experiment_cfg/conf.yaml

30.4 KB · 1325 lines · yaml Raw

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