Hugging Face's logo

EbotsVLA
/
IDM

Safetensors
Model card Files
xet
 Community
gayatrid-ebots commited on
Commit
828608d
·
verified ·
1 Parent(s): 78f5277

Uploaded a new model: cable_harness_v3_and_v3trivialmotions_cart checkpoint-20000

Browse files
Files changed (7) hide show
  1. cable_harness_v3_and_v3trivialmotions_cart/checkpoint-20000/config.json +79 -0
  2. cable_harness_v3_and_v3trivialmotions_cart/checkpoint-20000/experiment_cfg/metadata.json +467 -0
  3. cable_harness_v3_and_v3trivialmotions_cart/checkpoint-20000/model.safetensors +3 -0
  4. cable_harness_v3_and_v3trivialmotions_cart/checkpoint-20000/optimizer.pt +3 -0
  5. cable_harness_v3_and_v3trivialmotions_cart/checkpoint-20000/rng_state.pth +3 -0
  6. cable_harness_v3_and_v3trivialmotions_cart/checkpoint-20000/scheduler.pt +3 -0
  7. cable_harness_v3_and_v3trivialmotions_cart/checkpoint-20000/trainer_state.json +0 -0
cable_harness_v3_and_v3trivialmotions_cart/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_cart/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
+ 0.39572378993034363,
8
+ 0.06153039261698723,
9
+ 0.21977785229682922,
10
+ 2.98593807220459,
11
+ 2.118591785430908,
12
+ 2.164052724838257
13
+ ],
14
+ "min": [
15
+ -0.1006389930844307,
16
+ -0.3866279125213623,
17
+ -0.08715233206748962,
18
+ -2.9687929153442383,
19
+ -2.4328651428222656,
20
+ -2.165515184402466
21
+ ],
22
+ "mean": [
23
+ 0.18759986758232117,
24
+ -0.2185787558555603,
25
+ 0.0021240366622805595,
26
+ 0.5923742651939392,
27
+ -0.3791689872741699,
28
+ 0.23210103809833527
29
+ ],
30
+ "std": [
31
+ 0.12854869663715363,
32
+ 0.10552146285772324,
33
+ 0.04197557270526886,
34
+ 2.2150657176971436,
35
+ 0.6532875299453735,
36
+ 1.492536187171936
37
+ ],
38
+ "q01": [
39
+ -0.050147806853055955,
40
+ -0.36622276723384856,
41
+ -0.07304400220513343,
42
+ -2.601826858520508,
43
+ -1.8987161564826964,
44
+ -1.9595023703575134
45
+ ],
46
+ "q99": [
47
+ 0.36637878060340884,
48
+ -0.038548954948782764,
49
+ 0.10102849140763283,
50
+ 2.7363871192932128,
51
+ 0.9946757173538232,
52
+ 1.9522275614738467
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
+ -0.1352429986000061,
78
+ 0.16688400506973267,
79
+ 0.15862099826335907,
80
+ -1.0093860626220703,
81
+ -1.2007321119308472,
82
+ -0.3215712606906891
83
+ ],
84
+ "min": [
85
+ -0.1352429986000061,
86
+ 0.16688400506973267,
87
+ 0.15862099826335907,
88
+ -1.0093860626220703,
89
+ -1.2007321119308472,
90
+ -0.3215712606906891
91
+ ],
92
+ "mean": [
93
+ -0.13516156375408173,
94
+ 0.16697241365909576,
95
+ 0.15852481126785278,
96
+ -1.008612036705017,
97
+ -1.2009369134902954,
98
+ -0.3217256963253021
99
+ ],
100
+ "std": [
101
+ 8.145133324433118e-05,
102
+ 8.842043462209404e-05,
103
+ 9.620140917832032e-05,
104
+ 0.0007742791785858572,
105
+ 0.00020475195196922868,
106
+ 0.00015446406905539334
107
+ ],
108
+ "q01": [
109
+ -0.1352429986000061,
110
+ 0.16688400506973267,
111
+ 0.15862099826335907,
112
+ -1.0093860626220703,
113
+ -1.2007321119308472,
114
+ -0.3215712606906891
115
+ ],
116
+ "q99": [
117
+ -0.1352429986000061,
118
+ 0.16688400506973267,
119
+ 0.15862099826335907,
120
+ -1.0093860626220703,
121
+ -1.2007321119308472,
122
+ -0.3215712606906891
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
+ 0.39572378993034363,
182
+ 0.06153039261698723,
183
+ 0.21977785229682922,
184
+ 2.98593807220459,
185
+ 2.118591785430908,
186
+ 2.164052724838257
187
+ ],
188
+ "min": [
189
+ -0.1001979187130928,
190
+ -0.3866279125213623,
191
+ -0.08715233206748962,
192
+ -2.9687929153442383,
193
+ -2.4328651428222656,
194
+ -2.165515184402466
195
+ ],
196
+ "mean": [
197
+ 0.1877501755952835,
198
+ -0.2185070514678955,
199
+ 0.0020799352787435055,
200
+ 0.592729389667511,
201
+ -0.3777087330818176,
202
+ 0.23274624347686768
203
+ ],
204
+ "std": [
205
+ 0.1284850537776947,
206
+ 0.1055026650428772,
207
+ 0.04197055101394653,
208
+ 2.2154335975646973,
209
+ 0.6524918079376221,
210
+ 1.4930609464645386
211
+ ],
212
+ "q01": [
213
+ -0.0500591865926981,
214
+ -0.3662228584289551,
215
+ -0.07303985461592674,
216
+ -2.601826858520508,
217
+ -1.8931019878387452,
218
+ -1.959519124031067
219
+ ],
220
+ "q99": [
221
+ 0.3667649745941163,
222
+ -0.038420532345771786,
223
+ 0.10093420028686528,
224
+ 2.7363871192932128,
225
+ 0.9946757173538232,
226
+ 1.9528815174102787
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
+ -0.1352429986000061,
252
+ 0.16688400506973267,
253
+ 0.15862099826335907,
254
+ -1.0093860626220703,
255
+ -1.2007321119308472,
256
+ -0.3215712606906891
257
+ ],
258
+ "min": [
259
+ -0.1352429986000061,
260
+ 0.16688400506973267,
261
+ 0.15862099826335907,
262
+ -1.0093860626220703,
263
+ -1.2007321119308472,
264
+ -0.3215712606906891
265
+ ],
266
+ "mean": [
267
+ -0.13516156375408173,
268
+ 0.16697241365909576,
269
+ 0.15852481126785278,
270
+ -1.008612036705017,
271
+ -1.2009369134902954,
272
+ -0.3217256963253021
273
+ ],
274
+ "std": [
275
+ 8.145133324433118e-05,
276
+ 8.842043462209404e-05,
277
+ 9.620140917832032e-05,
278
+ 0.0007742791785858572,
279
+ 0.00020475195196922868,
280
+ 0.00015446406905539334
281
+ ],
282
+ "q01": [
283
+ -0.1352429986000061,
284
+ 0.16688400506973267,
285
+ 0.15862099826335907,
286
+ -1.0093860626220703,
287
+ -1.2007321119308472,
288
+ -0.3215712606906891
289
+ ],
290
+ "q99": [
291
+ -0.1352429986000061,
292
+ 0.16688400506973267,
293
+ 0.15862099826335907,
294
+ -1.0093860626220703,
295
+ -1.2007321119308472,
296
+ -0.3215712606906891
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_cart/checkpoint-20000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:325a1942e83f4824b4d553b1c498835a45d84e8e9209fd88f70eec91c74ba5f6
3
+ size 2549650952
cable_harness_v3_and_v3trivialmotions_cart/checkpoint-20000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e2a5cb24f400d77a18c683149bb1a28151ed461d87bde0860e5f8b370af92bf3
3
+ size 4898088807
cable_harness_v3_and_v3trivialmotions_cart/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_cart/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_cart/checkpoint-20000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff