
+ },
353
+ "modalities": {
354
+ "video": {
355
+ "cam_high": {
356
+ "resolution": [
357
+ 224,
358
+ 224
359
+ ],
360
+ "channels": 3,
361
+ "fps": 15.0
362
+ },
363
+ "cam_left_wrist": {
364
+ "resolution": [
365
+ 224,
366
+ 224
367
+ ],
368
+ "channels": 3,
369
+ "fps": 15.0
370
+ },
371
+ "cam_side": {
372
+ "resolution": [
373
+ 224,
374
+ 224
375
+ ],
376
+ "channels": 3,
377
+ "fps": 15.0
378
+ }
379
+ },
380
+ "state": {
381
+ "arm_l": {
382
+ "absolute": true,
383
+ "rotation_type": null,
384
+ "shape": [
385
+ 6
386
+ ],
387
+ "continuous": true
388
+ },
389
+ "arm_l_finger": {
390
+ "absolute": true,
391
+ "rotation_type": null,
392
+ "shape": [
393
+ 1
394
+ ],
395
+ "continuous": true
396
+ },
397
+ "arm_r": {
398
+ "absolute": true,
399
+ "rotation_type": null,
400
+ "shape": [
401
+ 6
402
+ ],
403
+ "continuous": true
404
+ },
405
+ "arm_r_finger": {
406
+ "absolute": true,
407
+ "rotation_type": null,
408
+ "shape": [
409
+ 1
410
+ ],
411
+ "continuous": true
412
+ },
413
+ "stage": {
414
+ "absolute": true,
415
+ "rotation_type": null,
416
+ "shape": [
417
+ 3
418
+ ],
419
+ "continuous": true
420
+ }
421
+ },
422
+ "action": {
423
+ "arm_l": {
424
+ "absolute": true,
425
+ "rotation_type": null,
426
+ "shape": [
427
+ 6
428
+ ],
429
+ "continuous": true
430
+ },
431
+ "arm_l_finger": {
432
+ "absolute": true,
433
+ "rotation_type": null,
434
+ "shape": [
435
+ 1
436
+ ],
437
+ "continuous": true
438
+ },
439
+ "arm_r": {
440
+ "absolute": true,
441
+ "rotation_type": null,
442
+ "shape": [
443
+ 6
444
+ ],
445
+ "continuous": true
446
+ },
447
+ "arm_r_finger": {
448
+ "absolute": true,
449
+ "rotation_type": null,
450
+ "shape": [
451
+ 1
452
+ ],
453
+ "continuous": true
454
+ },
455
+ "stage": {
456
+ "absolute": true,
457
+ "rotation_type": null,
458
+ "shape": [
459
+ 3
460
+ ],
461
+ "continuous": true
462
+ }
463
+ }
464
+ },
465
+ "embodiment_tag": "new_embodiment"
466
+ }
467
+ }
cable_harness_v3_and_v3trivialmotions_joint/checkpoint-20000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0864f561d33c9c5f0b8e7522abc752c58db6200c36fa783a1f56228ec8f2644e
3
+ size 2549650952
cable_harness_v3_and_v3trivialmotions_joint/checkpoint-20000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ed7f63384980188a1e313789a0cc7ea1d88d1281082958b7bb49bdae3b2e8d6e
3
+ size 4898088807
cable_harness_v3_and_v3trivialmotions_joint/checkpoint-20000/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bab03c93e25fba681070fac6dba362629c8c4e90d515ab1e70d08dfc2d57624f
3
+ size 14244
cable_harness_v3_and_v3trivialmotions_joint/checkpoint-20000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:514333fcc99a00a5ac2f0aadb8529bebf896b0c8433955834346694f8797b343
3
+ size 1064
cable_harness_v3_and_v3trivialmotions_joint/checkpoint-20000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff