experiment_cfg/config.yaml

34.2 KB · 1342 lines · yaml Raw

1	`!!python/object:groot.vla.omni.configs.base_config.Config`
2	`data: !!python/object:groot.vla.omni.configs.data.data_config.DataConfig`
3	`datasets:`
4	`- !!python/object:groot.vla.omni.configs.data.data_config.SingleDatasetConfig`
5	`dataset_paths:`
6	`- /mnt/aws-lfs-02/shared/datasets/xdof.yam_v7_all_merged_global_task_exclude_bad_subtasks`
7	`dataset_type: physical_embodiment`
8	`embodiment_tag: xdof_relative_eef_relative_joint`
9	`mix_ratio: 0.1`
10	`- !!python/object:groot.vla.omni.configs.data.data_config.SingleDatasetConfig`
11	`dataset_paths:`
12	`- /mnt/aws-lfs-02/shared/datasets/xdof.yam_v7_subtask_only_merged_global_task`
13	`dataset_type: physical_embodiment`
14	`embodiment_tag: xdof_relative_eef_relative_joint_subtask`
15	`mix_ratio: 0.2`
16	`- !!python/object:groot.vla.omni.configs.data.data_config.SingleDatasetConfig`
17	`dataset_paths:`
18	`- /mnt/aws-lfs-02/shared/datasets/droid_101_success_idlefiltered_n17`
19	`- /mnt/aws-lfs-02/shared/datasets/droid_101_success_idlefiltered_n17_swapped`
20	`dataset_type: physical_embodiment`
21	`embodiment_tag: oxe_droid_relative_eef_relative_joint`
22	`mix_ratio: 0.1`
23	`- !!python/object:groot.vla.omni.configs.data.data_config.SingleDatasetConfig`
24	`dataset_paths:`
25	`- /mnt/aws-lfs-02/shared/datasets/galaxea_sharpa/real_g1.g1-in-the-wild-merged`
26	`dataset_type: physical_embodiment`
27	`embodiment_tag: real_g1_relative_eef_relative_joints`
28	`mix_ratio: 0.05`
29	`- !!python/object:groot.vla.omni.configs.data.data_config.SingleDatasetConfig`
30	`dataset_paths:`
31	`- /mnt/aws-lfs-02/shared/datasets/galaxea_sharpa/real_r1_pro_sharpa.inlab_play_real_robot_batch_1`
32	`- /mnt/aws-lfs-02/shared/datasets/galaxea_sharpa/real_r1_pro_sharpa.inlab_play_real_robot_batch_2`
33	`- /mnt/aws-lfs-02/shared/datasets/galaxea_sharpa/real_r1_pro_sharpa.miscellaneous_1k_trajectories`
34	`dataset_type: physical_embodiment`
35	`embodiment_tag: real_r1_pro_sharpa_relative_eef`
36	`mix_ratio: 0.05`
37	`- !!python/object:groot.vla.omni.configs.data.data_config.SingleDatasetConfig`
38	`dataset_paths:`
39	`- /mnt/aws-lfs-02/shared/datasets/galaxea_sharpa/mecka_lerobot/real_r1_pro_sharpa.mecka_batch1-2025-12-10-merged`
40	`- /mnt/aws-lfs-02/shared/datasets/galaxea_sharpa/mecka_lerobot/real_r1_pro_sharpa.mecka_batch3_2026-01-04-merged_backup`
41	`- /mnt/aws-lfs-02/shared/datasets/galaxea_sharpa/mecka_lerobot/real_r1_pro_sharpa.mecka_batch4_2026-01-05-merged_backup`
42	`- /mnt/aws-lfs-02/shared/datasets/galaxea_sharpa/mecka_lerobot/real_r1_pro_sharpa.mecka_batch5_2026-01-05-merged_backup`
43	`- /mnt/aws-lfs-02/shared/datasets/galaxea_sharpa/mecka_lerobot/real_r1_pro_sharpa.mecka_batch6_2026-01-05-merged_backup`
44	`- /mnt/aws-lfs-02/shared/datasets/galaxea_sharpa/mecka_lerobot/real_r1_pro_sharpa.mecka_batch10_2026-01-10-merged_backup`
45	`- /mnt/aws-lfs-02/shared/datasets/galaxea_sharpa/mecka_lerobot/real_r1_pro_sharpa.mecka_batch11_2026-01-10-merged_backup`
46	`- /mnt/aws-lfs-02/shared/datasets/galaxea_sharpa/mecka_lerobot/real_r1_pro_sharpa.mecka_batch12_2026-01-10-merged_backup`
47	`- /mnt/aws-lfs-02/shared/datasets/galaxea_sharpa/mecka_lerobot/real_r1_pro_sharpa.mecka_batch8_2026-01-10-merged_backup`
48	`- /mnt/aws-lfs-02/shared/datasets/galaxea_sharpa/mecka_lerobot/real_r1_pro_sharpa.mecka_batch9_2026-01-10-merged_backup`
49	`dataset_type: physical_embodiment`
50	`embodiment_tag: real_r1_pro_sharpa_relative_eef_mecka`
51	`mix_ratio: 0.25`
52	`- !!python/object:groot.vla.omni.configs.data.data_config.SingleDatasetConfig`
53	`dataset_paths:`
54	`- /mnt/aws-lfs-02/shared/datasets/galaxea_sharpa/maxinsights_lerobot_updated/1530hrs/real_r1_pro_sharpa.maxinsights_1530hrs_updated_train_set_merged`
55	`dataset_type: physical_embodiment`
56	`embodiment_tag: real_r1_pro_sharpa_relative_eef_maxinsights`
57	`mix_ratio: 0.2`
58	`- !!python/object:groot.vla.omni.configs.data.data_config.SingleDatasetConfig`
59	`dataset_paths:`
60	`- /mnt/aws-lfs-02/shared/datasets/galaxea_sharpa/real_r1_pro_sharpa.inlab_play_human_batch1`
61	`- /mnt/aws-lfs-02/shared/datasets/galaxea_sharpa/real_r1_pro_sharpa.inlab_play_human_batch2`
62	`- /mnt/aws-lfs-02/shared/datasets/galaxea_sharpa/real_r1_pro_sharpa.shirt_rolling_task24_2000_human_video_filter_n6_keep1619_demo_stats`
63	`- /mnt/aws-lfs-02/shared/datasets/galaxea_sharpa/real_r1_pro_sharpa.shirt_rolling_task15_2000_human_video_filter_n6_keep572_demo_stats`
64	`- /mnt/aws-lfs-02/shared/datasets/galaxea_sharpa/real_r1_pro_sharpa.sort_cards_human_filter_n6_keep523_demo_stats_overwrite_left_side_stats`
65	`- /mnt/aws-lfs-02/shared/datasets/galaxea_sharpa/real_r1_pro_sharpa.tong_task38_2000_human_video_overwrite_left_side_stats`
66	`- /mnt/aws-lfs-02/shared/datasets/galaxea_sharpa/real_r1_pro_sharpa.syringe_task30i_2000_human_video_filtered`
67	`- /mnt/aws-lfs-02/shared/datasets/galaxea_sharpa/real_r1_pro_sharpa.unscrew_bottle_task43_2000_human_video_fixed-duration`
68	`- /mnt/aws-lfs-02/shared/datasets/galaxea_sharpa/real_r1_pro_sharpa.unscrew_Jim_bottle_task47_600_human_video`
69	`- /mnt/aws-lfs-02/shared/datasets/galaxea_sharpa/real_r1_pro_sharpa.fold_shirt_task30b_500_human_video_halfdone`
70	`- /mnt/aws-lfs-02/shared/datasets/galaxea_sharpa/real_r1_pro_sharpa.fold_towel_task30c_500_human_video_halfdone`
71	`- /mnt/aws-lfs-02/shared/datasets/galaxea_sharpa/real_r1_pro_sharpa.sort_cards_task32e_1000_human_video`
72	`dataset_type: physical_embodiment`
73	`embodiment_tag: real_r1_pro_sharpa_relative_eef_human`
74	`mix_ratio: 0.05`
75	`download_cache: false`
76	`episode_sampling_rate: 0.1`
77	`image_crop_size:`
78	`- 244`
79	`- 244`
80	`image_target_size:`
81	`- 224`
82	`- 224`
83	`max_prompt_trajectories: 5`
84	`mock_dataset_mode: false`
85	`modality_configs:`
86	`oxe_droid_relative_eef_relative_joint:`
87	`action: !!python/object:groot.vla.omni.data.types.ModalityConfig`
88	`action_format:`
89	`- &id004 !!python/object/apply:groot.vla.omni.data.types.ActionFormat`
90	`- xyz+rot6d`
91	`- &id001 !!python/object/apply:groot.vla.omni.data.types.ActionFormat`
92	`- default`
93	`- *id001`
94	`action_representation:`
95	`- &id002 !!python/object/apply:groot.vla.omni.data.types.ActionRepresentation`
96	`- relative`
97	`- &id005 !!python/object/apply:groot.vla.omni.data.types.ActionRepresentation`
98	`- absolute`
99	`- *id002`
100	`action_type:`
101	`- &id006 !!python/object/apply:groot.vla.omni.data.types.ActionType`
102	`- eef`
103	`- &id003 !!python/object/apply:groot.vla.omni.data.types.ActionType`
104	`- non_eef`
105	`- *id003`
106	`delta_indices:`
107	`- 0`
108	`- 1`
109	`- 2`
110	`- 3`
111	`- 4`
112	`- 5`
113	`- 6`
114	`- 7`
115	`- 8`
116	`- 9`
117	`- 10`
118	`- 11`
119	`- 12`
120	`- 13`
121	`- 14`
122	`- 15`
123	`- 16`
124	`- 17`
125	`- 18`
126	`- 19`
127	`- 20`
128	`- 21`
129	`- 22`
130	`- 23`
131	`- 24`
132	`- 25`
133	`- 26`
134	`- 27`
135	`- 28`
136	`- 29`
137	`- 30`
138	`- 31`
139	`- 32`
140	`- 33`
141	`- 34`
142	`- 35`
143	`- 36`
144	`- 37`
145	`- 38`
146	`- 39`
147	`exclude_state: false`
148	`extra_keys:`
149	`- joint_position`
150	`hand_keys:`
151	`- gripper_position`
152	`loss_weights: null`
153	`modality_keys:`
154	`- eef_9d`
155	`- gripper_position`
156	`- joint_position`
157	`normalization_mode: null`
158	`normalize_rotation: true`
159	`wrist_keys:`
160	`- eef_9d`
161	`language: !!python/object:groot.vla.omni.data.types.ModalityConfig`
162	`action_format: null`
163	`action_representation: null`
164	`action_type: null`
165	`delta_indices:`
166	`- 0`
167	`exclude_state: false`
168	`extra_keys: null`
169	`hand_keys: null`
170	`loss_weights: null`
171	`modality_keys:`
172	`- annotation.language.language_instruction`
173	`- annotation.language.language_instruction_2`
174	`- annotation.language.language_instruction_3`
175	`normalization_mode: null`
176	`normalize_rotation: true`
177	`wrist_keys: null`
178	`state: !!python/object:groot.vla.omni.data.types.ModalityConfig`
179	`action_format: null`
180	`action_representation: null`
181	`action_type: null`
182	`delta_indices:`
183	`- 0`
184	`exclude_state: false`
185	`extra_keys: null`
186	`hand_keys: null`
187	`loss_weights: null`
188	`modality_keys:`
189	`- eef_9d`
190	`- gripper_position`
191	`- joint_position`
192	`normalization_mode: null`
193	`normalize_rotation: true`
194	`wrist_keys: null`
195	`video: !!python/object:groot.vla.omni.data.types.ModalityConfig`
196	`action_format: null`
197	`action_representation: null`
198	`action_type: null`
199	`delta_indices:`
200	`- -15`
201	`- 0`
202	`exclude_state: false`
203	`extra_keys: null`
204	`hand_keys: null`
205	`loss_weights: null`
206	`modality_keys:`
207	`- exterior_image_1_left`
208	`- wrist_image_left`
209	`normalization_mode: null`
210	`normalize_rotation: true`
211	`wrist_keys: null`
212	`real_g1_relative_eef_relative_joints:`
213	`action: !!python/object:groot.vla.omni.data.types.ModalityConfig`
214	`action_format:`
215	`- *id004`
216	`- *id004`
217	`- *id001`
218	`- *id001`
219	`- *id001`
220	`- *id001`
221	`- *id001`
222	`- *id001`
223	`- *id001`
224	`action_representation:`
225	`- *id002`
226	`- *id002`
227	`- *id005`
228	`- *id005`
229	`- *id002`
230	`- *id002`
231	`- *id005`
232	`- *id005`
233	`- *id005`
234	`action_type:`
235	`- *id006`
236	`- *id006`
237	`- *id003`
238	`- *id003`
239	`- *id003`
240	`- *id003`
241	`- *id003`
242	`- *id003`
243	`- *id003`
244	`delta_indices:`
245	`- 0`
246	`- 1`
247	`- 2`
248	`- 3`
249	`- 4`
250	`- 5`
251	`- 6`
252	`- 7`
253	`- 8`
254	`- 9`
255	`- 10`
256	`- 11`
257	`- 12`
258	`- 13`
259	`- 14`
260	`- 15`
261	`- 16`
262	`- 17`
263	`- 18`
264	`- 19`
265	`- 20`
266	`- 21`
267	`- 22`
268	`- 23`
269	`- 24`
270	`- 25`
271	`- 26`
272	`- 27`
273	`- 28`
274	`- 29`
275	`- 30`
276	`- 31`
277	`- 32`
278	`- 33`
279	`- 34`
280	`- 35`
281	`- 36`
282	`- 37`
283	`- 38`
284	`- 39`
285	`exclude_state: false`
286	`extra_keys:`
287	`- left_arm`
288	`- right_arm`
289	`- waist`
290	`- base_height_command`
291	`- navigate_command`
292	`hand_keys:`
293	`- left_hand`
294	`- right_hand`
295	`loss_weights: null`
296	`modality_keys:`
297	`- left_wrist_eef_9d`
298	`- right_wrist_eef_9d`
299	`- left_hand`
300	`- right_hand`
301	`- left_arm`
302	`- right_arm`
303	`- waist`
304	`- base_height_command`
305	`- navigate_command`
306	`normalization_mode: null`
307	`normalize_rotation: true`
308	`wrist_keys:`
309	`- left_wrist_eef_9d`
310	`- right_wrist_eef_9d`
311	`language: !!python/object:groot.vla.omni.data.types.ModalityConfig`
312	`action_format: null`
313	`action_representation: null`
314	`action_type: null`
315	`delta_indices:`
316	`- 0`
317	`exclude_state: false`
318	`extra_keys: null`
319	`hand_keys: null`
320	`loss_weights: null`
321	`modality_keys:`
322	`- annotation.human.task_description`
323	`normalization_mode: null`
324	`normalize_rotation: true`
325	`wrist_keys: null`
326	`state: !!python/object:groot.vla.omni.data.types.ModalityConfig`
327	`action_format: null`
328	`action_representation: null`
329	`action_type: null`
330	`delta_indices:`
331	`- 0`
332	`exclude_state: false`
333	`extra_keys: null`
334	`hand_keys: null`
335	`loss_weights: null`
336	`modality_keys:`
337	`- left_wrist_eef_9d`
338	`- right_wrist_eef_9d`
339	`- left_hand`
340	`- right_hand`
341	`- left_arm`
342	`- right_arm`
343	`- waist`
344	`normalization_mode: null`
345	`normalize_rotation: true`
346	`wrist_keys: null`
347	`video: !!python/object:groot.vla.omni.data.types.ModalityConfig`
348	`action_format: null`
349	`action_representation: null`
350	`action_type: null`
351	`delta_indices:`
352	`- -20`
353	`- 0`
354	`exclude_state: false`
355	`extra_keys: null`
356	`hand_keys: null`
357	`loss_weights: null`
358	`modality_keys:`
359	`- ego_view`
360	`normalization_mode: null`
361	`normalize_rotation: true`
362	`wrist_keys: null`
363	`real_r1_pro_sharpa_relative_eef:`
364	`action: !!python/object:groot.vla.omni.data.types.ModalityConfig`
365	`action_format:`
366	`- *id004`
367	`- *id004`
368	`- *id001`
369	`- *id001`
370	`action_representation:`
371	`- *id002`
372	`- *id002`
373	`- *id005`
374	`- *id005`
375	`action_type:`
376	`- *id006`
377	`- *id006`
378	`- *id003`
379	`- *id003`
380	`delta_indices:`
381	`- 0`
382	`- 1`
383	`- 2`
384	`- 3`
385	`- 4`
386	`- 5`
387	`- 6`
388	`- 7`
389	`- 8`
390	`- 9`
391	`- 10`
392	`- 11`
393	`- 12`
394	`- 13`
395	`- 14`
396	`- 15`
397	`- 16`
398	`- 17`
399	`- 18`
400	`- 19`
401	`- 20`
402	`- 21`
403	`- 22`
404	`- 23`
405	`- 24`
406	`- 25`
407	`- 26`
408	`- 27`
409	`- 28`
410	`- 29`
411	`- 30`
412	`- 31`
413	`- 32`
414	`- 33`
415	`- 34`
416	`- 35`
417	`- 36`
418	`- 37`
419	`- 38`
420	`- 39`
421	`exclude_state: false`
422	`extra_keys: []`
423	`hand_keys:`
424	`- left_hand_joints`
425	`- right_hand_joints`
426	`loss_weights: null`
427	`modality_keys:`
428	`- left_wrist_eef`
429	`- right_wrist_eef`
430	`- left_hand_joints`
431	`- right_hand_joints`
432	`normalization_mode: null`
433	`normalize_rotation: true`
434	`wrist_keys:`
435	`- left_wrist_eef`
436	`- right_wrist_eef`
437	`language: !!python/object:groot.vla.omni.data.types.ModalityConfig`
438	`action_format: null`
439	`action_representation: null`
440	`action_type: null`
441	`delta_indices:`
442	`- 0`
443	`exclude_state: false`
444	`extra_keys: null`
445	`hand_keys: null`
446	`loss_weights: null`
447	`modality_keys:`
448	`- annotation.human.coarse_action`
449	`normalization_mode: null`
450	`normalize_rotation: true`
451	`wrist_keys: null`
452	`state: !!python/object:groot.vla.omni.data.types.ModalityConfig`
453	`action_format: null`
454	`action_representation: null`
455	`action_type: null`
456	`delta_indices:`
457	`- 0`
458	`exclude_state: false`
459	`extra_keys: null`
460	`hand_keys: null`
461	`loss_weights: null`
462	`modality_keys:`
463	`- left_wrist_eef`
464	`- right_wrist_eef`
465	`- left_hand_joints`
466	`- right_hand_joints`
467	`normalization_mode: null`
468	`normalize_rotation: true`
469	`wrist_keys: null`
470	`video: !!python/object:groot.vla.omni.data.types.ModalityConfig`
471	`action_format: null`
472	`action_representation: null`
473	`action_type: null`
474	`delta_indices:`
475	`- -20`
476	`- 0`
477	`exclude_state: false`
478	`extra_keys: null`
479	`hand_keys: null`
480	`loss_weights: null`
481	`modality_keys:`
482	`- ego_view_res320x240_freq20`
483	`- left_wrist_view_res320x240_freq20`
484	`- right_wrist_view_res320x240_freq20`
485	`normalization_mode: null`
486	`normalize_rotation: true`
487	`wrist_keys: null`
488	`real_r1_pro_sharpa_relative_eef_human:`
489	`action: !!python/object:groot.vla.omni.data.types.ModalityConfig`
490	`action_format:`
491	`- *id004`
492	`- *id004`
493	`- *id001`
494	`- *id001`
495	`action_representation:`
496	`- *id002`
497	`- *id002`
498	`- *id005`
499	`- *id005`
500	`action_type:`
501	`- *id006`
502	`- *id006`
503	`- *id003`
504	`- *id003`
505	`delta_indices:`
506	`- 0`
507	`- 1`
508	`- 2`
509	`- 3`
510	`- 4`
511	`- 5`
512	`- 6`
513	`- 7`
514	`- 8`
515	`- 9`
516	`- 10`
517	`- 11`
518	`- 12`
519	`- 13`
520	`- 14`
521	`- 15`
522	`- 16`
523	`- 17`
524	`- 18`
525	`- 19`
526	`- 20`
527	`- 21`
528	`- 22`
529	`- 23`
530	`- 24`
531	`- 25`
532	`- 26`
533	`- 27`
534	`- 28`
535	`- 29`
536	`- 30`
537	`- 31`
538	`- 32`
539	`- 33`
540	`- 34`
541	`- 35`
542	`- 36`
543	`- 37`
544	`- 38`
545	`- 39`
546	`exclude_state: false`
547	`extra_keys: []`
548	`hand_keys:`
549	`- left_hand_joints`
550	`- right_hand_joints`
551	`loss_weights: null`
552	`modality_keys:`
553	`- left_wrist_eef`
554	`- right_wrist_eef`
555	`- left_hand_joints`
556	`- right_hand_joints`
557	`normalization_mode: null`
558	`normalize_rotation: true`
559	`wrist_keys:`
560	`- left_wrist_eef`
561	`- right_wrist_eef`
562	`language: !!python/object:groot.vla.omni.data.types.ModalityConfig`
563	`action_format: null`
564	`action_representation: null`
565	`action_type: null`
566	`delta_indices:`
567	`- 0`
568	`exclude_state: false`
569	`extra_keys: null`
570	`hand_keys: null`
571	`loss_weights: null`
572	`modality_keys:`
573	`- annotation.human.coarse_action`
574	`normalization_mode: null`
575	`normalize_rotation: true`
576	`wrist_keys: null`
577	`state: !!python/object:groot.vla.omni.data.types.ModalityConfig`
578	`action_format: null`
579	`action_representation: null`
580	`action_type: null`
581	`delta_indices:`
582	`- 0`
583	`exclude_state: true`
584	`extra_keys: null`
585	`hand_keys: null`
586	`loss_weights: null`
587	`modality_keys:`
588	`- left_wrist_eef`
589	`- right_wrist_eef`
590	`- left_hand_joints`
591	`- right_hand_joints`
592	`normalization_mode: null`
593	`normalize_rotation: true`
594	`wrist_keys: null`
595	`video: !!python/object:groot.vla.omni.data.types.ModalityConfig`
596	`action_format: null`
597	`action_representation: null`
598	`action_type: null`
599	`delta_indices:`
600	`- -20`
601	`- 0`
602	`exclude_state: false`
603	`extra_keys: null`
604	`hand_keys: null`
605	`loss_weights: null`
606	`modality_keys:`
607	`- ego_view_res320x240_freq20`
608	`- left_wrist_view_res320x240_freq20`
609	`- right_wrist_view_res320x240_freq20`
610	`normalization_mode: null`
611	`normalize_rotation: true`
612	`wrist_keys: null`
613	`real_r1_pro_sharpa_relative_eef_maxinsights:`
614	`action: !!python/object:groot.vla.omni.data.types.ModalityConfig`
615	`action_format:`
616	`- *id004`
617	`- *id004`
618	`- *id001`
619	`- *id001`
620	`action_representation:`
621	`- *id002`
622	`- *id002`
623	`- *id005`
624	`- *id005`
625	`action_type:`
626	`- *id006`
627	`- *id006`
628	`- *id003`
629	`- *id003`
630	`delta_indices:`
631	`- 0`
632	`- 1`
633	`- 2`
634	`- 3`
635	`- 4`
636	`- 5`
637	`- 6`
638	`- 7`
639	`- 8`
640	`- 9`
641	`- 10`
642	`- 11`
643	`- 12`
644	`- 13`
645	`- 14`
646	`- 15`
647	`- 16`
648	`- 17`
649	`- 18`
650	`- 19`
651	`- 20`
652	`- 21`
653	`- 22`
654	`- 23`
655	`- 24`
656	`- 25`
657	`- 26`
658	`- 27`
659	`- 28`
660	`- 29`
661	`- 30`
662	`- 31`
663	`- 32`
664	`- 33`
665	`- 34`
666	`- 35`
667	`- 36`
668	`- 37`
669	`- 38`
670	`- 39`
671	`exclude_state: false`
672	`extra_keys: []`
673	`hand_keys:`
674	`- left_hand_joints`
675	`- right_hand_joints`
676	`loss_weights: null`
677	`modality_keys:`
678	`- left_wrist_eef`
679	`- right_wrist_eef`
680	`- left_hand_joints`
681	`- right_hand_joints`
682	`normalization_mode: null`
683	`normalize_rotation: true`
684	`wrist_keys:`
685	`- left_wrist_eef`
686	`- right_wrist_eef`
687	`language: !!python/object:groot.vla.omni.data.types.ModalityConfig`
688	`action_format: null`
689	`action_representation: null`
690	`action_type: null`
691	`delta_indices:`
692	`- 0`
693	`exclude_state: false`
694	`extra_keys: null`
695	`hand_keys: null`
696	`loss_weights: null`
697	`modality_keys:`
698	`- annotation.human.coarse_action`
699	`normalization_mode: null`
700	`normalize_rotation: true`
701	`wrist_keys: null`
702	`state: !!python/object:groot.vla.omni.data.types.ModalityConfig`
703	`action_format: null`
704	`action_representation: null`
705	`action_type: null`
706	`delta_indices:`
707	`- 0`
708	`exclude_state: true`
709	`extra_keys: null`
710	`hand_keys: null`
711	`loss_weights: null`
712	`modality_keys:`
713	`- left_wrist_eef`
714	`- right_wrist_eef`
715	`- left_hand_joints`
716	`- right_hand_joints`
717	`normalization_mode: null`
718	`normalize_rotation: true`
719	`wrist_keys: null`
720	`video: !!python/object:groot.vla.omni.data.types.ModalityConfig`
721	`action_format: null`
722	`action_representation: null`
723	`action_type: null`
724	`delta_indices:`
725	`- -30`
726	`- 0`
727	`exclude_state: false`
728	`extra_keys: null`
729	`hand_keys: null`
730	`loss_weights: null`
731	`modality_keys:`
732	`- ego_view_cropratio_res320x240_freq30`
733	`normalization_mode: null`
734	`normalize_rotation: true`
735	`wrist_keys: null`
736	`real_r1_pro_sharpa_relative_eef_mecka:`
737	`action: !!python/object:groot.vla.omni.data.types.ModalityConfig`
738	`action_format:`
739	`- *id004`
740	`- *id004`
741	`- *id001`
742	`- *id001`
743	`action_representation:`
744	`- *id002`
745	`- *id002`
746	`- *id005`
747	`- *id005`
748	`action_type:`
749	`- *id006`
750	`- *id006`
751	`- *id003`
752	`- *id003`
753	`delta_indices:`
754	`- 0`
755	`- 1`
756	`- 2`
757	`- 3`
758	`- 4`
759	`- 5`
760	`- 6`
761	`- 7`
762	`- 8`
763	`- 9`
764	`- 10`
765	`- 11`
766	`- 12`
767	`- 13`
768	`- 14`
769	`- 15`
770	`- 16`
771	`- 17`
772	`- 18`
773	`- 19`
774	`- 20`
775	`- 21`
776	`- 22`
777	`- 23`
778	`- 24`
779	`- 25`
780	`- 26`
781	`- 27`
782	`- 28`
783	`- 29`
784	`- 30`
785	`- 31`
786	`- 32`
787	`- 33`
788	`- 34`
789	`- 35`
790	`- 36`
791	`- 37`
792	`- 38`
793	`- 39`
794	`exclude_state: false`
795	`extra_keys: []`
796	`hand_keys:`
797	`- left_hand_joints`
798	`- right_hand_joints`
799	`loss_weights: null`
800	`modality_keys:`
801	`- left_wrist_eef`
802	`- right_wrist_eef`
803	`- left_hand_joints`
804	`- right_hand_joints`
805	`normalization_mode: null`
806	`normalize_rotation: true`
807	`wrist_keys:`
808	`- left_wrist_eef`
809	`- right_wrist_eef`
810	`language: !!python/object:groot.vla.omni.data.types.ModalityConfig`
811	`action_format: null`
812	`action_representation: null`
813	`action_type: null`
814	`delta_indices:`
815	`- 0`
816	`exclude_state: false`
817	`extra_keys: null`
818	`hand_keys: null`
819	`loss_weights: null`
820	`modality_keys:`
821	`- annotation.human.coarse_action`
822	`normalization_mode: null`
823	`normalize_rotation: true`
824	`wrist_keys: null`
825	`state: !!python/object:groot.vla.omni.data.types.ModalityConfig`
826	`action_format: null`
827	`action_representation: null`
828	`action_type: null`
829	`delta_indices:`
830	`- 0`
831	`exclude_state: true`
832	`extra_keys: null`
833	`hand_keys: null`
834	`loss_weights: null`
835	`modality_keys:`
836	`- left_wrist_eef`
837	`- right_wrist_eef`
838	`- left_hand_joints`
839	`- right_hand_joints`
840	`normalization_mode: null`
841	`normalize_rotation: true`
842	`wrist_keys: null`
843	`video: !!python/object:groot.vla.omni.data.types.ModalityConfig`
844	`action_format: null`
845	`action_representation: null`
846	`action_type: null`
847	`delta_indices:`
848	`- -30`
849	`- 0`
850	`exclude_state: false`
851	`extra_keys: null`
852	`hand_keys: null`
853	`loss_weights: null`
854	`modality_keys:`
855	`- ego_view_cropratio_res320x240_freq30`
856	`normalization_mode: null`
857	`normalize_rotation: true`
858	`wrist_keys: null`
859	`xdof_relative_eef_relative_joint:`
860	`action: !!python/object:groot.vla.omni.data.types.ModalityConfig`
861	`action_format:`
862	`- *id004`
863	`- *id004`
864	`- *id001`
865	`- *id001`
866	`- *id001`
867	`- *id001`
868	`action_representation:`
869	`- *id002`
870	`- *id002`
871	`- *id005`
872	`- *id005`
873	`- *id002`
874	`- *id002`
875	`action_type:`
876	`- *id006`
877	`- *id006`
878	`- *id003`
879	`- *id003`
880	`- *id003`
881	`- *id003`
882	`delta_indices:`
883	`- 0`
884	`- 1`
885	`- 2`
886	`- 3`
887	`- 4`
888	`- 5`
889	`- 6`
890	`- 7`
891	`- 8`
892	`- 9`
893	`- 10`
894	`- 11`
895	`- 12`
896	`- 13`
897	`- 14`
898	`- 15`
899	`- 16`
900	`- 17`
901	`- 18`
902	`- 19`
903	`- 20`
904	`- 21`
905	`- 22`
906	`- 23`
907	`- 24`
908	`- 25`
909	`- 26`
910	`- 27`
911	`- 28`
912	`- 29`
913	`- 30`
914	`- 31`
915	`- 32`
916	`- 33`
917	`- 34`
918	`- 35`
919	`- 36`
920	`- 37`
921	`- 38`
922	`- 39`
923	`exclude_state: false`
924	`extra_keys:`
925	`- left_joint_pos`
926	`- right_joint_pos`
927	`hand_keys:`
928	`- left_gripper_pos`
929	`- right_gripper_pos`
930	`loss_weights: null`
931	`modality_keys:`
932	`- left_wrist_eef`
933	`- right_wrist_eef`
934	`- left_gripper_pos`
935	`- right_gripper_pos`
936	`- left_joint_pos`
937	`- right_joint_pos`
938	`normalization_mode: null`
939	`normalize_rotation: true`
940	`wrist_keys:`
941	`- left_wrist_eef`
942	`- right_wrist_eef`
943	`language: !!python/object:groot.vla.omni.data.types.ModalityConfig`
944	`action_format: null`
945	`action_representation: null`
946	`action_type: null`
947	`delta_indices:`
948	`- 0`
949	`exclude_state: false`
950	`extra_keys: null`
951	`hand_keys: null`
952	`loss_weights: null`
953	`modality_keys:`
954	`- annotation.task`
955	`normalization_mode: null`
956	`normalize_rotation: true`
957	`wrist_keys: null`
958	`state: !!python/object:groot.vla.omni.data.types.ModalityConfig`
959	`action_format: null`
960	`action_representation: null`
961	`action_type: null`
962	`delta_indices:`
963	`- 0`
964	`exclude_state: false`
965	`extra_keys: null`
966	`hand_keys: null`
967	`loss_weights: null`
968	`modality_keys:`
969	`- left_wrist_eef`
970	`- right_wrist_eef`
971	`- left_gripper_pos`
972	`- right_gripper_pos`
973	`- left_joint_pos`
974	`- right_joint_pos`
975	`normalization_mode: null`
976	`normalize_rotation: true`
977	`wrist_keys: null`
978	`video: !!python/object:groot.vla.omni.data.types.ModalityConfig`
979	`action_format: null`
980	`action_representation: null`
981	`action_type: null`
982	`delta_indices:`
983	`- -30`
984	`- 0`
985	`exclude_state: false`
986	`extra_keys: null`
987	`hand_keys: null`
988	`loss_weights: null`
989	`modality_keys:`
990	`- top_camera-images-rgb_320_240`
991	`- left_camera-images-rgb_320_240`
992	`- right_camera-images-rgb_320_240`
993	`normalization_mode: null`
994	`normalize_rotation: true`
995	`wrist_keys: null`
996	`xdof_relative_eef_relative_joint_subtask:`
997	`action: !!python/object:groot.vla.omni.data.types.ModalityConfig`
998	`action_format:`
999	`- *id004`
1000	`- *id004`
1001	`- *id001`
1002	`- *id001`
1003	`- *id001`
1004	`- *id001`
1005	`action_representation:`
1006	`- *id002`
1007	`- *id002`
1008	`- *id005`
1009	`- *id005`
1010	`- *id002`
1011	`- *id002`
1012	`action_type:`
1013	`- *id006`
1014	`- *id006`
1015	`- *id003`
1016	`- *id003`
1017	`- *id003`
1018	`- *id003`
1019	`delta_indices:`
1020	`- 0`
1021	`- 1`
1022	`- 2`
1023	`- 3`
1024	`- 4`
1025	`- 5`
1026	`- 6`
1027	`- 7`
1028	`- 8`
1029	`- 9`
1030	`- 10`
1031	`- 11`
1032	`- 12`
1033	`- 13`
1034	`- 14`
1035	`- 15`
1036	`- 16`
1037	`- 17`
1038	`- 18`
1039	`- 19`
1040	`- 20`
1041	`- 21`
1042	`- 22`
1043	`- 23`
1044	`- 24`
1045	`- 25`
1046	`- 26`
1047	`- 27`
1048	`- 28`
1049	`- 29`
1050	`- 30`
1051	`- 31`
1052	`- 32`
1053	`- 33`
1054	`- 34`
1055	`- 35`
1056	`- 36`
1057	`- 37`
1058	`- 38`
1059	`- 39`
1060	`exclude_state: false`
1061	`extra_keys:`
1062	`- left_joint_pos`
1063	`- right_joint_pos`
1064	`hand_keys:`
1065	`- left_gripper_pos`
1066	`- right_gripper_pos`
1067	`loss_weights: null`
1068	`modality_keys:`
1069	`- left_wrist_eef`
1070	`- right_wrist_eef`
1071	`- left_gripper_pos`
1072	`- right_gripper_pos`
1073	`- left_joint_pos`
1074	`- right_joint_pos`
1075	`normalization_mode: null`
1076	`normalize_rotation: true`
1077	`wrist_keys:`
1078	`- left_wrist_eef`
1079	`- right_wrist_eef`
1080	`language: !!python/object:groot.vla.omni.data.types.ModalityConfig`
1081	`action_format: null`
1082	`action_representation: null`
1083	`action_type: null`
1084	`delta_indices:`
1085	`- 0`
1086	`exclude_state: false`
1087	`extra_keys: null`
1088	`hand_keys: null`
1089	`loss_weights: null`
1090	`modality_keys:`
1091	`- annotation.sub_task`
1092	`normalization_mode: null`
1093	`normalize_rotation: true`
1094	`wrist_keys: null`
1095	`state: !!python/object:groot.vla.omni.data.types.ModalityConfig`
1096	`action_format: null`
1097	`action_representation: null`
1098	`action_type: null`
1099	`delta_indices:`
1100	`- 0`
1101	`exclude_state: false`
1102	`extra_keys: null`
1103	`hand_keys: null`
1104	`loss_weights: null`
1105	`modality_keys:`
1106	`- left_wrist_eef`
1107	`- right_wrist_eef`
1108	`- left_gripper_pos`
1109	`- right_gripper_pos`
1110	`- left_joint_pos`
1111	`- right_joint_pos`
1112	`normalization_mode: null`
1113	`normalize_rotation: true`
1114	`wrist_keys: null`
1115	`video: !!python/object:groot.vla.omni.data.types.ModalityConfig`
1116	`action_format: null`
1117	`action_representation: null`
1118	`action_type: null`
1119	`delta_indices:`
1120	`- -30`
1121	`- 0`
1122	`exclude_state: false`
1123	`extra_keys: null`
1124	`hand_keys: null`
1125	`loss_weights: null`
1126	`modality_keys:`
1127	`- top_camera-images-rgb_320_240`
1128	`- left_camera-images-rgb_320_240`
1129	`- right_camera-images-rgb_320_240`
1130	`normalization_mode: null`
1131	`normalize_rotation: true`
1132	`wrist_keys: null`
1133	`mode: single_turn`
1134	`num_prompt_trajectories: 2`
1135	`num_shards_per_epoch: 100000`
1136	`override_pretraining_statistics: false`
1137	`random_chop: 0.0`
1138	`seed: 24`
1139	`shard_size: 1024`
1140	`shuffle: true`
1141	`subsample_ratio: 1.0`
1142	`variable_num_demos: false`
1143	`video_backend: torchcodec`
1144	`load_config_path: groot/vla/omni/configs/experiments/r1_pro/sharpa/n17_pretrain/n17_pretrain_human_robot_cross_embodiment_fix_yam_absolute_hand_2step.yaml`
1145	`model: !!python/object:groot.vla.omni.configs.model.groot_n1d5_qwen.GrootN1d5QwenConfig`
1146	`_attn_implementation_internal: null`
1147	`_commit_hash: null`
1148	`_name_or_path: ''`
1149	`_output_attentions: false`
1150	`action_horizon: 40`
1151	`action_space_prompt: false`
1152	`add_cross_attention: false`
1153	`add_pos_embed: true`
1154	`apply_sincos_state_encoding: false`
1155	`architectures: null`
1156	`attn_dropout: 0.2`
1157	`backbone_embedding_dim: 2048`
1158	`bad_words_ids: null`
1159	`begin_suppress_tokens: null`
1160	`bos_token_id: null`
1161	`chunk_size_feed_forward: 0`
1162	`color_jitter_params:`
1163	`brightness: 0.3`
1164	`contrast: 0.4`
1165	`hue: 0.08`
1166	`saturation: 0.5`
1167	`crop_fraction: 0.95`
1168	`cross_attention_hidden_size: null`
1169	`decoder_start_token_id: null`
1170	`diffusion_model_cfg:`
1171	`attention_head_dim: 48`
1172	`cross_attention_dim: 2048`
1173	`dropout: 0.2`
1174	`final_dropout: true`
1175	`interleave_self_attention: true`
1176	`norm_type: ada_norm`
1177	`num_attention_heads: 32`
1178	`num_layers: 32`
1179	`output_dim: 1024`
1180	`positional_embeddings: null`
1181	`dit_latent_dim: 1536`
1182	`diversity_penalty: 0.0`
1183	`do_human_interpolation: false`
1184	`do_sample: false`
1185	`dtype: null`
1186	`early_stopping: false`
1187	`encoder_no_repeat_ngram_size: 0`
1188	`eos_token_id: null`
1189	`exclude_state: false`
1190	`exponential_decay_length_penalty: null`
1191	`finetuning_task: null`
1192	`forced_bos_token_id: null`
1193	`forced_eos_token_id: null`
1194	`formalize_language: true`
1195	`hidden_size: 1024`
1196	`human_embodiment_tags: null`
1197	`id2label:`
1198	`0: LABEL_0`
1199	`1: LABEL_1`
1200	`image_crop_size: !!python/tuple`
1201	`- 230`
1202	`- 230`
1203	`image_target_size: !!python/tuple`
1204	`- 256`
1205	`- 256`
1206	`interpolation_steps: 20`
1207	`is_decoder: false`
1208	`is_encoder_decoder: false`
1209	`label2id:`
1210	`LABEL_0: 0`
1211	`LABEL_1: 1`
1212	`language_dropout_prob: 0.0`
1213	`length_penalty: 1.0`
1214	`letter_box_transform: false`
1215	`load_bf16: true`
1216	`max_action_dim: 132`
1217	`max_length: 20`
1218	`max_num_embodiments: 32`
1219	`max_seq_len: 1024`
1220	`max_state_dim: 132`
1221	`min_length: 0`
1222	`model_dtype: bfloat16`
1223	`model_type: GrootN1d5Qwen`
1224	`no_repeat_ngram_size: 0`
1225	`noise_beta_alpha: 1.5`
1226	`noise_beta_beta: 1.0`
1227	`noise_s: 0.999`
1228	`num_beam_groups: 1`
1229	`num_beams: 1`
1230	`num_inference_timesteps: 4`
1231	`num_return_sequences: 1`
1232	`num_timestep_buckets: 1000`
1233	`output_hidden_states: false`
1234	`output_scores: false`
1235	`pad_token_id: null`
1236	`prefix: null`
1237	`problem_type: null`
1238	`pruned_heads: {}`
1239	`random_history_crop: true`
1240	`random_rotation_angle: 0`
1241	`remove_invalid_values: false`
1242	`repetition_penalty: 1.0`
1243	`reproject_vision: false`
1244	`return_dict: true`
1245	`return_dict_in_generate: false`
1246	`rtc_ramp_rate: 6.0`
1247	`select_layer: 16`
1248	`sep_token_id: null`
1249	`shortest_image_edge: 256`
1250	`state_dropout_prob: 0.2`
1251	`state_gaussian_noise_std: 0.0`
1252	`suppress_tokens: null`
1253	`task_specific_params: null`
1254	`temperature: 1.0`
1255	`tf_legacy_loss: false`
1256	`tie_encoder_decoder: false`
1257	`tie_word_embeddings: true`
1258	`tokenizer_class: null`
1259	`top_k: 50`
1260	`top_p: 1.0`
1261	`torchscript: false`
1262	`transformers_version: null`
1263	`tune_diffusion_model: true`
1264	`tune_linear: true`
1265	`tune_llm: false`
1266	`tune_projector: true`
1267	`tune_top_llm_layers: 0`
1268	`tune_visual: false`
1269	`tune_vlln: true`
1270	`typical_p: 1.0`
1271	`use_albumentations: true`
1272	`use_alternate_vl_dit: true`
1273	`use_bfloat16: false`
1274	`use_flash_attention: true`
1275	`use_future_tokens: false`
1276	`use_mean_std: false`
1277	`use_percentiles: true`
1278	`use_vl_self_attention: true`
1279	`use_vlln: true`
1280	`vl_self_attention_cfg:`
1281	`attention_head_dim: 64`
1282	`dropout: 0.2`
1283	`final_dropout: true`
1284	`num_attention_heads: 32`
1285	`num_layers: 4`
1286	`positional_embeddings: null`
1287	`vlm_backend: qwen3`
1288	`vlm_model_path: nvidia/Cosmos-Reason2-2B`
1289	`training: !!python/object:groot.vla.omni.configs.training.training_config.TrainingConfig`
1290	`assert_loss_less_than: null`
1291	`batch_size: 32`
1292	`bf16: true`
1293	`dataloader_num_workers: 4`
1294	`deepspeed_stage: 2`
1295	`enable_profiling: false`
1296	`eval_batch_size: 2`
1297	`eval_bf16: true`
1298	`eval_set_split_ratio: 0.1`
1299	`eval_steps: 500`
1300	`eval_strategy: 'no'`
1301	`experiment_name: null`
1302	`fp16: false`
1303	`global_batch_size: 1024`
1304	`gradient_accumulation_steps: 1`
1305	`gradient_checkpointing: false`
1306	`learning_rate: 5.0e-05`
1307	`logging_steps: 10`
1308	`lr_scheduler_type: cosine`
1309	`max_concurrent_uploads: 2`
1310	`max_grad_norm: 1.0`
1311	`max_retries: 3`
1312	`max_steps: 200000`
1313	`muon_lr: 0.005`
1314	`num_gpus: 256`
1315	`optim: adamw_torch_fused`
1316	`output_dir: nvidia/Cosmos-Reason2-2B`
1317	`remove_unused_columns: false`
1318	`save_best_eval_metric_greater_is_better: true`
1319	`save_best_eval_metric_name: ''`
1320	`save_steps: 1000`
1321	`save_total_limit: 5`
1322	`save_vl_model: false`
1323	`skip_spike: true`
1324	`skip_spike_ema_alpha: 0.99`
1325	`skip_spike_max_consecutive: 10`
1326	`skip_spike_threshold: 5.0`
1327	`start_from_checkpoint: null`
1328	`tf32: true`
1329	`upload_checkpoints: true`
1330	`upload_every: 1000`
1331	`upload_last_n_checkpoints: 5`
1332	`use_ddp: false`
1333	`use_legacy_wd_application: false`
1334	`use_muon: false`
1335	`use_wandb: true`
1336	`wandb_project: human_pretraining_n15_galaxea_sharpa`
1337	`warmup_ratio: 0.05`
1338	`warmup_steps: 0`
1339	`weight_decay: 1.0e-05`
1340	`wsd_decay_type: cosine`
1341	`wsd_stable_ratio: 0.8`
1342