Yewandou commited on
Commit
22cf30e
·
verified ·
1 Parent(s): 28374a2

psix VLA: psix/psix_sonic_neck_round3_2609100359/run_config.json

Browse files
psix/psix_sonic_neck_round3_2609100359/run_config.json ADDED
@@ -0,0 +1,2234 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "exp": "finetune-sonic-neck-psix-no-vlm-tok-v1-1-round3",
3
+ "seed": 292285,
4
+ "auto_tag_run": false,
5
+ "eval": false,
6
+ "debug": false,
7
+ "timestamp": "2609100359",
8
+ "log": {
9
+ "logging_dir": "logs",
10
+ "report_to": "wandb",
11
+ "log_freq": 100
12
+ },
13
+ "wandb": {
14
+ "project": "psix",
15
+ "entity": "jhy906219458-university-of-southern-california",
16
+ "group": "psix_finetune",
17
+ "id": "lb1c1n9o",
18
+ "name": "finetune-sonic-neck-psix-no-vlm-tok-v1-1-round3.sonic-neck-no-vlm.flow1000.cosine.lr1.0e-04.b256.gpus2.2609100359",
19
+ "resume": "allow"
20
+ },
21
+ "train": {
22
+ "num_workers": 8,
23
+ "overfit_single_batch": false,
24
+ "name": "psix_finetune",
25
+ "run_tag": "sonic-neck-no-vlm",
26
+ "resume_from_checkpoint": ".runs/psix_finetune/finetune-sonic-neck-psix-no-vlm-tok-v1-1-round3.sonic-neck-no-vlm.flow1000.cosine.lr1.0e-04.b256.gpus2.2609100359",
27
+ "skip_resumed_steps": false,
28
+ "hf_token": "PosixPath('.hf_token')",
29
+ "lora": false,
30
+ "output_dir": ".runs",
31
+ "gradient_accumulation_steps": 4,
32
+ "mixed_precision": "bf16",
33
+ "max_grad_norm": 1.0,
34
+ "optimizer_foreach": null,
35
+ "grad_spike_factor": null,
36
+ "grad_spike_window": 200,
37
+ "train_batch_size": 32,
38
+ "val_batch_size": 16,
39
+ "val_num_batches": 10,
40
+ "checkpointing_steps": 2500,
41
+ "max_checkpoints_to_keep": 5,
42
+ "preserve_checkpoint_every": 10000,
43
+ "validation_steps": 5000,
44
+ "learning_rate": 0.0001,
45
+ "lr_scheduler_type": "cosine",
46
+ "lr_scheduler_kwargs": {
47
+ "betas": [
48
+ 0.95,
49
+ 0.999
50
+ ],
51
+ "weight_decay": 1e-06,
52
+ "eps": 1e-08
53
+ },
54
+ "scheduler_specific_kwargs": {},
55
+ "data_parallel": "ddp",
56
+ "sharding_strategy": "full-shard",
57
+ "deepspeed_config": "/scratch/project/prj-02-spatial-intelligence/jwang/h_tmp/psi/scripts/deepspeed/zero3.json",
58
+ "enable_gradient_checkpointing": true,
59
+ "enable_mixed_precision_training": true,
60
+ "reduce_in_full_precision": true,
61
+ "max_training_steps": 40000,
62
+ "num_train_epochs": null,
63
+ "warmup_steps": 1000,
64
+ "warmup_ratio": null
65
+ },
66
+ "data": {
67
+ "transform": {
68
+ "repack": {
69
+ "dataset_name": "simple+decoupled_wbc_g1+decoupled_wbc_h1+psix_he_g1_sonic+psix_he_h1_sonic+psix_g1_sonic_neck+egodex_psix",
70
+ "goal_sampling": null,
71
+ "goal_future_offsets": [
72
+ 1.0,
73
+ 2.0,
74
+ 3.0,
75
+ 4.0
76
+ ],
77
+ "goal_end_segment_prob": null,
78
+ "use_paraphrases": false,
79
+ "pad_action_dim": 80,
80
+ "pad_state_dim": 45,
81
+ "action_chunk_size": null,
82
+ "simple_g1": {
83
+ "dataset_name": "simple",
84
+ "goal_sampling": false,
85
+ "goal_future_offsets": [
86
+ 1.0,
87
+ 2.0,
88
+ 3.0,
89
+ 4.0
90
+ ],
91
+ "goal_end_segment_prob": 0.5,
92
+ "use_paraphrases": false,
93
+ "num_past_frames": 0,
94
+ "history_stride": 1,
95
+ "num_past_state_frames": 0,
96
+ "num_past_image_frames": 0,
97
+ "action_chunk_size": 30,
98
+ "pad_action_dim": 80,
99
+ "pad_state_dim": 45,
100
+ "image_keys": [
101
+ "video.rs_view"
102
+ ],
103
+ "state_key": [
104
+ "state.left_hand",
105
+ "state.right_hand",
106
+ "state.left_arm",
107
+ "state.right_arm",
108
+ "state.rpy",
109
+ "state.height"
110
+ ],
111
+ "action_key": [
112
+ "action.left_hand",
113
+ "action.right_hand",
114
+ "action.left_arm",
115
+ "action.right_arm",
116
+ "action.rpy",
117
+ "action.height",
118
+ "action.torso_vx",
119
+ "action.torso_vy",
120
+ "action.torso_vyaw",
121
+ "action.target_yaw"
122
+ ],
123
+ "instruction_key": [
124
+ "annotation.human.task_description"
125
+ ],
126
+ "video": {
127
+ "delta_indices": [
128
+ 0
129
+ ],
130
+ "eval_delta_indices": [
131
+ 0
132
+ ],
133
+ "modality_keys": [
134
+ "video.rs_view"
135
+ ]
136
+ },
137
+ "state": {
138
+ "delta_indices": [
139
+ 0
140
+ ],
141
+ "eval_delta_indices": [
142
+ 0
143
+ ],
144
+ "modality_keys": [
145
+ "state.left_hand",
146
+ "state.right_hand",
147
+ "state.left_arm",
148
+ "state.right_arm",
149
+ "state.rpy",
150
+ "state.height"
151
+ ]
152
+ },
153
+ "action": {
154
+ "delta_indices": [
155
+ 0,
156
+ 1,
157
+ 2,
158
+ 3,
159
+ 4,
160
+ 5,
161
+ 6,
162
+ 7,
163
+ 8,
164
+ 9,
165
+ 10,
166
+ 11,
167
+ 12,
168
+ 13,
169
+ 14,
170
+ 15,
171
+ 16,
172
+ 17,
173
+ 18,
174
+ 19,
175
+ 20,
176
+ 21,
177
+ 22,
178
+ 23,
179
+ 24,
180
+ 25,
181
+ 26,
182
+ 27,
183
+ 28,
184
+ 29
185
+ ],
186
+ "eval_delta_indices": [
187
+ 0,
188
+ 1,
189
+ 2,
190
+ 3,
191
+ 4,
192
+ 5,
193
+ 6,
194
+ 7,
195
+ 8,
196
+ 9,
197
+ 10,
198
+ 11,
199
+ 12,
200
+ 13,
201
+ 14,
202
+ 15,
203
+ 16,
204
+ 17,
205
+ 18,
206
+ 19,
207
+ 20,
208
+ 21,
209
+ 22,
210
+ 23,
211
+ 24,
212
+ 25,
213
+ 26,
214
+ 27,
215
+ 28,
216
+ 29,
217
+ 30,
218
+ 31,
219
+ 32,
220
+ 33,
221
+ 34,
222
+ 35
223
+ ],
224
+ "modality_keys": [
225
+ "action.left_hand",
226
+ "action.right_hand",
227
+ "action.left_arm",
228
+ "action.right_arm",
229
+ "action.rpy",
230
+ "action.height",
231
+ "action.torso_vx",
232
+ "action.torso_vy",
233
+ "action.torso_vyaw",
234
+ "action.target_yaw"
235
+ ]
236
+ },
237
+ "language": {
238
+ "delta_indices": [
239
+ 0
240
+ ],
241
+ "eval_delta_indices": [
242
+ 0
243
+ ],
244
+ "modality_keys": [
245
+ "annotation.human.task_description"
246
+ ]
247
+ }
248
+ },
249
+ "decoupled_wbc_g1": {
250
+ "dataset_name": "decoupled_wbc_g1",
251
+ "goal_sampling": false,
252
+ "goal_future_offsets": [
253
+ 1.0,
254
+ 2.0,
255
+ 3.0,
256
+ 4.0
257
+ ],
258
+ "goal_end_segment_prob": 0.5,
259
+ "use_paraphrases": false,
260
+ "image_keys": [
261
+ "video.egocentric"
262
+ ],
263
+ "state_key": [
264
+ "state.joint_positions"
265
+ ],
266
+ "action_key": [
267
+ "action.hand_joints",
268
+ "action.arm_joints",
269
+ "action.torso_rpy",
270
+ "action.base_height",
271
+ "action.base_vx",
272
+ "action.base_vy",
273
+ "action.base_vyaw",
274
+ "action.target_yaw"
275
+ ],
276
+ "subgoal_key": [
277
+ "subgoal.egocentric"
278
+ ],
279
+ "instruction_key": [
280
+ "annotation.human_instruction"
281
+ ],
282
+ "subtask_prompt_key": [
283
+ "annotation.sub_task"
284
+ ],
285
+ "memory_text_key": [
286
+ "annotation.memory_desc"
287
+ ],
288
+ "sub_task_delta_key": [
289
+ "annotation.sub_task_delta"
290
+ ],
291
+ "next_subtask_key": [
292
+ "annotation.next_sub_task"
293
+ ],
294
+ "prev_subtask_key": [
295
+ "annotation.prev_sub_task"
296
+ ],
297
+ "memory_items_key": [
298
+ "annotation.memory_items"
299
+ ],
300
+ "meta_keys": [
301
+ "meta.quality",
302
+ "meta.speed",
303
+ "meta.control_mode",
304
+ "meta.embodiment_tag",
305
+ "meta.failure_type",
306
+ "meta.scene_type",
307
+ "meta.data_source",
308
+ "meta.operator_id",
309
+ "meta.task_description",
310
+ "meta.episode_length",
311
+ "meta.number_of_tasks"
312
+ ],
313
+ "num_past_frames": 0,
314
+ "history_stride": 1,
315
+ "num_past_state_frames": 0,
316
+ "num_past_image_frames": 0,
317
+ "action_chunk_size": 30,
318
+ "pad_action_dim": 80,
319
+ "pad_state_dim": 45,
320
+ "video": {
321
+ "delta_indices": [
322
+ 0
323
+ ],
324
+ "eval_delta_indices": [
325
+ 0
326
+ ],
327
+ "modality_keys": [
328
+ "video.egocentric"
329
+ ]
330
+ },
331
+ "state": {
332
+ "delta_indices": [
333
+ 0
334
+ ],
335
+ "eval_delta_indices": [
336
+ 0
337
+ ],
338
+ "modality_keys": [
339
+ "state.joint_positions"
340
+ ]
341
+ },
342
+ "action": {
343
+ "delta_indices": [
344
+ 0,
345
+ 1,
346
+ 2,
347
+ 3,
348
+ 4,
349
+ 5,
350
+ 6,
351
+ 7,
352
+ 8,
353
+ 9,
354
+ 10,
355
+ 11,
356
+ 12,
357
+ 13,
358
+ 14,
359
+ 15,
360
+ 16,
361
+ 17,
362
+ 18,
363
+ 19,
364
+ 20,
365
+ 21,
366
+ 22,
367
+ 23,
368
+ 24,
369
+ 25,
370
+ 26,
371
+ 27,
372
+ 28,
373
+ 29
374
+ ],
375
+ "eval_delta_indices": [
376
+ 0,
377
+ 1,
378
+ 2,
379
+ 3,
380
+ 4,
381
+ 5,
382
+ 6,
383
+ 7,
384
+ 8,
385
+ 9,
386
+ 10,
387
+ 11,
388
+ 12,
389
+ 13,
390
+ 14,
391
+ 15,
392
+ 16,
393
+ 17,
394
+ 18,
395
+ 19,
396
+ 20,
397
+ 21,
398
+ 22,
399
+ 23,
400
+ 24,
401
+ 25,
402
+ 26,
403
+ 27,
404
+ 28,
405
+ 29,
406
+ 30,
407
+ 31,
408
+ 32,
409
+ 33,
410
+ 34,
411
+ 35
412
+ ],
413
+ "modality_keys": [
414
+ "action.hand_joints",
415
+ "action.arm_joints",
416
+ "action.torso_rpy",
417
+ "action.base_height",
418
+ "action.base_vx",
419
+ "action.base_vy",
420
+ "action.base_vyaw",
421
+ "action.target_yaw"
422
+ ]
423
+ },
424
+ "language": {
425
+ "delta_indices": [
426
+ 0
427
+ ],
428
+ "eval_delta_indices": [
429
+ 0
430
+ ],
431
+ "modality_keys": [
432
+ "annotation.human_instruction",
433
+ "annotation.sub_task",
434
+ "annotation.memory_desc",
435
+ "annotation.sub_task_delta",
436
+ "annotation.next_sub_task",
437
+ "annotation.prev_sub_task",
438
+ "annotation.memory_items"
439
+ ]
440
+ },
441
+ "subgoal": {
442
+ "delta_indices": [
443
+ 0
444
+ ],
445
+ "eval_delta_indices": [
446
+ 0
447
+ ],
448
+ "modality_keys": [
449
+ "subgoal.egocentric"
450
+ ],
451
+ "kwargs": {
452
+ "subgoal_prob": 0.25,
453
+ "sample_from_gt_goal_prob": 0.25,
454
+ "future_horizon": [
455
+ 30,
456
+ 120
457
+ ]
458
+ }
459
+ },
460
+ "meta": {
461
+ "delta_indices": [
462
+ 0
463
+ ],
464
+ "eval_delta_indices": [
465
+ 0
466
+ ],
467
+ "modality_keys": [
468
+ "meta.quality",
469
+ "meta.speed",
470
+ "meta.control_mode",
471
+ "meta.embodiment_tag",
472
+ "meta.failure_type",
473
+ "meta.scene_type",
474
+ "meta.data_source",
475
+ "meta.operator_id",
476
+ "meta.task_description",
477
+ "meta.episode_length",
478
+ "meta.number_of_tasks"
479
+ ]
480
+ }
481
+ },
482
+ "decoupled_wbc_h1": {
483
+ "dataset_name": "decoupled_wbc_h1",
484
+ "goal_sampling": false,
485
+ "goal_future_offsets": [
486
+ 1.0,
487
+ 2.0,
488
+ 3.0,
489
+ 4.0
490
+ ],
491
+ "goal_end_segment_prob": 0.5,
492
+ "use_paraphrases": false,
493
+ "image_keys": [
494
+ "video.egocentric"
495
+ ],
496
+ "state_key": [
497
+ "state.joint_positions"
498
+ ],
499
+ "action_key": [
500
+ "action.hand_joints",
501
+ "action.arm_joints",
502
+ "action.torso_rpy",
503
+ "action.base_height",
504
+ "action.base_vx",
505
+ "action.base_vy",
506
+ "action.base_vyaw",
507
+ "action.target_yaw"
508
+ ],
509
+ "subgoal_key": [
510
+ "subgoal.egocentric"
511
+ ],
512
+ "instruction_key": [
513
+ "annotation.human_instruction"
514
+ ],
515
+ "subtask_prompt_key": [
516
+ "annotation.sub_task"
517
+ ],
518
+ "memory_text_key": [
519
+ "annotation.memory_desc"
520
+ ],
521
+ "sub_task_delta_key": [
522
+ "annotation.sub_task_delta"
523
+ ],
524
+ "next_subtask_key": [
525
+ "annotation.next_sub_task"
526
+ ],
527
+ "prev_subtask_key": [
528
+ "annotation.prev_sub_task"
529
+ ],
530
+ "memory_items_key": [
531
+ "annotation.memory_items"
532
+ ],
533
+ "meta_keys": [
534
+ "meta.quality",
535
+ "meta.speed",
536
+ "meta.control_mode",
537
+ "meta.embodiment_tag",
538
+ "meta.failure_type",
539
+ "meta.scene_type",
540
+ "meta.data_source",
541
+ "meta.operator_id",
542
+ "meta.task_description",
543
+ "meta.episode_length",
544
+ "meta.number_of_tasks"
545
+ ],
546
+ "num_past_frames": 0,
547
+ "num_past_state_frames": 0,
548
+ "num_past_image_frames": 0,
549
+ "action_chunk_size": 30,
550
+ "pad_action_dim": 80,
551
+ "pad_state_dim": 45,
552
+ "video": {
553
+ "delta_indices": [
554
+ 0
555
+ ],
556
+ "eval_delta_indices": [
557
+ 0
558
+ ],
559
+ "modality_keys": [
560
+ "video.egocentric"
561
+ ]
562
+ },
563
+ "state": {
564
+ "delta_indices": [
565
+ 0
566
+ ],
567
+ "eval_delta_indices": [
568
+ 0
569
+ ],
570
+ "modality_keys": [
571
+ "state.joint_positions"
572
+ ]
573
+ },
574
+ "action": {
575
+ "delta_indices": [
576
+ 0,
577
+ 1,
578
+ 2,
579
+ 3,
580
+ 4,
581
+ 5,
582
+ 6,
583
+ 7,
584
+ 8,
585
+ 9,
586
+ 10,
587
+ 11,
588
+ 12,
589
+ 13,
590
+ 14,
591
+ 15,
592
+ 16,
593
+ 17,
594
+ 18,
595
+ 19,
596
+ 20,
597
+ 21,
598
+ 22,
599
+ 23,
600
+ 24,
601
+ 25,
602
+ 26,
603
+ 27,
604
+ 28,
605
+ 29
606
+ ],
607
+ "eval_delta_indices": [
608
+ 0,
609
+ 1,
610
+ 2,
611
+ 3,
612
+ 4,
613
+ 5,
614
+ 6,
615
+ 7,
616
+ 8,
617
+ 9,
618
+ 10,
619
+ 11,
620
+ 12,
621
+ 13,
622
+ 14,
623
+ 15,
624
+ 16,
625
+ 17,
626
+ 18,
627
+ 19,
628
+ 20,
629
+ 21,
630
+ 22,
631
+ 23,
632
+ 24,
633
+ 25,
634
+ 26,
635
+ 27,
636
+ 28,
637
+ 29,
638
+ 30,
639
+ 31,
640
+ 32,
641
+ 33,
642
+ 34,
643
+ 35
644
+ ],
645
+ "modality_keys": [
646
+ "action.hand_joints",
647
+ "action.arm_joints",
648
+ "action.torso_rpy",
649
+ "action.base_height",
650
+ "action.base_vx",
651
+ "action.base_vy",
652
+ "action.base_vyaw",
653
+ "action.target_yaw"
654
+ ]
655
+ },
656
+ "language": {
657
+ "delta_indices": [
658
+ 0
659
+ ],
660
+ "eval_delta_indices": [
661
+ 0
662
+ ],
663
+ "modality_keys": [
664
+ "annotation.human_instruction",
665
+ "annotation.sub_task",
666
+ "annotation.memory_desc",
667
+ "annotation.sub_task_delta",
668
+ "annotation.next_sub_task",
669
+ "annotation.prev_sub_task",
670
+ "annotation.memory_items"
671
+ ]
672
+ },
673
+ "subgoal": {
674
+ "delta_indices": [
675
+ 0
676
+ ],
677
+ "eval_delta_indices": [
678
+ 0
679
+ ],
680
+ "modality_keys": [
681
+ "subgoal.egocentric"
682
+ ],
683
+ "kwargs": {
684
+ "subgoal_prob": 0.25,
685
+ "sample_from_gt_goal_prob": 0.25,
686
+ "future_horizon": [
687
+ 30,
688
+ 120
689
+ ]
690
+ }
691
+ },
692
+ "meta": {
693
+ "delta_indices": [
694
+ 0
695
+ ],
696
+ "eval_delta_indices": [
697
+ 0
698
+ ],
699
+ "modality_keys": [
700
+ "meta.quality",
701
+ "meta.speed",
702
+ "meta.control_mode",
703
+ "meta.embodiment_tag",
704
+ "meta.failure_type",
705
+ "meta.scene_type",
706
+ "meta.data_source",
707
+ "meta.operator_id",
708
+ "meta.task_description",
709
+ "meta.episode_length",
710
+ "meta.number_of_tasks"
711
+ ]
712
+ }
713
+ },
714
+ "psix_he_g1_sonic": {
715
+ "dataset_name": "psix_he_g1_sonic",
716
+ "goal_sampling": false,
717
+ "goal_future_offsets": [
718
+ 1.0,
719
+ 2.0,
720
+ 3.0,
721
+ 4.0
722
+ ],
723
+ "goal_end_segment_prob": 0.5,
724
+ "use_paraphrases": false,
725
+ "image_keys": [
726
+ "video.egocentric"
727
+ ],
728
+ "state_key": [
729
+ "state.joint_positions"
730
+ ],
731
+ "action_key": [
732
+ "action.hand_joints",
733
+ "action.body_token"
734
+ ],
735
+ "subgoal_key": [
736
+ "subgoal.egocentric"
737
+ ],
738
+ "instruction_key": [
739
+ "annotation.human_instruction"
740
+ ],
741
+ "subtask_prompt_key": [
742
+ "annotation.sub_task"
743
+ ],
744
+ "memory_text_key": [
745
+ "annotation.memory_desc"
746
+ ],
747
+ "sub_task_delta_key": [
748
+ "annotation.sub_task_delta"
749
+ ],
750
+ "next_subtask_key": [
751
+ "annotation.next_sub_task"
752
+ ],
753
+ "prev_subtask_key": [
754
+ "annotation.prev_sub_task"
755
+ ],
756
+ "memory_items_key": [
757
+ "annotation.memory_items"
758
+ ],
759
+ "meta_keys": [
760
+ "meta.quality",
761
+ "meta.speed",
762
+ "meta.control_mode",
763
+ "meta.embodiment_tag",
764
+ "meta.failure_type",
765
+ "meta.scene_type",
766
+ "meta.data_source",
767
+ "meta.operator_id",
768
+ "meta.task_description",
769
+ "meta.episode_length",
770
+ "meta.number_of_tasks"
771
+ ],
772
+ "num_past_frames": 0,
773
+ "history_stride": 1,
774
+ "num_past_state_frames": 0,
775
+ "num_past_image_frames": 0,
776
+ "action_chunk_size": 30,
777
+ "pad_action_dim": 80,
778
+ "pad_state_dim": 45,
779
+ "subgoal_prob": 0.25,
780
+ "sample_from_gt_goal_prob": 0.25,
781
+ "future_horizon_min": 30,
782
+ "future_horizon_max": 120,
783
+ "val_future_horizon": null,
784
+ "wm_goal_video_key": null,
785
+ "wm_goal_prob": 0.0,
786
+ "wm_goal_horizon": 48,
787
+ "wm_source_prob": 0.0,
788
+ "video": {
789
+ "delta_indices": [
790
+ 0
791
+ ],
792
+ "eval_delta_indices": [
793
+ 0
794
+ ],
795
+ "modality_keys": [
796
+ "video.egocentric"
797
+ ]
798
+ },
799
+ "state": {
800
+ "delta_indices": [
801
+ 0
802
+ ],
803
+ "eval_delta_indices": [
804
+ 0
805
+ ],
806
+ "modality_keys": [
807
+ "state.joint_positions"
808
+ ]
809
+ },
810
+ "action": {
811
+ "delta_indices": [
812
+ 0,
813
+ 1,
814
+ 2,
815
+ 3,
816
+ 4,
817
+ 5,
818
+ 6,
819
+ 7,
820
+ 8,
821
+ 9,
822
+ 10,
823
+ 11,
824
+ 12,
825
+ 13,
826
+ 14,
827
+ 15,
828
+ 16,
829
+ 17,
830
+ 18,
831
+ 19,
832
+ 20,
833
+ 21,
834
+ 22,
835
+ 23,
836
+ 24,
837
+ 25,
838
+ 26,
839
+ 27,
840
+ 28,
841
+ 29
842
+ ],
843
+ "eval_delta_indices": [
844
+ 0,
845
+ 1,
846
+ 2,
847
+ 3,
848
+ 4,
849
+ 5,
850
+ 6,
851
+ 7,
852
+ 8,
853
+ 9,
854
+ 10,
855
+ 11,
856
+ 12,
857
+ 13,
858
+ 14,
859
+ 15,
860
+ 16,
861
+ 17,
862
+ 18,
863
+ 19,
864
+ 20,
865
+ 21,
866
+ 22,
867
+ 23,
868
+ 24,
869
+ 25,
870
+ 26,
871
+ 27,
872
+ 28,
873
+ 29,
874
+ 30,
875
+ 31,
876
+ 32,
877
+ 33,
878
+ 34,
879
+ 35
880
+ ],
881
+ "modality_keys": [
882
+ "action.hand_joints",
883
+ "action.body_token"
884
+ ]
885
+ },
886
+ "language": {
887
+ "delta_indices": [
888
+ 0
889
+ ],
890
+ "eval_delta_indices": [
891
+ 0
892
+ ],
893
+ "modality_keys": [
894
+ "annotation.human_instruction",
895
+ "annotation.sub_task",
896
+ "annotation.memory_desc",
897
+ "annotation.sub_task_delta",
898
+ "annotation.next_sub_task",
899
+ "annotation.prev_sub_task",
900
+ "annotation.memory_items"
901
+ ]
902
+ },
903
+ "subgoal": {
904
+ "delta_indices": [
905
+ 0
906
+ ],
907
+ "eval_delta_indices": [
908
+ 0
909
+ ],
910
+ "modality_keys": [
911
+ "subgoal.egocentric"
912
+ ],
913
+ "kwargs": {
914
+ "subgoal_prob": 0.25,
915
+ "sample_from_gt_goal_prob": 0.25,
916
+ "future_horizon": [
917
+ 30,
918
+ 120
919
+ ],
920
+ "val_future_horizon": null,
921
+ "wm_goal_video_key": null,
922
+ "wm_goal_prob": 0.0,
923
+ "wm_goal_horizon": 48,
924
+ "wm_source_prob": 0.0
925
+ }
926
+ },
927
+ "meta": {
928
+ "delta_indices": [
929
+ 0
930
+ ],
931
+ "eval_delta_indices": [
932
+ 0
933
+ ],
934
+ "modality_keys": [
935
+ "meta.quality",
936
+ "meta.speed",
937
+ "meta.control_mode",
938
+ "meta.embodiment_tag",
939
+ "meta.failure_type",
940
+ "meta.scene_type",
941
+ "meta.data_source",
942
+ "meta.operator_id",
943
+ "meta.task_description",
944
+ "meta.episode_length",
945
+ "meta.number_of_tasks"
946
+ ]
947
+ }
948
+ },
949
+ "psix_he_h1_sonic": {
950
+ "dataset_name": "psix_he_h1_sonic",
951
+ "goal_sampling": false,
952
+ "goal_future_offsets": [
953
+ 1.0,
954
+ 2.0,
955
+ 3.0,
956
+ 4.0
957
+ ],
958
+ "goal_end_segment_prob": 0.5,
959
+ "use_paraphrases": false,
960
+ "image_keys": [
961
+ "video.egocentric"
962
+ ],
963
+ "state_key": [
964
+ "state.joint_positions"
965
+ ],
966
+ "action_key": [
967
+ "action.hand_joints",
968
+ "action.body_token"
969
+ ],
970
+ "subgoal_key": [
971
+ "subgoal.egocentric"
972
+ ],
973
+ "instruction_key": [
974
+ "annotation.human_instruction"
975
+ ],
976
+ "subtask_prompt_key": [
977
+ "annotation.sub_task"
978
+ ],
979
+ "memory_text_key": [
980
+ "annotation.memory_desc"
981
+ ],
982
+ "sub_task_delta_key": [
983
+ "annotation.sub_task_delta"
984
+ ],
985
+ "next_subtask_key": [
986
+ "annotation.next_sub_task"
987
+ ],
988
+ "prev_subtask_key": [
989
+ "annotation.prev_sub_task"
990
+ ],
991
+ "memory_items_key": [
992
+ "annotation.memory_items"
993
+ ],
994
+ "meta_keys": [
995
+ "meta.quality",
996
+ "meta.speed",
997
+ "meta.control_mode",
998
+ "meta.embodiment_tag",
999
+ "meta.failure_type",
1000
+ "meta.scene_type",
1001
+ "meta.data_source",
1002
+ "meta.operator_id",
1003
+ "meta.task_description",
1004
+ "meta.episode_length",
1005
+ "meta.number_of_tasks"
1006
+ ],
1007
+ "num_past_frames": 0,
1008
+ "history_stride": 1,
1009
+ "num_past_state_frames": 0,
1010
+ "num_past_image_frames": 0,
1011
+ "action_chunk_size": 30,
1012
+ "pad_action_dim": 80,
1013
+ "pad_state_dim": 45,
1014
+ "hand_pad_to_g1": true,
1015
+ "subgoal_prob": 0.25,
1016
+ "sample_from_gt_goal_prob": 0.25,
1017
+ "future_horizon_min": 30,
1018
+ "future_horizon_max": 120,
1019
+ "val_future_horizon": null,
1020
+ "wm_goal_video_key": null,
1021
+ "wm_goal_prob": 0.0,
1022
+ "wm_goal_horizon": 48,
1023
+ "wm_source_prob": 0.0,
1024
+ "video": {
1025
+ "delta_indices": [
1026
+ 0
1027
+ ],
1028
+ "eval_delta_indices": [
1029
+ 0
1030
+ ],
1031
+ "modality_keys": [
1032
+ "video.egocentric"
1033
+ ]
1034
+ },
1035
+ "state": {
1036
+ "delta_indices": [
1037
+ 0
1038
+ ],
1039
+ "eval_delta_indices": [
1040
+ 0
1041
+ ],
1042
+ "modality_keys": [
1043
+ "state.joint_positions"
1044
+ ]
1045
+ },
1046
+ "action": {
1047
+ "delta_indices": [
1048
+ 0,
1049
+ 1,
1050
+ 2,
1051
+ 3,
1052
+ 4,
1053
+ 5,
1054
+ 6,
1055
+ 7,
1056
+ 8,
1057
+ 9,
1058
+ 10,
1059
+ 11,
1060
+ 12,
1061
+ 13,
1062
+ 14,
1063
+ 15,
1064
+ 16,
1065
+ 17,
1066
+ 18,
1067
+ 19,
1068
+ 20,
1069
+ 21,
1070
+ 22,
1071
+ 23,
1072
+ 24,
1073
+ 25,
1074
+ 26,
1075
+ 27,
1076
+ 28,
1077
+ 29
1078
+ ],
1079
+ "eval_delta_indices": [
1080
+ 0,
1081
+ 1,
1082
+ 2,
1083
+ 3,
1084
+ 4,
1085
+ 5,
1086
+ 6,
1087
+ 7,
1088
+ 8,
1089
+ 9,
1090
+ 10,
1091
+ 11,
1092
+ 12,
1093
+ 13,
1094
+ 14,
1095
+ 15,
1096
+ 16,
1097
+ 17,
1098
+ 18,
1099
+ 19,
1100
+ 20,
1101
+ 21,
1102
+ 22,
1103
+ 23,
1104
+ 24,
1105
+ 25,
1106
+ 26,
1107
+ 27,
1108
+ 28,
1109
+ 29,
1110
+ 30,
1111
+ 31,
1112
+ 32,
1113
+ 33,
1114
+ 34,
1115
+ 35
1116
+ ],
1117
+ "modality_keys": [
1118
+ "action.hand_joints",
1119
+ "action.body_token"
1120
+ ]
1121
+ },
1122
+ "language": {
1123
+ "delta_indices": [
1124
+ 0
1125
+ ],
1126
+ "eval_delta_indices": [
1127
+ 0
1128
+ ],
1129
+ "modality_keys": [
1130
+ "annotation.human_instruction",
1131
+ "annotation.sub_task",
1132
+ "annotation.memory_desc",
1133
+ "annotation.sub_task_delta",
1134
+ "annotation.next_sub_task",
1135
+ "annotation.prev_sub_task",
1136
+ "annotation.memory_items"
1137
+ ]
1138
+ },
1139
+ "subgoal": {
1140
+ "delta_indices": [
1141
+ 0
1142
+ ],
1143
+ "eval_delta_indices": [
1144
+ 0
1145
+ ],
1146
+ "modality_keys": [
1147
+ "subgoal.egocentric"
1148
+ ],
1149
+ "kwargs": {
1150
+ "subgoal_prob": 0.25,
1151
+ "sample_from_gt_goal_prob": 0.25,
1152
+ "future_horizon": [
1153
+ 30,
1154
+ 120
1155
+ ],
1156
+ "val_future_horizon": null,
1157
+ "wm_goal_video_key": null,
1158
+ "wm_goal_prob": 0.0,
1159
+ "wm_goal_horizon": 48,
1160
+ "wm_source_prob": 0.0
1161
+ }
1162
+ },
1163
+ "meta": {
1164
+ "delta_indices": [
1165
+ 0
1166
+ ],
1167
+ "eval_delta_indices": [
1168
+ 0
1169
+ ],
1170
+ "modality_keys": [
1171
+ "meta.quality",
1172
+ "meta.speed",
1173
+ "meta.control_mode",
1174
+ "meta.embodiment_tag",
1175
+ "meta.failure_type",
1176
+ "meta.scene_type",
1177
+ "meta.data_source",
1178
+ "meta.operator_id",
1179
+ "meta.task_description",
1180
+ "meta.episode_length",
1181
+ "meta.number_of_tasks"
1182
+ ]
1183
+ }
1184
+ },
1185
+ "psix_g1_sonic_neck": {
1186
+ "dataset_name": "psix_g1_sonic_neck",
1187
+ "goal_sampling": false,
1188
+ "goal_future_offsets": [
1189
+ 1.0,
1190
+ 2.0,
1191
+ 3.0,
1192
+ 4.0
1193
+ ],
1194
+ "goal_end_segment_prob": 0.5,
1195
+ "use_paraphrases": false,
1196
+ "image_keys": [
1197
+ "video.egocentric"
1198
+ ],
1199
+ "state_key": [
1200
+ "state.joint_positions"
1201
+ ],
1202
+ "action_key": [
1203
+ "action.hand_joints",
1204
+ "action.body_token_v1_1",
1205
+ "action.neck"
1206
+ ],
1207
+ "subgoal_key": [
1208
+ "subgoal.egocentric"
1209
+ ],
1210
+ "instruction_key": [
1211
+ "annotation.human_instruction"
1212
+ ],
1213
+ "subtask_prompt_key": [
1214
+ "annotation.sub_task"
1215
+ ],
1216
+ "memory_text_key": [
1217
+ "annotation.memory_desc"
1218
+ ],
1219
+ "sub_task_delta_key": [
1220
+ "annotation.sub_task_delta"
1221
+ ],
1222
+ "next_subtask_key": [
1223
+ "annotation.next_sub_task"
1224
+ ],
1225
+ "prev_subtask_key": [
1226
+ "annotation.prev_sub_task"
1227
+ ],
1228
+ "memory_items_key": [
1229
+ "annotation.memory_items"
1230
+ ],
1231
+ "meta_keys": [
1232
+ "meta.quality",
1233
+ "meta.speed",
1234
+ "meta.control_mode",
1235
+ "meta.embodiment_tag",
1236
+ "meta.failure_type",
1237
+ "meta.scene_type",
1238
+ "meta.data_source",
1239
+ "meta.operator_id",
1240
+ "meta.task_description",
1241
+ "meta.episode_length",
1242
+ "meta.number_of_tasks"
1243
+ ],
1244
+ "num_past_frames": 0,
1245
+ "history_stride": 1,
1246
+ "num_past_state_frames": 0,
1247
+ "num_past_image_frames": 0,
1248
+ "action_chunk_size": 30,
1249
+ "pad_action_dim": 80,
1250
+ "pad_state_dim": 45,
1251
+ "subgoal_prob": 1.0,
1252
+ "sample_from_gt_goal_prob": 0.25,
1253
+ "future_horizon_min": 30,
1254
+ "future_horizon_max": 120,
1255
+ "val_future_horizon": 48,
1256
+ "wm_goal_video_key": "video.egocentric_wm",
1257
+ "wm_goal_prob": 0.0,
1258
+ "wm_goal_horizon": 48,
1259
+ "wm_source_prob": 0.5,
1260
+ "video": {
1261
+ "delta_indices": [
1262
+ 0
1263
+ ],
1264
+ "eval_delta_indices": [
1265
+ 0
1266
+ ],
1267
+ "modality_keys": [
1268
+ "video.egocentric",
1269
+ "video.egocentric_wm"
1270
+ ]
1271
+ },
1272
+ "state": {
1273
+ "delta_indices": [
1274
+ 0
1275
+ ],
1276
+ "eval_delta_indices": [
1277
+ 0
1278
+ ],
1279
+ "modality_keys": [
1280
+ "state.joint_positions"
1281
+ ]
1282
+ },
1283
+ "action": {
1284
+ "delta_indices": [
1285
+ 0,
1286
+ 1,
1287
+ 2,
1288
+ 3,
1289
+ 4,
1290
+ 5,
1291
+ 6,
1292
+ 7,
1293
+ 8,
1294
+ 9,
1295
+ 10,
1296
+ 11,
1297
+ 12,
1298
+ 13,
1299
+ 14,
1300
+ 15,
1301
+ 16,
1302
+ 17,
1303
+ 18,
1304
+ 19,
1305
+ 20,
1306
+ 21,
1307
+ 22,
1308
+ 23,
1309
+ 24,
1310
+ 25,
1311
+ 26,
1312
+ 27,
1313
+ 28,
1314
+ 29
1315
+ ],
1316
+ "eval_delta_indices": [
1317
+ 0,
1318
+ 1,
1319
+ 2,
1320
+ 3,
1321
+ 4,
1322
+ 5,
1323
+ 6,
1324
+ 7,
1325
+ 8,
1326
+ 9,
1327
+ 10,
1328
+ 11,
1329
+ 12,
1330
+ 13,
1331
+ 14,
1332
+ 15,
1333
+ 16,
1334
+ 17,
1335
+ 18,
1336
+ 19,
1337
+ 20,
1338
+ 21,
1339
+ 22,
1340
+ 23,
1341
+ 24,
1342
+ 25,
1343
+ 26,
1344
+ 27,
1345
+ 28,
1346
+ 29,
1347
+ 30,
1348
+ 31,
1349
+ 32,
1350
+ 33,
1351
+ 34,
1352
+ 35
1353
+ ],
1354
+ "modality_keys": [
1355
+ "action.hand_joints",
1356
+ "action.body_token_v1_1",
1357
+ "action.neck"
1358
+ ]
1359
+ },
1360
+ "language": {
1361
+ "delta_indices": [
1362
+ 0
1363
+ ],
1364
+ "eval_delta_indices": [
1365
+ 0
1366
+ ],
1367
+ "modality_keys": [
1368
+ "annotation.human_instruction",
1369
+ "annotation.sub_task",
1370
+ "annotation.memory_desc",
1371
+ "annotation.sub_task_delta",
1372
+ "annotation.next_sub_task",
1373
+ "annotation.prev_sub_task",
1374
+ "annotation.memory_items"
1375
+ ]
1376
+ },
1377
+ "subgoal": {
1378
+ "delta_indices": [
1379
+ 0
1380
+ ],
1381
+ "eval_delta_indices": [
1382
+ 0
1383
+ ],
1384
+ "modality_keys": [
1385
+ "subgoal.egocentric"
1386
+ ],
1387
+ "kwargs": {
1388
+ "subgoal_prob": 1.0,
1389
+ "sample_from_gt_goal_prob": 0.25,
1390
+ "future_horizon": [
1391
+ 30,
1392
+ 120
1393
+ ],
1394
+ "val_future_horizon": 48,
1395
+ "wm_goal_video_key": "video.egocentric_wm",
1396
+ "wm_goal_prob": 0.0,
1397
+ "wm_goal_horizon": 48,
1398
+ "wm_source_prob": 0.5
1399
+ }
1400
+ },
1401
+ "meta": {
1402
+ "delta_indices": [
1403
+ 0
1404
+ ],
1405
+ "eval_delta_indices": [
1406
+ 0
1407
+ ],
1408
+ "modality_keys": [
1409
+ "meta.quality",
1410
+ "meta.speed",
1411
+ "meta.control_mode",
1412
+ "meta.embodiment_tag",
1413
+ "meta.failure_type",
1414
+ "meta.scene_type",
1415
+ "meta.data_source",
1416
+ "meta.operator_id",
1417
+ "meta.task_description",
1418
+ "meta.episode_length",
1419
+ "meta.number_of_tasks"
1420
+ ]
1421
+ }
1422
+ },
1423
+ "egodex_psix": {
1424
+ "dataset_name": "egodex_psix",
1425
+ "goal_sampling": false,
1426
+ "goal_future_offsets": [
1427
+ 1.0,
1428
+ 2.0,
1429
+ 3.0,
1430
+ 4.0
1431
+ ],
1432
+ "goal_end_segment_prob": 0.5,
1433
+ "use_paraphrases": false,
1434
+ "image_keys": [
1435
+ "video.egocentric"
1436
+ ],
1437
+ "state_key": [
1438
+ "state.pose"
1439
+ ],
1440
+ "action_key": [
1441
+ "action.pose"
1442
+ ],
1443
+ "subgoal_key": [
1444
+ "subgoal.egocentric"
1445
+ ],
1446
+ "instruction_key": [
1447
+ "annotation.human_instruction"
1448
+ ],
1449
+ "subtask_prompt_key": [
1450
+ "annotation.sub_task"
1451
+ ],
1452
+ "memory_text_key": [
1453
+ "annotation.memory_desc"
1454
+ ],
1455
+ "sub_task_delta_key": [
1456
+ "annotation.sub_task_delta"
1457
+ ],
1458
+ "next_subtask_key": [
1459
+ "annotation.next_sub_task"
1460
+ ],
1461
+ "prev_subtask_key": [
1462
+ "annotation.prev_sub_task"
1463
+ ],
1464
+ "memory_items_key": [
1465
+ "annotation.memory_items"
1466
+ ],
1467
+ "meta_keys": [
1468
+ "meta.quality",
1469
+ "meta.speed",
1470
+ "meta.control_mode",
1471
+ "meta.embodiment_tag",
1472
+ "meta.failure_type",
1473
+ "meta.scene_type",
1474
+ "meta.data_source",
1475
+ "meta.operator_id",
1476
+ "meta.task_description",
1477
+ "meta.episode_length",
1478
+ "meta.number_of_tasks"
1479
+ ],
1480
+ "num_past_frames": 0,
1481
+ "history_stride": 1,
1482
+ "num_past_state_frames": 0,
1483
+ "num_past_image_frames": 0,
1484
+ "action_chunk_size": 31,
1485
+ "pad_action_dim": 80,
1486
+ "pad_state_dim": 45,
1487
+ "subgoal_prob": 0.25,
1488
+ "sample_from_gt_goal_prob": 0.25,
1489
+ "future_horizon_min": 30,
1490
+ "future_horizon_max": 120,
1491
+ "val_future_horizon": null,
1492
+ "wm_goal_video_key": null,
1493
+ "wm_goal_prob": 0.0,
1494
+ "wm_goal_horizon": 48,
1495
+ "wm_source_prob": 0.0,
1496
+ "video": {
1497
+ "delta_indices": [
1498
+ 0
1499
+ ],
1500
+ "eval_delta_indices": [
1501
+ 0
1502
+ ],
1503
+ "modality_keys": [
1504
+ "video.egocentric"
1505
+ ]
1506
+ },
1507
+ "state": {
1508
+ "delta_indices": [
1509
+ 0
1510
+ ],
1511
+ "eval_delta_indices": [
1512
+ 0
1513
+ ],
1514
+ "modality_keys": [
1515
+ "state.pose",
1516
+ "state.camera_extrinsic"
1517
+ ]
1518
+ },
1519
+ "action": {
1520
+ "delta_indices": [
1521
+ 0,
1522
+ 1,
1523
+ 2,
1524
+ 3,
1525
+ 4,
1526
+ 5,
1527
+ 6,
1528
+ 7,
1529
+ 8,
1530
+ 9,
1531
+ 10,
1532
+ 11,
1533
+ 12,
1534
+ 13,
1535
+ 14,
1536
+ 15,
1537
+ 16,
1538
+ 17,
1539
+ 18,
1540
+ 19,
1541
+ 20,
1542
+ 21,
1543
+ 22,
1544
+ 23,
1545
+ 24,
1546
+ 25,
1547
+ 26,
1548
+ 27,
1549
+ 28,
1550
+ 29,
1551
+ 30
1552
+ ],
1553
+ "eval_delta_indices": [
1554
+ 0,
1555
+ 1,
1556
+ 2,
1557
+ 3,
1558
+ 4,
1559
+ 5,
1560
+ 6,
1561
+ 7,
1562
+ 8,
1563
+ 9,
1564
+ 10,
1565
+ 11,
1566
+ 12,
1567
+ 13,
1568
+ 14,
1569
+ 15,
1570
+ 16,
1571
+ 17,
1572
+ 18,
1573
+ 19,
1574
+ 20,
1575
+ 21,
1576
+ 22,
1577
+ 23,
1578
+ 24,
1579
+ 25,
1580
+ 26,
1581
+ 27,
1582
+ 28,
1583
+ 29,
1584
+ 30,
1585
+ 31,
1586
+ 32,
1587
+ 33,
1588
+ 34,
1589
+ 35
1590
+ ],
1591
+ "modality_keys": [
1592
+ "action.pose"
1593
+ ]
1594
+ },
1595
+ "language": {
1596
+ "delta_indices": [
1597
+ 0
1598
+ ],
1599
+ "eval_delta_indices": [
1600
+ 0
1601
+ ],
1602
+ "modality_keys": [
1603
+ "annotation.human_instruction",
1604
+ "annotation.sub_task",
1605
+ "annotation.memory_desc",
1606
+ "annotation.sub_task_delta",
1607
+ "annotation.next_sub_task",
1608
+ "annotation.prev_sub_task",
1609
+ "annotation.memory_items"
1610
+ ]
1611
+ },
1612
+ "subgoal": {
1613
+ "delta_indices": [
1614
+ 0
1615
+ ],
1616
+ "eval_delta_indices": [
1617
+ 0
1618
+ ],
1619
+ "modality_keys": [
1620
+ "subgoal.egocentric"
1621
+ ],
1622
+ "kwargs": {
1623
+ "subgoal_prob": 0.25,
1624
+ "sample_from_gt_goal_prob": 0.25,
1625
+ "future_horizon": [
1626
+ 30,
1627
+ 120
1628
+ ],
1629
+ "val_future_horizon": null,
1630
+ "wm_goal_video_key": null,
1631
+ "wm_goal_prob": 0.0,
1632
+ "wm_goal_horizon": 48,
1633
+ "wm_source_prob": 0.0
1634
+ }
1635
+ },
1636
+ "meta": {
1637
+ "delta_indices": [
1638
+ 0
1639
+ ],
1640
+ "eval_delta_indices": [
1641
+ 0
1642
+ ],
1643
+ "modality_keys": [
1644
+ "meta.quality",
1645
+ "meta.speed",
1646
+ "meta.control_mode",
1647
+ "meta.embodiment_tag",
1648
+ "meta.failure_type",
1649
+ "meta.scene_type",
1650
+ "meta.data_source",
1651
+ "meta.operator_id",
1652
+ "meta.task_description",
1653
+ "meta.episode_length",
1654
+ "meta.number_of_tasks"
1655
+ ]
1656
+ },
1657
+ "action_stride": 1,
1658
+ "variable_horizon": false
1659
+ }
1660
+ },
1661
+ "model": {
1662
+ "resize": {
1663
+ "size": [
1664
+ 270,
1665
+ 480
1666
+ ]
1667
+ },
1668
+ "center_crop": {
1669
+ "size": [
1670
+ 270,
1671
+ 480
1672
+ ]
1673
+ },
1674
+ "color_jitter": {
1675
+ "brightness": 0.2,
1676
+ "contrast": [
1677
+ 0.8,
1678
+ 1.2
1679
+ ],
1680
+ "saturation": [
1681
+ 0.8,
1682
+ 1.2
1683
+ ],
1684
+ "hue": 0.05
1685
+ },
1686
+ "gaussian_noise": {
1687
+ "mean": 0.0,
1688
+ "std": 3.0,
1689
+ "prob_skip": 0.1
1690
+ },
1691
+ "img_aug": true,
1692
+ "min_pixels": 4096,
1693
+ "max_pixels": 16777216,
1694
+ "adaptive_resize": false,
1695
+ "use_subtask_text": true,
1696
+ "drop_subtask_text_prob": 0.5,
1697
+ "subtask_only": false
1698
+ },
1699
+ "field": {
1700
+ "simple_g1": {
1701
+ "state_to_tensor": {
1702
+ "apply_to": [],
1703
+ "training": true,
1704
+ "input_dtypes": {},
1705
+ "output_dtypes": {}
1706
+ },
1707
+ "normalize_state": true,
1708
+ "pad_action_dim": null,
1709
+ "pad_state_dim": null,
1710
+ "state_transform": {
1711
+ "apply_to": [],
1712
+ "training": true,
1713
+ "normalization_modes": {
1714
+ "state.left_hand": "min_max",
1715
+ "state.right_hand": "min_max",
1716
+ "state.left_arm": "min_max",
1717
+ "state.right_arm": "min_max",
1718
+ "state.rpy": "min_max",
1719
+ "state.height": "min_max"
1720
+ },
1721
+ "target_rotations": {},
1722
+ "normalization_statistics": {},
1723
+ "modality_metadata": {}
1724
+ },
1725
+ "action_to_tensor": {
1726
+ "apply_to": [],
1727
+ "training": true,
1728
+ "input_dtypes": {},
1729
+ "output_dtypes": {}
1730
+ },
1731
+ "action_transform": {
1732
+ "apply_to": [],
1733
+ "training": true,
1734
+ "normalization_modes": {
1735
+ "action.left_hand": "min_max",
1736
+ "action.right_hand": "min_max",
1737
+ "action.left_arm": "min_max",
1738
+ "action.right_arm": "min_max",
1739
+ "action.rpy": "min_max",
1740
+ "action.height": "min_max",
1741
+ "action.torso_vx": "min_max",
1742
+ "action.torso_vy": "min_max",
1743
+ "action.torso_vyaw": "min_max",
1744
+ "action.target_yaw": "min_max"
1745
+ },
1746
+ "target_rotations": {},
1747
+ "normalization_statistics": {},
1748
+ "modality_metadata": {}
1749
+ },
1750
+ "concat": {
1751
+ "apply_to": [],
1752
+ "training": true,
1753
+ "video_concat_order": [],
1754
+ "subgoal_concat_order": null,
1755
+ "state_concat_order": [],
1756
+ "action_concat_order": [],
1757
+ "action_dims": {},
1758
+ "state_dims": {},
1759
+ "action_dims_post_transform": {},
1760
+ "state_dims_post_transform": {}
1761
+ }
1762
+ },
1763
+ "decoupled_wbc_g1": {
1764
+ "video_apply_to": [
1765
+ "video.egocentric"
1766
+ ],
1767
+ "state_norm_mode": "min_max",
1768
+ "action_norm_mode": "min_max",
1769
+ "state_to_tensor": {
1770
+ "apply_to": [],
1771
+ "training": true,
1772
+ "input_dtypes": {},
1773
+ "output_dtypes": {}
1774
+ },
1775
+ "state_transform": {
1776
+ "apply_to": [],
1777
+ "training": true,
1778
+ "normalization_modes": {},
1779
+ "target_rotations": {},
1780
+ "normalization_statistics": {},
1781
+ "modality_metadata": {}
1782
+ },
1783
+ "action_to_tensor": {
1784
+ "apply_to": [],
1785
+ "training": true,
1786
+ "input_dtypes": {},
1787
+ "output_dtypes": {}
1788
+ },
1789
+ "action_transform": {
1790
+ "apply_to": [],
1791
+ "training": true,
1792
+ "normalization_modes": {},
1793
+ "target_rotations": {},
1794
+ "normalization_statistics": {},
1795
+ "modality_metadata": {}
1796
+ },
1797
+ "concat": {
1798
+ "apply_to": [],
1799
+ "training": true,
1800
+ "video_concat_order": [],
1801
+ "subgoal_concat_order": [],
1802
+ "state_concat_order": [],
1803
+ "action_concat_order": [],
1804
+ "action_dims": {},
1805
+ "state_dims": {},
1806
+ "action_dims_post_transform": {},
1807
+ "state_dims_post_transform": {}
1808
+ }
1809
+ },
1810
+ "decoupled_wbc_h1": {
1811
+ "video_apply_to": [
1812
+ "video.egocentric"
1813
+ ],
1814
+ "video_to_tensor": {
1815
+ "apply_to": [],
1816
+ "training": true,
1817
+ "backend": "torchvision",
1818
+ "output_on_cuda": false
1819
+ },
1820
+ "video_crop": {
1821
+ "apply_to": [],
1822
+ "training": true,
1823
+ "backend": "torchvision",
1824
+ "height": null,
1825
+ "width": null,
1826
+ "scale": 0.95
1827
+ },
1828
+ "video_resize": {
1829
+ "apply_to": [],
1830
+ "training": true,
1831
+ "backend": "torchvision",
1832
+ "height": 176,
1833
+ "width": 320,
1834
+ "interpolation": "linear",
1835
+ "antialias": true
1836
+ },
1837
+ "video_color_jitter": {
1838
+ "apply_to": [],
1839
+ "training": true,
1840
+ "backend": "torchvision",
1841
+ "brightness": 0.3,
1842
+ "contrast": 0.4,
1843
+ "saturation": 0.5,
1844
+ "hue": 0.08
1845
+ },
1846
+ "video_to_numpy": {
1847
+ "apply_to": [],
1848
+ "training": true,
1849
+ "backend": "torchvision"
1850
+ },
1851
+ "subgoal_to_tensor": {
1852
+ "apply_to": [],
1853
+ "training": true,
1854
+ "backend": "torchvision",
1855
+ "output_on_cuda": false
1856
+ },
1857
+ "subgoal_resize": {
1858
+ "apply_to": [],
1859
+ "training": true,
1860
+ "backend": "torchvision",
1861
+ "height": 240,
1862
+ "width": 320,
1863
+ "interpolation": "nearest",
1864
+ "antialias": true
1865
+ },
1866
+ "subgoal_to_numpy": {
1867
+ "apply_to": [],
1868
+ "training": true,
1869
+ "backend": "torchvision"
1870
+ },
1871
+ "state_norm_mode": "min_max",
1872
+ "action_norm_mode": "min_max",
1873
+ "state_to_tensor": {
1874
+ "apply_to": [],
1875
+ "training": true,
1876
+ "input_dtypes": {},
1877
+ "output_dtypes": {}
1878
+ },
1879
+ "state_transform": {
1880
+ "apply_to": [],
1881
+ "training": true,
1882
+ "normalization_modes": {},
1883
+ "target_rotations": {},
1884
+ "normalization_statistics": {},
1885
+ "modality_metadata": {}
1886
+ },
1887
+ "action_to_tensor": {
1888
+ "apply_to": [],
1889
+ "training": true,
1890
+ "input_dtypes": {},
1891
+ "output_dtypes": {}
1892
+ },
1893
+ "action_transform": {
1894
+ "apply_to": [],
1895
+ "training": true,
1896
+ "normalization_modes": {},
1897
+ "target_rotations": {},
1898
+ "normalization_statistics": {},
1899
+ "modality_metadata": {}
1900
+ },
1901
+ "concat": {
1902
+ "apply_to": [],
1903
+ "training": true,
1904
+ "video_concat_order": [],
1905
+ "subgoal_concat_order": [],
1906
+ "state_concat_order": [],
1907
+ "action_concat_order": [],
1908
+ "action_dims": {},
1909
+ "state_dims": {},
1910
+ "action_dims_post_transform": {},
1911
+ "state_dims_post_transform": {}
1912
+ }
1913
+ },
1914
+ "psix_he_g1_sonic": {
1915
+ "video_apply_to": [
1916
+ "video.egocentric"
1917
+ ],
1918
+ "state_norm_mode": "min_max",
1919
+ "action_norm_mode": "min_max",
1920
+ "state_to_tensor": {
1921
+ "apply_to": [],
1922
+ "training": true,
1923
+ "input_dtypes": {},
1924
+ "output_dtypes": {}
1925
+ },
1926
+ "state_transform": {
1927
+ "apply_to": [],
1928
+ "training": true,
1929
+ "normalization_modes": {},
1930
+ "target_rotations": {},
1931
+ "normalization_statistics": {},
1932
+ "modality_metadata": {}
1933
+ },
1934
+ "action_to_tensor": {
1935
+ "apply_to": [],
1936
+ "training": true,
1937
+ "input_dtypes": {},
1938
+ "output_dtypes": {}
1939
+ },
1940
+ "action_transform": {
1941
+ "apply_to": [],
1942
+ "training": true,
1943
+ "normalization_modes": {},
1944
+ "target_rotations": {},
1945
+ "normalization_statistics": {},
1946
+ "modality_metadata": {}
1947
+ },
1948
+ "concat": {
1949
+ "apply_to": [],
1950
+ "training": true,
1951
+ "video_concat_order": [],
1952
+ "subgoal_concat_order": [],
1953
+ "state_concat_order": [],
1954
+ "action_concat_order": [],
1955
+ "action_dims": {},
1956
+ "state_dims": {},
1957
+ "action_dims_post_transform": {},
1958
+ "state_dims_post_transform": {}
1959
+ }
1960
+ },
1961
+ "psix_he_h1_sonic": {
1962
+ "video_apply_to": [
1963
+ "video.egocentric"
1964
+ ],
1965
+ "state_norm_mode": "min_max",
1966
+ "action_norm_mode": "min_max",
1967
+ "state_to_tensor": {
1968
+ "apply_to": [],
1969
+ "training": true,
1970
+ "input_dtypes": {},
1971
+ "output_dtypes": {}
1972
+ },
1973
+ "state_transform": {
1974
+ "apply_to": [],
1975
+ "training": true,
1976
+ "normalization_modes": {},
1977
+ "target_rotations": {},
1978
+ "normalization_statistics": {},
1979
+ "modality_metadata": {}
1980
+ },
1981
+ "action_to_tensor": {
1982
+ "apply_to": [],
1983
+ "training": true,
1984
+ "input_dtypes": {},
1985
+ "output_dtypes": {}
1986
+ },
1987
+ "action_transform": {
1988
+ "apply_to": [],
1989
+ "training": true,
1990
+ "normalization_modes": {},
1991
+ "target_rotations": {},
1992
+ "normalization_statistics": {},
1993
+ "modality_metadata": {}
1994
+ },
1995
+ "concat": {
1996
+ "apply_to": [],
1997
+ "training": true,
1998
+ "video_concat_order": [],
1999
+ "subgoal_concat_order": [],
2000
+ "state_concat_order": [],
2001
+ "action_concat_order": [],
2002
+ "action_dims": {},
2003
+ "state_dims": {},
2004
+ "action_dims_post_transform": {},
2005
+ "state_dims_post_transform": {}
2006
+ }
2007
+ },
2008
+ "psix_g1_sonic_neck": {
2009
+ "video_apply_to": [
2010
+ "video.egocentric"
2011
+ ],
2012
+ "state_norm_mode": "min_max",
2013
+ "action_norm_mode": "min_max",
2014
+ "state_to_tensor": {
2015
+ "apply_to": [
2016
+ "state.joint_positions"
2017
+ ],
2018
+ "training": true,
2019
+ "input_dtypes": {},
2020
+ "output_dtypes": {
2021
+ "state.joint_positions": "torch.float32"
2022
+ }
2023
+ },
2024
+ "state_transform": {
2025
+ "apply_to": [
2026
+ "state.joint_positions"
2027
+ ],
2028
+ "training": true,
2029
+ "normalization_modes": {
2030
+ "state.joint_positions": "min_max"
2031
+ },
2032
+ "target_rotations": {},
2033
+ "normalization_statistics": {},
2034
+ "modality_metadata": {}
2035
+ },
2036
+ "action_to_tensor": {
2037
+ "apply_to": [
2038
+ "action.hand_joints",
2039
+ "action.body_token_v1_1",
2040
+ "action.neck"
2041
+ ],
2042
+ "training": true,
2043
+ "input_dtypes": {},
2044
+ "output_dtypes": {
2045
+ "action.hand_joints": "torch.float32",
2046
+ "action.body_token_v1_1": "torch.float32",
2047
+ "action.neck": "torch.float32"
2048
+ }
2049
+ },
2050
+ "action_transform": {
2051
+ "apply_to": [
2052
+ "action.hand_joints",
2053
+ "action.body_token_v1_1",
2054
+ "action.neck"
2055
+ ],
2056
+ "training": true,
2057
+ "normalization_modes": {
2058
+ "action.hand_joints": "min_max",
2059
+ "action.body_token_v1_1": "min_max",
2060
+ "action.neck": "min_max"
2061
+ },
2062
+ "target_rotations": {},
2063
+ "normalization_statistics": {},
2064
+ "modality_metadata": {}
2065
+ },
2066
+ "concat": {
2067
+ "apply_to": [],
2068
+ "training": true,
2069
+ "video_concat_order": [
2070
+ "video.egocentric"
2071
+ ],
2072
+ "subgoal_concat_order": [
2073
+ "subgoal.egocentric"
2074
+ ],
2075
+ "state_concat_order": [
2076
+ "state.joint_positions"
2077
+ ],
2078
+ "action_concat_order": [
2079
+ "action.hand_joints",
2080
+ "action.body_token_v1_1",
2081
+ "action.neck"
2082
+ ],
2083
+ "action_dims": {},
2084
+ "state_dims": {},
2085
+ "action_dims_post_transform": {},
2086
+ "state_dims_post_transform": {}
2087
+ }
2088
+ }
2089
+ }
2090
+ },
2091
+ "label": "psix",
2092
+ "num_workers": 8,
2093
+ "val_no_aug": true,
2094
+ "dataset_paths": [
2095
+ "psix_g1_sonic_neck:/scratch/project/prj-02-spatial-intelligence/jwang/dataset/finetune/teleop5_round2_generic_0909/teleop5_round2_generic_0909_train/g1"
2096
+ ],
2097
+ "dataset_weights": null,
2098
+ "val_dataset_paths": [
2099
+ "psix_g1_sonic_neck:/scratch/project/prj-02-spatial-intelligence/jwang/dataset/finetune/teleop5_round2_generic_0909/teleop5_round2_generic_0909_val/g1"
2100
+ ],
2101
+ "mixture_spec": [
2102
+ {
2103
+ "dataset_path": {
2104
+ "psix_g1_sonic_neck": [
2105
+ "PosixPath('/scratch/project/prj-02-spatial-intelligence/jwang/dataset/finetune/teleop5_round2_generic_0909/teleop5_round2_generic_0909_train/g1')"
2106
+ ]
2107
+ },
2108
+ "dataset_weight": 1.0,
2109
+ "distribute_weights": true
2110
+ }
2111
+ ],
2112
+ "mixture_kwwargs": {
2113
+ "balance_dataset_weights": false,
2114
+ "seed": 42,
2115
+ "shard_sampling_rate": 1.0,
2116
+ "allow_padding": true,
2117
+ "training": true
2118
+ },
2119
+ "dataset_kwargs": {
2120
+ "video_backend": "decord",
2121
+ "video_backend_kwargs": {
2122
+ "num_threads": 1
2123
+ },
2124
+ "use_global_metadata": false,
2125
+ "relative_action": false,
2126
+ "shard_size": 2000
2127
+ },
2128
+ "metadata_versions": {
2129
+ "psi0_sonic_g1": "0221",
2130
+ "simple_g1": "0221",
2131
+ "decoupled_wbc_g1": "0512",
2132
+ "decoupled_wbc_h1": "0512",
2133
+ "psix_he_g1_sonic": "0526",
2134
+ "psix_g1_sonic_neck": "0618",
2135
+ "psix_he_h1_sonic": "0526",
2136
+ "egodex_psix": "0726"
2137
+ },
2138
+ "fps": {
2139
+ "psi0_sonic_g1": 30,
2140
+ "simple_g1": 50,
2141
+ "decoupled_wbc_g1": 30,
2142
+ "decoupled_wbc_h1": 30,
2143
+ "psix_he_g1_sonic": 30,
2144
+ "psix_g1_sonic_neck": 30,
2145
+ "psix_he_h1_sonic": 30,
2146
+ "egodex_psix": 30
2147
+ }
2148
+ },
2149
+ "model": {
2150
+ "resnet_store_path": null,
2151
+ "pretrained_action_header_path": null,
2152
+ "pretrained_action_header_load_mode": "legacy_auto",
2153
+ "rtc": false,
2154
+ "max_delay": 8,
2155
+ "rtc_dropout_prob": 0.0,
2156
+ "action_dim": 80,
2157
+ "action_chunk_size": 30,
2158
+ "action_exec_horizon": 30,
2159
+ "observation_horizon": 1,
2160
+ "img_chunk": 1,
2161
+ "n_cams": 1,
2162
+ "use_obs": "add_token",
2163
+ "dropout": 0.1,
2164
+ "noise_scheduler": "flow",
2165
+ "train_diffusion_steps": 1000,
2166
+ "eval_diffusion_steps": 10,
2167
+ "share_cam_features": false,
2168
+ "early_fusion": false,
2169
+ "odim": 45,
2170
+ "n_conditions": 0,
2171
+ "token_fusion": "concat",
2172
+ "loss_w": [
2173
+ 0.1,
2174
+ 0.2,
2175
+ 0.1
2176
+ ],
2177
+ "time_dim": 256,
2178
+ "hidden_dim": 1536,
2179
+ "num_blocks": 6,
2180
+ "dim_feedforward": 2048,
2181
+ "nhead": 24,
2182
+ "activation": "gelu",
2183
+ "view_feature_dim": 2048,
2184
+ "use_film": false,
2185
+ "combined_temb": false,
2186
+ "pooled_text_encoder": null,
2187
+ "pooled_text_encoder_path": "openai/clip-vit-large-patch14",
2188
+ "pooled_projection_dim": 768,
2189
+ "pooled_cache_path": null,
2190
+ "pooled_zero_init": false,
2191
+ "final_layer_norm": true,
2192
+ "qk_norm": "rms_norm",
2193
+ "alt_dit_qk_norm": null,
2194
+ "state_dropout_prob": 0.0,
2195
+ "state_feature_dropout": 0.0,
2196
+ "state_as_action_token": true,
2197
+ "use_dit": false,
2198
+ "expert_arch": "mmdit",
2199
+ "alt_dit_num_blocks": 14,
2200
+ "alt_dit_cross_roles": "vl",
2201
+ "mmdit_num_blocks": 12,
2202
+ "weight_decay": 0.01,
2203
+ "model_name_or_path": "/scratch/project/prj-02-spatial-intelligence/jwang/h_tmp/checkpoints/psix/posttrain/pre.fast.he-unifolm-v2.2608232200.ckpt75k",
2204
+ "vlm_ckpt_step": null,
2205
+ "tune_vlm": false,
2206
+ "vlm_layer_indices": [
2207
+ 3,
2208
+ 5,
2209
+ 8,
2210
+ 10,
2211
+ 12,
2212
+ 14,
2213
+ 17,
2214
+ 19,
2215
+ 21,
2216
+ 23,
2217
+ 26,
2218
+ 28
2219
+ ],
2220
+ "tune_mm_llm": false,
2221
+ "tune_mm_vision": false,
2222
+ "tune_mm_mlp": false,
2223
+ "gradient_checkpointing": false,
2224
+ "lang_backbone_lr": 1e-07,
2225
+ "mm_projector_lr": 1e-05,
2226
+ "vision_tower_lr": 1e-06,
2227
+ "optim": "adamw_torch",
2228
+ "model_max_length": 4096,
2229
+ "data_flatten": true,
2230
+ "data_packing": true,
2231
+ "max_pixels": 16777216,
2232
+ "min_pixels": 4096
2233
+ }
2234
+ }