Hugging Face's logo

rail-berkeley
/
crossformer

Robotics
crossformer
Model card Files Community
2
rail-berkeley commited on
Commit
bd5ed29
•
1 Parent(s): 15910f0

Upload config.json

Browse files
Files changed (1) hide show
  1. config.json +1886 -0
config.json ADDED
@@ -0,0 +1,1886 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset_kwargs": {
3
+ "balance_weights": false,
4
+ "batch_size": 8,
5
+ "dataset_kwargs_list": [
6
+ {
7
+ "name": "kuka",
8
+ "data_dir": null,
9
+ "image_obs_keys": {
10
+ "primary": "image",
11
+ "high": null,
12
+ "nav": null,
13
+ "left_wrist": null,
14
+ "right_wrist": null
15
+ },
16
+ "proprio_obs_keys": {
17
+ "bimanual": null,
18
+ "quadruped": null
19
+ },
20
+ "proprio_obs_dims": {
21
+ "bimanual": 14,
22
+ "quadruped": 59
23
+ },
24
+ "action_normalization_mask": [
25
+ true,
26
+ true,
27
+ true,
28
+ true,
29
+ true,
30
+ true,
31
+ false
32
+ ],
33
+ "language_key": "language_instruction",
34
+ "action_proprio_normalization_type": "normal",
35
+ "standardize_fn": {
36
+ "module": "crossformer.data.oxe.oxe_standardization_transforms",
37
+ "name": "kuka_dataset_transform",
38
+ "args": [],
39
+ "kwargs": {}
40
+ }
41
+ },
42
+ {
43
+ "name": "taco_play",
44
+ "data_dir": null,
45
+ "image_obs_keys": {
46
+ "primary": "rgb_static",
47
+ "high": null,
48
+ "nav": null,
49
+ "left_wrist": null,
50
+ "right_wrist": null
51
+ },
52
+ "proprio_obs_keys": {
53
+ "bimanual": null,
54
+ "quadruped": null
55
+ },
56
+ "proprio_obs_dims": {
57
+ "bimanual": 14,
58
+ "quadruped": 59
59
+ },
60
+ "action_normalization_mask": [
61
+ true,
62
+ true,
63
+ true,
64
+ true,
65
+ true,
66
+ true,
67
+ false
68
+ ],
69
+ "language_key": "language_instruction",
70
+ "action_proprio_normalization_type": "normal",
71
+ "standardize_fn": {
72
+ "module": "crossformer.data.oxe.oxe_standardization_transforms",
73
+ "name": "taco_dataset_transform",
74
+ "args": [],
75
+ "kwargs": {}
76
+ }
77
+ },
78
+ {
79
+ "name": "taco_extra",
80
+ "data_dir": null,
81
+ "image_obs_keys": {
82
+ "primary": "rgb_static",
83
+ "high": null,
84
+ "nav": null,
85
+ "left_wrist": null,
86
+ "right_wrist": null
87
+ },
88
+ "proprio_obs_keys": {
89
+ "bimanual": null,
90
+ "quadruped": null
91
+ },
92
+ "proprio_obs_dims": {
93
+ "bimanual": 14,
94
+ "quadruped": 59
95
+ },
96
+ "action_normalization_mask": [
97
+ true,
98
+ true,
99
+ true,
100
+ true,
101
+ true,
102
+ true,
103
+ false
104
+ ],
105
+ "language_key": "language_instruction",
106
+ "action_proprio_normalization_type": "normal",
107
+ "standardize_fn": {
108
+ "module": "crossformer.data.oxe.oxe_standardization_transforms",
109
+ "name": "taco_dataset_transform",
110
+ "args": [],
111
+ "kwargs": {}
112
+ }
113
+ },
114
+ {
115
+ "name": "jaco_play",
116
+ "data_dir": null,
117
+ "image_obs_keys": {
118
+ "primary": "image",
119
+ "high": null,
120
+ "nav": null,
121
+ "left_wrist": null,
122
+ "right_wrist": null
123
+ },
124
+ "proprio_obs_keys": {
125
+ "bimanual": null,
126
+ "quadruped": null
127
+ },
128
+ "proprio_obs_dims": {
129
+ "bimanual": 14,
130
+ "quadruped": 59
131
+ },
132
+ "action_normalization_mask": [
133
+ true,
134
+ true,
135
+ true,
136
+ true,
137
+ true,
138
+ true,
139
+ false
140
+ ],
141
+ "language_key": "language_instruction",
142
+ "action_proprio_normalization_type": "normal",
143
+ "standardize_fn": {
144
+ "module": "crossformer.data.oxe.oxe_standardization_transforms",
145
+ "name": "jaco_play_dataset_transform",
146
+ "args": [],
147
+ "kwargs": {}
148
+ }
149
+ },
150
+ {
151
+ "name": "berkeley_cable_routing",
152
+ "data_dir": null,
153
+ "image_obs_keys": {
154
+ "primary": "image",
155
+ "high": null,
156
+ "nav": null,
157
+ "left_wrist": null,
158
+ "right_wrist": null
159
+ },
160
+ "proprio_obs_keys": {
161
+ "bimanual": null,
162
+ "quadruped": null
163
+ },
164
+ "proprio_obs_dims": {
165
+ "bimanual": 14,
166
+ "quadruped": 59
167
+ },
168
+ "action_normalization_mask": [
169
+ true,
170
+ true,
171
+ true,
172
+ true,
173
+ true,
174
+ true,
175
+ false
176
+ ],
177
+ "language_key": "language_instruction",
178
+ "action_proprio_normalization_type": "normal",
179
+ "standardize_fn": {
180
+ "module": "crossformer.data.oxe.oxe_standardization_transforms",
181
+ "name": "berkeley_cable_routing_dataset_transform",
182
+ "args": [],
183
+ "kwargs": {}
184
+ }
185
+ },
186
+ {
187
+ "name": "roboturk",
188
+ "data_dir": null,
189
+ "image_obs_keys": {
190
+ "primary": "front_rgb",
191
+ "high": null,
192
+ "nav": null,
193
+ "left_wrist": null,
194
+ "right_wrist": null
195
+ },
196
+ "proprio_obs_keys": {
197
+ "bimanual": null,
198
+ "quadruped": null
199
+ },
200
+ "proprio_obs_dims": {
201
+ "bimanual": 14,
202
+ "quadruped": 59
203
+ },
204
+ "action_normalization_mask": [
205
+ true,
206
+ true,
207
+ true,
208
+ true,
209
+ true,
210
+ true,
211
+ false
212
+ ],
213
+ "language_key": "language_instruction",
214
+ "action_proprio_normalization_type": "normal",
215
+ "standardize_fn": {
216
+ "module": "crossformer.data.oxe.oxe_standardization_transforms",
217
+ "name": "roboturk_dataset_transform",
218
+ "args": [],
219
+ "kwargs": {}
220
+ }
221
+ },
222
+ {
223
+ "name": "viola",
224
+ "data_dir": null,
225
+ "image_obs_keys": {
226
+ "primary": "agentview_rgb",
227
+ "high": null,
228
+ "nav": null,
229
+ "left_wrist": null,
230
+ "right_wrist": null
231
+ },
232
+ "proprio_obs_keys": {
233
+ "bimanual": null,
234
+ "quadruped": null
235
+ },
236
+ "proprio_obs_dims": {
237
+ "bimanual": 14,
238
+ "quadruped": 59
239
+ },
240
+ "action_normalization_mask": [
241
+ true,
242
+ true,
243
+ true,
244
+ true,
245
+ true,
246
+ true,
247
+ false
248
+ ],
249
+ "language_key": "language_instruction",
250
+ "action_proprio_normalization_type": "normal",
251
+ "standardize_fn": {
252
+ "module": "crossformer.data.oxe.oxe_standardization_transforms",
253
+ "name": "viola_dataset_transform",
254
+ "args": [],
255
+ "kwargs": {}
256
+ }
257
+ },
258
+ {
259
+ "name": "berkeley_autolab_ur5",
260
+ "data_dir": null,
261
+ "image_obs_keys": {
262
+ "primary": "image",
263
+ "high": null,
264
+ "nav": null,
265
+ "left_wrist": null,
266
+ "right_wrist": null
267
+ },
268
+ "proprio_obs_keys": {
269
+ "bimanual": null,
270
+ "quadruped": null
271
+ },
272
+ "proprio_obs_dims": {
273
+ "bimanual": 14,
274
+ "quadruped": 59
275
+ },
276
+ "action_normalization_mask": [
277
+ true,
278
+ true,
279
+ true,
280
+ true,
281
+ true,
282
+ true,
283
+ false
284
+ ],
285
+ "language_key": "language_instruction",
286
+ "action_proprio_normalization_type": "normal",
287
+ "standardize_fn": {
288
+ "module": "crossformer.data.oxe.oxe_standardization_transforms",
289
+ "name": "berkeley_autolab_ur5_dataset_transform",
290
+ "args": [],
291
+ "kwargs": {}
292
+ }
293
+ },
294
+ {
295
+ "name": "toto",
296
+ "data_dir": null,
297
+ "image_obs_keys": {
298
+ "primary": "image",
299
+ "high": null,
300
+ "nav": null,
301
+ "left_wrist": null,
302
+ "right_wrist": null
303
+ },
304
+ "proprio_obs_keys": {
305
+ "bimanual": null,
306
+ "quadruped": null
307
+ },
308
+ "proprio_obs_dims": {
309
+ "bimanual": 14,
310
+ "quadruped": 59
311
+ },
312
+ "action_normalization_mask": [
313
+ true,
314
+ true,
315
+ true,
316
+ true,
317
+ true,
318
+ true,
319
+ false
320
+ ],
321
+ "language_key": "language_instruction",
322
+ "action_proprio_normalization_type": "normal",
323
+ "standardize_fn": {
324
+ "module": "crossformer.data.oxe.oxe_standardization_transforms",
325
+ "name": "toto_dataset_transform",
326
+ "args": [],
327
+ "kwargs": {}
328
+ }
329
+ },
330
+ {
331
+ "name": "language_table",
332
+ "data_dir": null,
333
+ "image_obs_keys": {
334
+ "primary": "rgb",
335
+ "high": null,
336
+ "nav": null,
337
+ "left_wrist": null,
338
+ "right_wrist": null
339
+ },
340
+ "proprio_obs_keys": {
341
+ "bimanual": null,
342
+ "quadruped": null
343
+ },
344
+ "proprio_obs_dims": {
345
+ "bimanual": 14,
346
+ "quadruped": 59
347
+ },
348
+ "action_normalization_mask": [
349
+ true,
350
+ true,
351
+ true,
352
+ true,
353
+ true,
354
+ true,
355
+ false
356
+ ],
357
+ "language_key": "language_instruction",
358
+ "action_proprio_normalization_type": "normal",
359
+ "standardize_fn": {
360
+ "module": "crossformer.data.oxe.oxe_standardization_transforms",
361
+ "name": "language_table_dataset_transform",
362
+ "args": [],
363
+ "kwargs": {}
364
+ }
365
+ },
366
+ {
367
+ "name": "stanford_hydra_dataset_converted_externally_to_rlds",
368
+ "data_dir": null,
369
+ "image_obs_keys": {
370
+ "primary": "image",
371
+ "high": null,
372
+ "nav": null,
373
+ "left_wrist": null,
374
+ "right_wrist": null
375
+ },
376
+ "proprio_obs_keys": {
377
+ "bimanual": null,
378
+ "quadruped": null
379
+ },
380
+ "proprio_obs_dims": {
381
+ "bimanual": 14,
382
+ "quadruped": 59
383
+ },
384
+ "action_normalization_mask": [
385
+ true,
386
+ true,
387
+ true,
388
+ true,
389
+ true,
390
+ true,
391
+ false
392
+ ],
393
+ "language_key": "language_instruction",
394
+ "action_proprio_normalization_type": "normal",
395
+ "standardize_fn": {
396
+ "module": "crossformer.data.oxe.oxe_standardization_transforms",
397
+ "name": "stanford_hydra_dataset_transform",
398
+ "args": [],
399
+ "kwargs": {}
400
+ }
401
+ },
402
+ {
403
+ "name": "austin_buds_dataset_converted_externally_to_rlds",
404
+ "data_dir": null,
405
+ "image_obs_keys": {
406
+ "primary": "image",
407
+ "high": null,
408
+ "nav": null,
409
+ "left_wrist": null,
410
+ "right_wrist": null
411
+ },
412
+ "proprio_obs_keys": {
413
+ "bimanual": null,
414
+ "quadruped": null
415
+ },
416
+ "proprio_obs_dims": {
417
+ "bimanual": 14,
418
+ "quadruped": 59
419
+ },
420
+ "action_normalization_mask": [
421
+ true,
422
+ true,
423
+ true,
424
+ true,
425
+ true,
426
+ true,
427
+ false
428
+ ],
429
+ "language_key": "language_instruction",
430
+ "action_proprio_normalization_type": "normal",
431
+ "standardize_fn": {
432
+ "module": "crossformer.data.oxe.oxe_standardization_transforms",
433
+ "name": "austin_buds_dataset_transform",
434
+ "args": [],
435
+ "kwargs": {}
436
+ }
437
+ },
438
+ {
439
+ "name": "nyu_franka_play_dataset_converted_externally_to_rlds",
440
+ "data_dir": null,
441
+ "image_obs_keys": {
442
+ "primary": "image",
443
+ "high": null,
444
+ "nav": null,
445
+ "left_wrist": null,
446
+ "right_wrist": null
447
+ },
448
+ "proprio_obs_keys": {
449
+ "bimanual": null,
450
+ "quadruped": null
451
+ },
452
+ "proprio_obs_dims": {
453
+ "bimanual": 14,
454
+ "quadruped": 59
455
+ },
456
+ "action_normalization_mask": [
457
+ true,
458
+ true,
459
+ true,
460
+ true,
461
+ true,
462
+ true,
463
+ false
464
+ ],
465
+ "language_key": "language_instruction",
466
+ "action_proprio_normalization_type": "normal",
467
+ "standardize_fn": {
468
+ "module": "crossformer.data.oxe.oxe_standardization_transforms",
469
+ "name": "nyu_franka_play_dataset_transform",
470
+ "args": [],
471
+ "kwargs": {}
472
+ }
473
+ },
474
+ {
475
+ "name": "furniture_bench_dataset_converted_externally_to_rlds",
476
+ "data_dir": null,
477
+ "image_obs_keys": {
478
+ "primary": "image",
479
+ "high": null,
480
+ "nav": null,
481
+ "left_wrist": null,
482
+ "right_wrist": null
483
+ },
484
+ "proprio_obs_keys": {
485
+ "bimanual": null,
486
+ "quadruped": null
487
+ },
488
+ "proprio_obs_dims": {
489
+ "bimanual": 14,
490
+ "quadruped": 59
491
+ },
492
+ "action_normalization_mask": [
493
+ true,
494
+ true,
495
+ true,
496
+ true,
497
+ true,
498
+ true,
499
+ false
500
+ ],
501
+ "language_key": "language_instruction",
502
+ "action_proprio_normalization_type": "normal",
503
+ "standardize_fn": {
504
+ "module": "crossformer.data.oxe.oxe_standardization_transforms",
505
+ "name": "furniture_bench_dataset_transform",
506
+ "args": [],
507
+ "kwargs": {}
508
+ }
509
+ },
510
+ {
511
+ "name": "austin_sailor_dataset_converted_externally_to_rlds",
512
+ "data_dir": null,
513
+ "image_obs_keys": {
514
+ "primary": "image",
515
+ "nav": null,
516
+ "high": null,
517
+ "left_wrist": null,
518
+ "right_wrist": null
519
+ },
520
+ "proprio_obs_keys": {
521
+ "bimanual": null,
522
+ "quadruped": null
523
+ },
524
+ "proprio_obs_dims": {
525
+ "bimanual": 14,
526
+ "quadruped": 59
527
+ },
528
+ "action_normalization_mask": [
529
+ true,
530
+ true,
531
+ true,
532
+ true,
533
+ true,
534
+ true,
535
+ false
536
+ ],
537
+ "language_key": "language_instruction",
538
+ "action_proprio_normalization_type": "normal",
539
+ "standardize_fn": {
540
+ "module": "crossformer.data.oxe.oxe_standardization_transforms",
541
+ "name": "austin_sailor_dataset_transform",
542
+ "args": [],
543
+ "kwargs": {}
544
+ }
545
+ },
546
+ {
547
+ "name": "austin_sirius_dataset_converted_externally_to_rlds",
548
+ "data_dir": null,
549
+ "image_obs_keys": {
550
+ "primary": "image",
551
+ "nav": null,
552
+ "high": null,
553
+ "left_wrist": null,
554
+ "right_wrist": null
555
+ },
556
+ "proprio_obs_keys": {
557
+ "bimanual": null,
558
+ "quadruped": null
559
+ },
560
+ "proprio_obs_dims": {
561
+ "bimanual": 14,
562
+ "quadruped": 59
563
+ },
564
+ "action_normalization_mask": [
565
+ true,
566
+ true,
567
+ true,
568
+ true,
569
+ true,
570
+ true,
571
+ false
572
+ ],
573
+ "language_key": "language_instruction",
574
+ "action_proprio_normalization_type": "normal",
575
+ "standardize_fn": {
576
+ "module": "crossformer.data.oxe.oxe_standardization_transforms",
577
+ "name": "austin_sirius_dataset_transform",
578
+ "args": [],
579
+ "kwargs": {}
580
+ }
581
+ },
582
+ {
583
+ "name": "bc_z",
584
+ "data_dir": null,
585
+ "image_obs_keys": {
586
+ "primary": "image",
587
+ "high": null,
588
+ "nav": null,
589
+ "left_wrist": null,
590
+ "right_wrist": null
591
+ },
592
+ "proprio_obs_keys": {
593
+ "bimanual": null,
594
+ "quadruped": null
595
+ },
596
+ "proprio_obs_dims": {
597
+ "bimanual": 14,
598
+ "quadruped": 59
599
+ },
600
+ "action_normalization_mask": [
601
+ true,
602
+ true,
603
+ true,
604
+ true,
605
+ true,
606
+ true,
607
+ false
608
+ ],
609
+ "language_key": "language_instruction",
610
+ "action_proprio_normalization_type": "normal",
611
+ "standardize_fn": {
612
+ "module": "crossformer.data.oxe.oxe_standardization_transforms",
613
+ "name": "bc_z_dataset_transform",
614
+ "args": [],
615
+ "kwargs": {}
616
+ }
617
+ },
618
+ {
619
+ "name": "dlr_edan_shared_control_converted_externally_to_rlds",
620
+ "data_dir": null,
621
+ "image_obs_keys": {
622
+ "primary": "image",
623
+ "high": null,
624
+ "nav": null,
625
+ "left_wrist": null,
626
+ "right_wrist": null
627
+ },
628
+ "proprio_obs_keys": {
629
+ "bimanual": null,
630
+ "quadruped": null
631
+ },
632
+ "proprio_obs_dims": {
633
+ "bimanual": 14,
634
+ "quadruped": 59
635
+ },
636
+ "action_normalization_mask": [
637
+ true,
638
+ true,
639
+ true,
640
+ true,
641
+ true,
642
+ true,
643
+ false
644
+ ],
645
+ "language_key": "language_instruction",
646
+ "action_proprio_normalization_type": "normal",
647
+ "standardize_fn": {
648
+ "module": "crossformer.data.oxe.oxe_standardization_transforms",
649
+ "name": "dlr_edan_shared_control_dataset_transform",
650
+ "args": [],
651
+ "kwargs": {}
652
+ }
653
+ },
654
+ {
655
+ "name": "iamlab_cmu_pickup_insert_converted_externally_to_rlds",
656
+ "data_dir": null,
657
+ "image_obs_keys": {
658
+ "primary": "image",
659
+ "high": null,
660
+ "nav": null,
661
+ "left_wrist": null,
662
+ "right_wrist": null
663
+ },
664
+ "proprio_obs_keys": {
665
+ "bimanual": null,
666
+ "quadruped": null
667
+ },
668
+ "proprio_obs_dims": {
669
+ "bimanual": 14,
670
+ "quadruped": 59
671
+ },
672
+ "action_normalization_mask": [
673
+ true,
674
+ true,
675
+ true,
676
+ true,
677
+ true,
678
+ true,
679
+ false
680
+ ],
681
+ "language_key": "language_instruction",
682
+ "action_proprio_normalization_type": "normal",
683
+ "standardize_fn": {
684
+ "module": "crossformer.data.oxe.oxe_standardization_transforms",
685
+ "name": "iamlab_pick_insert_dataset_transform",
686
+ "args": [],
687
+ "kwargs": {}
688
+ }
689
+ },
690
+ {
691
+ "name": "utaustin_mutex",
692
+ "data_dir": null,
693
+ "image_obs_keys": {
694
+ "primary": "image",
695
+ "high": null,
696
+ "nav": null,
697
+ "left_wrist": null,
698
+ "right_wrist": null
699
+ },
700
+ "proprio_obs_keys": {
701
+ "bimanual": null,
702
+ "quadruped": null
703
+ },
704
+ "proprio_obs_dims": {
705
+ "bimanual": 14,
706
+ "quadruped": 59
707
+ },
708
+ "action_normalization_mask": [
709
+ true,
710
+ true,
711
+ true,
712
+ true,
713
+ true,
714
+ true,
715
+ false
716
+ ],
717
+ "language_key": "language_instruction",
718
+ "action_proprio_normalization_type": "normal",
719
+ "standardize_fn": {
720
+ "module": "crossformer.data.oxe.oxe_standardization_transforms",
721
+ "name": "utaustin_mutex_dataset_transform",
722
+ "args": [],
723
+ "kwargs": {}
724
+ }
725
+ },
726
+ {
727
+ "name": "berkeley_fanuc_manipulation",
728
+ "data_dir": null,
729
+ "image_obs_keys": {
730
+ "primary": "image",
731
+ "high": null,
732
+ "nav": null,
733
+ "left_wrist": null,
734
+ "right_wrist": null
735
+ },
736
+ "proprio_obs_keys": {
737
+ "bimanual": null,
738
+ "quadruped": null
739
+ },
740
+ "proprio_obs_dims": {
741
+ "bimanual": 14,
742
+ "quadruped": 59
743
+ },
744
+ "action_normalization_mask": [
745
+ true,
746
+ true,
747
+ true,
748
+ true,
749
+ true,
750
+ true,
751
+ false
752
+ ],
753
+ "language_key": "language_instruction",
754
+ "action_proprio_normalization_type": "normal",
755
+ "standardize_fn": {
756
+ "module": "crossformer.data.oxe.oxe_standardization_transforms",
757
+ "name": "berkeley_fanuc_dataset_transform",
758
+ "args": [],
759
+ "kwargs": {}
760
+ }
761
+ },
762
+ {
763
+ "name": "cmu_stretch",
764
+ "data_dir": null,
765
+ "image_obs_keys": {
766
+ "primary": "image",
767
+ "high": null,
768
+ "nav": null,
769
+ "left_wrist": null,
770
+ "right_wrist": null
771
+ },
772
+ "proprio_obs_keys": {
773
+ "bimanual": null,
774
+ "quadruped": null
775
+ },
776
+ "proprio_obs_dims": {
777
+ "bimanual": 14,
778
+ "quadruped": 59
779
+ },
780
+ "action_normalization_mask": [
781
+ true,
782
+ true,
783
+ true,
784
+ true,
785
+ true,
786
+ true,
787
+ false
788
+ ],
789
+ "language_key": "language_instruction",
790
+ "action_proprio_normalization_type": "normal",
791
+ "standardize_fn": {
792
+ "module": "crossformer.data.oxe.oxe_standardization_transforms",
793
+ "name": "cmu_stretch_dataset_transform",
794
+ "args": [],
795
+ "kwargs": {}
796
+ }
797
+ },
798
+ {
799
+ "name": "droid",
800
+ "data_dir": null,
801
+ "image_obs_keys": {
802
+ "primary": "exterior_image_1_left",
803
+ "high": null,
804
+ "nav": null,
805
+ "left_wrist": null,
806
+ "right_wrist": null
807
+ },
808
+ "proprio_obs_keys": {
809
+ "bimanual": null,
810
+ "quadruped": null
811
+ },
812
+ "proprio_obs_dims": {
813
+ "bimanual": 14,
814
+ "quadruped": 59
815
+ },
816
+ "filter_functions": [
817
+ {
818
+ "module": "experiments.homer.scripts.configs.cross_embodiment",
819
+ "name": "filter_success",
820
+ "args": [],
821
+ "kwargs": {}
822
+ }
823
+ ],
824
+ "action_normalization_mask": [
825
+ true,
826
+ true,
827
+ true,
828
+ true,
829
+ true,
830
+ true,
831
+ false
832
+ ],
833
+ "language_key": "language_instruction",
834
+ "action_proprio_normalization_type": "normal",
835
+ "standardize_fn": {
836
+ "module": "crossformer.data.oxe.oxe_standardization_transforms",
837
+ "name": "droid_dataset_transform",
838
+ "args": [],
839
+ "kwargs": {}
840
+ }
841
+ },
842
+ {
843
+ "name": "aloha_pen_uncap_diverse_dataset",
844
+ "data_dir": null,
845
+ "image_obs_keys": {
846
+ "primary": null,
847
+ "high": "cam_high",
848
+ "nav": null,
849
+ "left_wrist": "cam_left_wrist",
850
+ "right_wrist": "cam_right_wrist"
851
+ },
852
+ "proprio_obs_keys": {
853
+ "bimanual": "proprio",
854
+ "quadruped": null
855
+ },
856
+ "proprio_obs_dims": {
857
+ "bimanual": 14,
858
+ "quadruped": 59
859
+ },
860
+ "action_normalization_mask": [
861
+ true,
862
+ true,
863
+ true,
864
+ true,
865
+ true,
866
+ true,
867
+ false,
868
+ true,
869
+ true,
870
+ true,
871
+ true,
872
+ true,
873
+ true,
874
+ false
875
+ ],
876
+ "language_key": "language_instruction",
877
+ "action_proprio_normalization_type": "normal",
878
+ "standardize_fn": {
879
+ "module": "crossformer.data.oxe.oxe_standardization_transforms",
880
+ "name": "aloha_pen_uncap_dataset_transform",
881
+ "args": [],
882
+ "kwargs": {}
883
+ }
884
+ },
885
+ {
886
+ "name": "aloha_dough_cut_dataset",
887
+ "data_dir": null,
888
+ "image_obs_keys": {
889
+ "primary": null,
890
+ "high": "cam_high",
891
+ "nav": null,
892
+ "left_wrist": "cam_left_wrist",
893
+ "right_wrist": "cam_right_wrist"
894
+ },
895
+ "proprio_obs_keys": {
896
+ "bimanual": "proprio",
897
+ "quadruped": null
898
+ },
899
+ "proprio_obs_dims": {
900
+ "bimanual": 14,
901
+ "quadruped": 59
902
+ },
903
+ "action_normalization_mask": [
904
+ true,
905
+ true,
906
+ true,
907
+ true,
908
+ true,
909
+ true,
910
+ false,
911
+ true,
912
+ true,
913
+ true,
914
+ true,
915
+ true,
916
+ true,
917
+ false
918
+ ],
919
+ "language_key": "language_instruction",
920
+ "action_proprio_normalization_type": "normal",
921
+ "standardize_fn": {
922
+ "module": "crossformer.data.oxe.oxe_standardization_transforms",
923
+ "name": "aloha_dough_dataset_transform",
924
+ "args": [],
925
+ "kwargs": {}
926
+ }
927
+ },
928
+ {
929
+ "name": "aloha_lucy_dataset",
930
+ "data_dir": null,
931
+ "image_obs_keys": {
932
+ "primary": null,
933
+ "high": "cam_high",
934
+ "nav": null,
935
+ "left_wrist": "cam_left_wrist",
936
+ "right_wrist": "cam_right_wrist"
937
+ },
938
+ "proprio_obs_keys": {
939
+ "bimanual": "proprio",
940
+ "quadruped": null
941
+ },
942
+ "proprio_obs_dims": {
943
+ "bimanual": 14,
944
+ "quadruped": 59
945
+ },
946
+ "action_normalization_mask": [
947
+ true,
948
+ true,
949
+ true,
950
+ true,
951
+ true,
952
+ true,
953
+ false,
954
+ true,
955
+ true,
956
+ true,
957
+ true,
958
+ true,
959
+ true,
960
+ false
961
+ ],
962
+ "language_key": "language_instruction",
963
+ "action_proprio_normalization_type": "normal",
964
+ "standardize_fn": {
965
+ "module": "crossformer.data.oxe.oxe_standardization_transforms",
966
+ "name": "aloha_dough_dataset_transform",
967
+ "args": [],
968
+ "kwargs": {}
969
+ }
970
+ },
971
+ {
972
+ "name": "aloha_drawer_dataset",
973
+ "data_dir": null,
974
+ "image_obs_keys": {
975
+ "primary": null,
976
+ "high": "cam_high",
977
+ "nav": null,
978
+ "left_wrist": "cam_left_wrist",
979
+ "right_wrist": "cam_right_wrist"
980
+ },
981
+ "proprio_obs_keys": {
982
+ "bimanual": "proprio",
983
+ "quadruped": null
984
+ },
985
+ "proprio_obs_dims": {
986
+ "bimanual": 14,
987
+ "quadruped": 59
988
+ },
989
+ "action_normalization_mask": [
990
+ true,
991
+ true,
992
+ true,
993
+ true,
994
+ true,
995
+ true,
996
+ false,
997
+ true,
998
+ true,
999
+ true,
1000
+ true,
1001
+ true,
1002
+ true,
1003
+ false
1004
+ ],
1005
+ "language_key": "language_instruction",
1006
+ "action_proprio_normalization_type": "normal",
1007
+ "standardize_fn": {
1008
+ "module": "crossformer.data.oxe.oxe_standardization_transforms",
1009
+ "name": "aloha_dough_dataset_transform",
1010
+ "args": [],
1011
+ "kwargs": {}
1012
+ }
1013
+ },
1014
+ {
1015
+ "name": "aloha_pick_place_dataset",
1016
+ "data_dir": null,
1017
+ "image_obs_keys": {
1018
+ "primary": null,
1019
+ "high": "cam_high",
1020
+ "nav": null,
1021
+ "left_wrist": "cam_left_wrist",
1022
+ "right_wrist": "cam_right_wrist"
1023
+ },
1024
+ "proprio_obs_keys": {
1025
+ "bimanual": "proprio",
1026
+ "quadruped": null
1027
+ },
1028
+ "proprio_obs_dims": {
1029
+ "bimanual": 14,
1030
+ "quadruped": 59
1031
+ },
1032
+ "action_normalization_mask": [
1033
+ true,
1034
+ true,
1035
+ true,
1036
+ true,
1037
+ true,
1038
+ true,
1039
+ false,
1040
+ true,
1041
+ true,
1042
+ true,
1043
+ true,
1044
+ true,
1045
+ true,
1046
+ false
1047
+ ],
1048
+ "language_key": "language_instruction",
1049
+ "action_proprio_normalization_type": "normal",
1050
+ "standardize_fn": {
1051
+ "module": "crossformer.data.oxe.oxe_standardization_transforms",
1052
+ "name": "aloha_dough_dataset_transform",
1053
+ "args": [],
1054
+ "kwargs": {}
1055
+ }
1056
+ },
1057
+ {
1058
+ "name": "aloha_static_dataset",
1059
+ "data_dir": null,
1060
+ "image_obs_keys": {
1061
+ "primary": null,
1062
+ "high": "cam_high",
1063
+ "nav": null,
1064
+ "left_wrist": "cam_left_wrist",
1065
+ "right_wrist": "cam_right_wrist"
1066
+ },
1067
+ "proprio_obs_keys": {
1068
+ "bimanual": "proprio",
1069
+ "quadruped": null
1070
+ },
1071
+ "proprio_obs_dims": {
1072
+ "bimanual": 14,
1073
+ "quadruped": 59
1074
+ },
1075
+ "action_normalization_mask": [
1076
+ true,
1077
+ true,
1078
+ true,
1079
+ true,
1080
+ true,
1081
+ true,
1082
+ false,
1083
+ true,
1084
+ true,
1085
+ true,
1086
+ true,
1087
+ true,
1088
+ true,
1089
+ false
1090
+ ],
1091
+ "language_key": "language_instruction",
1092
+ "action_proprio_normalization_type": "normal",
1093
+ "standardize_fn": {
1094
+ "module": "crossformer.data.oxe.oxe_standardization_transforms",
1095
+ "name": "aloha_dough_dataset_transform",
1096
+ "args": [],
1097
+ "kwargs": {}
1098
+ }
1099
+ },
1100
+ {
1101
+ "name": "aloha_sushi_cut_full_dataset",
1102
+ "data_dir": null,
1103
+ "image_obs_keys": {
1104
+ "primary": null,
1105
+ "high": "cam_high",
1106
+ "nav": null,
1107
+ "left_wrist": "cam_left_wrist",
1108
+ "right_wrist": "cam_right_wrist"
1109
+ },
1110
+ "proprio_obs_keys": {
1111
+ "bimanual": "proprio",
1112
+ "quadruped": null
1113
+ },
1114
+ "proprio_obs_dims": {
1115
+ "bimanual": 14,
1116
+ "quadruped": 59
1117
+ },
1118
+ "action_normalization_mask": [
1119
+ true,
1120
+ true,
1121
+ true,
1122
+ true,
1123
+ true,
1124
+ true,
1125
+ false,
1126
+ true,
1127
+ true,
1128
+ true,
1129
+ true,
1130
+ true,
1131
+ true,
1132
+ false
1133
+ ],
1134
+ "language_key": "language_instruction",
1135
+ "action_proprio_normalization_type": "normal",
1136
+ "standardize_fn": {
1137
+ "module": "crossformer.data.oxe.oxe_standardization_transforms",
1138
+ "name": "aloha_dough_dataset_transform",
1139
+ "args": [],
1140
+ "kwargs": {}
1141
+ }
1142
+ },
1143
+ {
1144
+ "name": "bridge_dataset",
1145
+ "data_dir": null,
1146
+ "image_obs_keys": {
1147
+ "primary": "image_0",
1148
+ "high": null,
1149
+ "nav": null,
1150
+ "left_wrist": null,
1151
+ "right_wrist": null
1152
+ },
1153
+ "proprio_obs_keys": {
1154
+ "bimanual": null,
1155
+ "quadruped": null
1156
+ },
1157
+ "proprio_obs_dims": {
1158
+ "bimanual": 14,
1159
+ "quadruped": 59
1160
+ },
1161
+ "action_normalization_mask": [
1162
+ true,
1163
+ true,
1164
+ true,
1165
+ true,
1166
+ true,
1167
+ true,
1168
+ false
1169
+ ],
1170
+ "language_key": "language_instruction",
1171
+ "action_proprio_normalization_type": "normal",
1172
+ "standardize_fn": {
1173
+ "module": "crossformer.data.oxe.oxe_standardization_transforms",
1174
+ "name": "bridge_dataset_transform",
1175
+ "args": [],
1176
+ "kwargs": {}
1177
+ }
1178
+ },
1179
+ {
1180
+ "name": "go1",
1181
+ "data_dir": null,
1182
+ "image_obs_keys": {
1183
+ "primary": null,
1184
+ "high": null,
1185
+ "nav": null,
1186
+ "left_wrist": null,
1187
+ "right_wrist": null
1188
+ },
1189
+ "proprio_obs_keys": {
1190
+ "bimanual": null,
1191
+ "quadruped": "proprio"
1192
+ },
1193
+ "proprio_obs_dims": {
1194
+ "bimanual": 14,
1195
+ "quadruped": 59
1196
+ },
1197
+ "action_normalization_mask": [
1198
+ true,
1199
+ true,
1200
+ true,
1201
+ true,
1202
+ true,
1203
+ true,
1204
+ true,
1205
+ true,
1206
+ true,
1207
+ true,
1208
+ true,
1209
+ true
1210
+ ],
1211
+ "language_key": "language_instruction",
1212
+ "action_proprio_normalization_type": "normal",
1213
+ "standardize_fn": {
1214
+ "module": "crossformer.data.oxe.oxe_standardization_transforms",
1215
+ "name": "go1_dataset_transform",
1216
+ "args": [],
1217
+ "kwargs": {}
1218
+ }
1219
+ },
1220
+ {
1221
+ "name": "droid_wipe",
1222
+ "data_dir": null,
1223
+ "image_obs_keys": {
1224
+ "primary": "exterior_image_2_left",
1225
+ "high": null,
1226
+ "nav": null,
1227
+ "left_wrist": null,
1228
+ "right_wrist": null
1229
+ },
1230
+ "proprio_obs_keys": {
1231
+ "bimanual": null,
1232
+ "quadruped": null
1233
+ },
1234
+ "proprio_obs_dims": {
1235
+ "bimanual": 14,
1236
+ "quadruped": 59
1237
+ },
1238
+ "action_normalization_mask": [
1239
+ true,
1240
+ true,
1241
+ true,
1242
+ true,
1243
+ true,
1244
+ true,
1245
+ false
1246
+ ],
1247
+ "language_key": "language_instruction",
1248
+ "action_proprio_normalization_type": "normal",
1249
+ "standardize_fn": {
1250
+ "module": "crossformer.data.oxe.oxe_standardization_transforms",
1251
+ "name": "droid_dataset_transform",
1252
+ "args": [],
1253
+ "kwargs": {}
1254
+ }
1255
+ },
1256
+ {
1257
+ "name": "omnimimic_gnm_dataset",
1258
+ "data_dir": null,
1259
+ "image_obs_keys": {
1260
+ "primary": null,
1261
+ "high": null,
1262
+ "nav": "image",
1263
+ "left_wrist": null,
1264
+ "right_wrist": null
1265
+ },
1266
+ "proprio_obs_keys": {
1267
+ "bimanual": null,
1268
+ "quadruped": null
1269
+ },
1270
+ "proprio_obs_dims": {
1271
+ "bimanual": 14,
1272
+ "quadruped": 59
1273
+ },
1274
+ "action_normalization_mask": [
1275
+ true,
1276
+ true
1277
+ ],
1278
+ "language_key": "language_instruction",
1279
+ "action_proprio_normalization_type": "normal",
1280
+ "standardize_fn": {
1281
+ "module": "crossformer.data.oxe.oxe_standardization_transforms",
1282
+ "name": "omnimimic_gnm_transform",
1283
+ "args": [],
1284
+ "kwargs": {}
1285
+ }
1286
+ },
1287
+ {
1288
+ "name": "fractal20220817_data",
1289
+ "data_dir": null,
1290
+ "image_obs_keys": {
1291
+ "primary": "image",
1292
+ "high": null,
1293
+ "nav": null,
1294
+ "left_wrist": null,
1295
+ "right_wrist": null
1296
+ },
1297
+ "proprio_obs_keys": {
1298
+ "bimanual": null,
1299
+ "quadruped": null
1300
+ },
1301
+ "proprio_obs_dims": {
1302
+ "bimanual": 14,
1303
+ "quadruped": 59
1304
+ },
1305
+ "action_normalization_mask": [
1306
+ true,
1307
+ true,
1308
+ true,
1309
+ true,
1310
+ true,
1311
+ true,
1312
+ false
1313
+ ],
1314
+ "language_key": "language_instruction",
1315
+ "action_proprio_normalization_type": "normal",
1316
+ "standardize_fn": {
1317
+ "module": "crossformer.data.oxe.oxe_standardization_transforms",
1318
+ "name": "rt1_dataset_transform",
1319
+ "args": [],
1320
+ "kwargs": {}
1321
+ }
1322
+ }
1323
+ ],
1324
+ "frame_transform_kwargs": {
1325
+ "image_augment_kwargs": {
1326
+ "high": {
1327
+ "augment_order": [
1328
+ "random_resized_crop",
1329
+ "random_brightness",
1330
+ "random_contrast",
1331
+ "random_saturation",
1332
+ "random_hue"
1333
+ ],
1334
+ "random_brightness": [
1335
+ 0.1
1336
+ ],
1337
+ "random_contrast": [
1338
+ 0.9,
1339
+ 1.1
1340
+ ],
1341
+ "random_hue": [
1342
+ 0.05
1343
+ ],
1344
+ "random_resized_crop": {
1345
+ "ratio": [
1346
+ 0.75,
1347
+ 1.3333333333333333
1348
+ ],
1349
+ "scale": [
1350
+ 0.9,
1351
+ 1.0
1352
+ ]
1353
+ },
1354
+ "random_saturation": [
1355
+ 0.9,
1356
+ 1.1
1357
+ ]
1358
+ },
1359
+ "left_wrist": {
1360
+ "augment_order": [
1361
+ "random_resized_crop",
1362
+ "random_brightness",
1363
+ "random_contrast",
1364
+ "random_saturation",
1365
+ "random_hue"
1366
+ ],
1367
+ "random_brightness": [
1368
+ 0.1
1369
+ ],
1370
+ "random_contrast": [
1371
+ 0.9,
1372
+ 1.1
1373
+ ],
1374
+ "random_hue": [
1375
+ 0.05
1376
+ ],
1377
+ "random_resized_crop": {
1378
+ "ratio": [
1379
+ 0.75,
1380
+ 1.3333333333333333
1381
+ ],
1382
+ "scale": [
1383
+ 0.9,
1384
+ 1.0
1385
+ ]
1386
+ },
1387
+ "random_saturation": [
1388
+ 0.9,
1389
+ 1.1
1390
+ ]
1391
+ },
1392
+ "nav": {
1393
+ "augment_order": [
1394
+ "random_resized_crop",
1395
+ "random_brightness",
1396
+ "random_contrast",
1397
+ "random_saturation",
1398
+ "random_hue"
1399
+ ],
1400
+ "random_brightness": [
1401
+ 0.1
1402
+ ],
1403
+ "random_contrast": [
1404
+ 0.9,
1405
+ 1.1
1406
+ ],
1407
+ "random_hue": [
1408
+ 0.05
1409
+ ],
1410
+ "random_resized_crop": {
1411
+ "ratio": [
1412
+ 0.9,
1413
+ 1.1
1414
+ ],
1415
+ "scale": [
1416
+ 0.8,
1417
+ 1.0
1418
+ ]
1419
+ },
1420
+ "random_saturation": [
1421
+ 0.9,
1422
+ 1.1
1423
+ ]
1424
+ },
1425
+ "primary": {
1426
+ "augment_order": [
1427
+ "random_resized_crop",
1428
+ "random_brightness",
1429
+ "random_contrast",
1430
+ "random_saturation",
1431
+ "random_hue"
1432
+ ],
1433
+ "random_brightness": [
1434
+ 0.1
1435
+ ],
1436
+ "random_contrast": [
1437
+ 0.9,
1438
+ 1.1
1439
+ ],
1440
+ "random_hue": [
1441
+ 0.05
1442
+ ],
1443
+ "random_resized_crop": {
1444
+ "ratio": [
1445
+ 0.9,
1446
+ 1.1
1447
+ ],
1448
+ "scale": [
1449
+ 0.8,
1450
+ 1.0
1451
+ ]
1452
+ },
1453
+ "random_saturation": [
1454
+ 0.9,
1455
+ 1.1
1456
+ ]
1457
+ },
1458
+ "right_wrist": {
1459
+ "augment_order": [
1460
+ "random_resized_crop",
1461
+ "random_brightness",
1462
+ "random_contrast",
1463
+ "random_saturation",
1464
+ "random_hue"
1465
+ ],
1466
+ "random_brightness": [
1467
+ 0.1
1468
+ ],
1469
+ "random_contrast": [
1470
+ 0.9,
1471
+ 1.1
1472
+ ],
1473
+ "random_hue": [
1474
+ 0.05
1475
+ ],
1476
+ "random_resized_crop": {
1477
+ "ratio": [
1478
+ 0.75,
1479
+ 1.3333333333333333
1480
+ ],
1481
+ "scale": [
1482
+ 0.9,
1483
+ 1.0
1484
+ ]
1485
+ },
1486
+ "random_saturation": [
1487
+ 0.9,
1488
+ 1.1
1489
+ ]
1490
+ }
1491
+ },
1492
+ "num_parallel_calls": 200,
1493
+ "resize_size": {
1494
+ "high": [
1495
+ 224,
1496
+ 224
1497
+ ],
1498
+ "left_wrist": [
1499
+ 224,
1500
+ 224
1501
+ ],
1502
+ "nav": [
1503
+ 224,
1504
+ 224
1505
+ ],
1506
+ "primary": [
1507
+ 224,
1508
+ 224
1509
+ ],
1510
+ "right_wrist": [
1511
+ 224,
1512
+ 224
1513
+ ]
1514
+ }
1515
+ },
1516
+ "sample_weights": [
1517
+ 0.021755552811740044,
1518
+ 0.009986998241552502,
1519
+ 0.0023179438302583103,
1520
+ 0.0016371801232614254,
1521
+ 0.0008888419195460733,
1522
+ 0.007874885740329124,
1523
+ 0.0032054418194781,
1524
+ 0.004113224757060238,
1525
+ 0.006839324427193265,
1526
+ 0.01479473259297815,
1527
+ 0.015045048015813084,
1528
+ 0.0007163148639093663,
1529
+ 0.002826978440249719,
1530
+ 0.008290489894059044,
1531
+ 0.007414589603576858,
1532
+ 0.005878414243900214,
1533
+ 0.021755552811740044,
1534
+ 0.0001874782805166165,
1535
+ 0.0030709017944702635,
1536
+ 0.007599149035416076,
1537
+ 0.0026296096724881066,
1538
+ 0.0005253087662862543,
1539
+ 0.0002175555281174004,
1540
+ 0.085,
1541
+ 0.014166666666666666,
1542
+ 0.014166666666666666,
1543
+ 0.014166666666666666,
1544
+ 0.014166666666666666,
1545
+ 0.014166666666666666,
1546
+ 0.014166666666666666,
1547
+ 0.17,
1548
+ 0.085,
1549
+ 0.085,
1550
+ 0.17,
1551
+ 0.17
1552
+ ],
1553
+ "shuffle_buffer_size": 25000,
1554
+ "traj_read_threads": 48,
1555
+ "traj_transform_kwargs": {
1556
+ "action_horizon": 100,
1557
+ "goal_relabeling_kwargs": {
1558
+ "max_goal_distance": 15
1559
+ },
1560
+ "goal_relabeling_strategy": "uniform",
1561
+ "head_to_dataset": {
1562
+ "bimanual": [
1563
+ "aloha_pen_uncap_diverse_dataset",
1564
+ "aloha_dough_cut_dataset",
1565
+ "aloha_lucy_dataset",
1566
+ "aloha_drawer_dataset",
1567
+ "aloha_pick_place_dataset",
1568
+ "aloha_static_dataset",
1569
+ "aloha_sushi_cut_full_dataset"
1570
+ ],
1571
+ "nav": [
1572
+ "omnimimic_gnm_dataset"
1573
+ ],
1574
+ "quadruped": [
1575
+ "go1"
1576
+ ],
1577
+ "single_arm": [
1578
+ "bridge_dataset",
1579
+ "fractal20220817_data",
1580
+ "kuka",
1581
+ "taco_play",
1582
+ "taco_extra",
1583
+ "jaco_play",
1584
+ "berkeley_cable_routing",
1585
+ "roboturk",
1586
+ "nyu_door_opening_surprising_effectiveness",
1587
+ "viola",
1588
+ "berkeley_autolab_ur5",
1589
+ "toto",
1590
+ "language_table",
1591
+ "stanford_hydra_dataset_converted_externally_to_rlds",
1592
+ "austin_buds_dataset_converted_externally_to_rlds",
1593
+ "nyu_franka_play_dataset_converted_externally_to_rlds",
1594
+ "furniture_bench_dataset_converted_externally_to_rlds",
1595
+ "austin_sailor_dataset_converted_externally_to_rlds",
1596
+ "austin_sirius_dataset_converted_externally_to_rlds",
1597
+ "bc_z",
1598
+ "dlr_edan_shared_control_converted_externally_to_rlds",
1599
+ "iamlab_cmu_pickup_insert_converted_externally_to_rlds",
1600
+ "utaustin_mutex",
1601
+ "berkeley_fanuc_manipulation",
1602
+ "cmu_stretch",
1603
+ "droid",
1604
+ "droid_wipe"
1605
+ ]
1606
+ },
1607
+ "max_action_dim": 14,
1608
+ "subsample_length": 100,
1609
+ "task_augment_kwargs": {
1610
+ "keep_image_prob": 1.0
1611
+ },
1612
+ "task_augment_strategy": "delete_task_conditioning",
1613
+ "window_size": 5
1614
+ },
1615
+ "traj_transform_threads": 48
1616
+ },
1617
+ "eval_datasets": [],
1618
+ "eval_interval": 5e+20,
1619
+ "log_interval": 500,
1620
+ "model": {
1621
+ "heads": {
1622
+ "bimanual": {
1623
+ "args": [],
1624
+ "kwargs": {
1625
+ "action_dim": 14,
1626
+ "action_horizon": 100,
1627
+ "clip_pred": false,
1628
+ "constrain_loss_dims": true,
1629
+ "loss_weight": 1.0,
1630
+ "num_preds": 14,
1631
+ "pool_strategy": "pass",
1632
+ "readout_key": "readout_bimanual"
1633
+ },
1634
+ "module": "crossformer.model.components.action_heads",
1635
+ "name": "L1ActionHead"
1636
+ },
1637
+ "nav": {
1638
+ "args": [],
1639
+ "kwargs": {
1640
+ "action_dim": 2,
1641
+ "action_horizon": 4,
1642
+ "clip_pred": false,
1643
+ "constrain_loss_dims": true,
1644
+ "loss_weight": 1.0,
1645
+ "num_preds": 2,
1646
+ "pool_strategy": "pass",
1647
+ "readout_key": "readout_nav"
1648
+ },
1649
+ "module": "crossformer.model.components.action_heads",
1650
+ "name": "L1ActionHead"
1651
+ },
1652
+ "quadruped": {
1653
+ "args": [],
1654
+ "kwargs": {
1655
+ "action_dim": 12,
1656
+ "action_horizon": 1,
1657
+ "clip_pred": false,
1658
+ "constrain_loss_dims": true,
1659
+ "loss_weight": 1.0,
1660
+ "num_preds": 12,
1661
+ "pool_strategy": "pass",
1662
+ "readout_key": "readout_quadruped"
1663
+ },
1664
+ "module": "crossformer.model.components.action_heads",
1665
+ "name": "L1ActionHead"
1666
+ },
1667
+ "single_arm": {
1668
+ "args": [],
1669
+ "kwargs": {
1670
+ "action_dim": 7,
1671
+ "action_horizon": 4,
1672
+ "clip_pred": false,
1673
+ "constrain_loss_dims": true,
1674
+ "loss_weight": 1.0,
1675
+ "num_preds": 7,
1676
+ "pool_strategy": "pass",
1677
+ "readout_key": "readout_single_arm"
1678
+ },
1679
+ "module": "crossformer.model.components.action_heads",
1680
+ "name": "L1ActionHead"
1681
+ }
1682
+ },
1683
+ "max_horizon": 10,
1684
+ "observation_tokenizers": {
1685
+ "bimanual": {
1686
+ "args": [],
1687
+ "kwargs": {
1688
+ "dropout_rate": 0.2,
1689
+ "obs_keys": [
1690
+ "proprio_bimanual"
1691
+ ]
1692
+ },
1693
+ "module": "crossformer.model.components.tokenizers",
1694
+ "name": "LowdimObsTokenizer"
1695
+ },
1696
+ "high": {
1697
+ "args": [],
1698
+ "kwargs": {
1699
+ "encoder": {
1700
+ "args": [],
1701
+ "kwargs": {},
1702
+ "module": "crossformer.model.components.vit_encoders",
1703
+ "name": "ResNet26FILM"
1704
+ },
1705
+ "obs_stack_keys": [
1706
+ "image_high"
1707
+ ],
1708
+ "task_film_keys": [
1709
+ "language_instruction"
1710
+ ],
1711
+ "task_stack_keys": [
1712
+ "image_high"
1713
+ ]
1714
+ },
1715
+ "module": "crossformer.model.components.tokenizers",
1716
+ "name": "ImageTokenizer"
1717
+ },
1718
+ "left": {
1719
+ "args": [],
1720
+ "kwargs": {
1721
+ "encoder": {
1722
+ "args": [],
1723
+ "kwargs": {},
1724
+ "module": "crossformer.model.components.vit_encoders",
1725
+ "name": "ResNet26FILM"
1726
+ },
1727
+ "obs_stack_keys": [
1728
+ "image_left_wrist"
1729
+ ],
1730
+ "task_film_keys": [
1731
+ "language_instruction"
1732
+ ],
1733
+ "task_stack_keys": []
1734
+ },
1735
+ "module": "crossformer.model.components.tokenizers",
1736
+ "name": "ImageTokenizer"
1737
+ },
1738
+ "nav": {
1739
+ "args": [],
1740
+ "kwargs": {
1741
+ "encoder": {
1742
+ "args": [],
1743
+ "kwargs": {},
1744
+ "module": "crossformer.model.components.vit_encoders",
1745
+ "name": "ResNet26"
1746
+ },
1747
+ "obs_stack_keys": [
1748
+ "image_nav"
1749
+ ],
1750
+ "task_film_keys": [],
1751
+ "task_stack_keys": [
1752
+ "image_nav"
1753
+ ]
1754
+ },
1755
+ "module": "crossformer.model.components.tokenizers",
1756
+ "name": "ImageTokenizer"
1757
+ },
1758
+ "primary": {
1759
+ "args": [],
1760
+ "kwargs": {
1761
+ "encoder": {
1762
+ "args": [],
1763
+ "kwargs": {},
1764
+ "module": "crossformer.model.components.vit_encoders",
1765
+ "name": "ResNet26FILM"
1766
+ },
1767
+ "obs_stack_keys": [
1768
+ "image_primary"
1769
+ ],
1770
+ "task_film_keys": [
1771
+ "language_instruction"
1772
+ ],
1773
+ "task_stack_keys": [
1774
+ "image_primary"
1775
+ ]
1776
+ },
1777
+ "module": "crossformer.model.components.tokenizers",
1778
+ "name": "ImageTokenizer"
1779
+ },
1780
+ "quadruped": {
1781
+ "args": [],
1782
+ "kwargs": {
1783
+ "obs_keys": [
1784
+ "proprio_quadruped"
1785
+ ]
1786
+ },
1787
+ "module": "crossformer.model.components.tokenizers",
1788
+ "name": "LowdimObsTokenizer"
1789
+ },
1790
+ "right": {
1791
+ "args": [],
1792
+ "kwargs": {
1793
+ "encoder": {
1794
+ "args": [],
1795
+ "kwargs": {},
1796
+ "module": "crossformer.model.components.vit_encoders",
1797
+ "name": "ResNet26FILM"
1798
+ },
1799
+ "obs_stack_keys": [
1800
+ "image_right_wrist"
1801
+ ],
1802
+ "task_film_keys": [
1803
+ "language_instruction"
1804
+ ],
1805
+ "task_stack_keys": []
1806
+ },
1807
+ "module": "crossformer.model.components.tokenizers",
1808
+ "name": "ImageTokenizer"
1809
+ }
1810
+ },
1811
+ "readouts": {
1812
+ "bimanual": 100,
1813
+ "nav": 4,
1814
+ "quadruped": 1,
1815
+ "single_arm": 4
1816
+ },
1817
+ "repeat_task_tokens": true,
1818
+ "task_tokenizers": {},
1819
+ "token_embedding_size": 512,
1820
+ "transformer_kwargs": {
1821
+ "attention_dropout_rate": 0.0,
1822
+ "dropout_rate": 0.1,
1823
+ "mlp_dim": 2048,
1824
+ "num_attention_heads": 8,
1825
+ "num_layers": 12,
1826
+ "repeat_pos_enc": true
1827
+ }
1828
+ },
1829
+ "num_steps": 300000,
1830
+ "optimizer": {
1831
+ "clip_gradient": 1.0,
1832
+ "frozen_keys": [],
1833
+ "learning_rate": {
1834
+ "init_value": 0.0,
1835
+ "name": "rsqrt",
1836
+ "peak_value": 0.0003,
1837
+ "timescale": 10000,
1838
+ "warmup_steps": 2000
1839
+ },
1840
+ "weight_decay": 0.1
1841
+ },
1842
+ "prefetch_num_batches": 0,
1843
+ "pretrained_loaders": [
1844
+ {
1845
+ "module": "experiments.sudeep.aloha.resnet_pt",
1846
+ "name": "resnet_26_loader",
1847
+ "args": [],
1848
+ "kwargs": {
1849
+ "restore_path": ""
1850
+ }
1851
+ }
1852
+ ],
1853
+ "resume_path": null,
1854
+ "save_dir": null,
1855
+ "save_interval": 10000,
1856
+ "seed": 42,
1857
+ "skip_norm_keys": [
1858
+ "proprio_bimanual"
1859
+ ],
1860
+ "start_step": null,
1861
+ "text_processor": {
1862
+ "args": [],
1863
+ "kwargs": {},
1864
+ "module": "crossformer.data.utils.text_processing",
1865
+ "name": "UniversalSentenceEncoder"
1866
+ },
1867
+ "val_kwargs": {
1868
+ "num_val_batches": 16,
1869
+ "val_shuffle_buffer_size": 1000
1870
+ },
1871
+ "viz_datasets": [],
1872
+ "viz_interval": 5e+20,
1873
+ "viz_kwargs": {
1874
+ "eval_batch_size": 128,
1875
+ "samples_per_state": 8,
1876
+ "trajs_for_metrics": 100,
1877
+ "trajs_for_viz": 8
1878
+ },
1879
+ "wandb": {
1880
+ "entity": null,
1881
+ "group": null,
1882
+ "project": "crossformer"
1883
+ },
1884
+ "wandb_resume_id": null,
1885
+ "window_size": 5
1886
+ }