processor_config.json
20.6 KB · 956 lines · json Raw
1 {
2 "processor_class": "Gr00tN1d7Processor",
3 "processor_kwargs": {
4 "modality_configs": {
5 "real_g1_relative_eef_relative_joints": {
6 "video": {
7 "delta_indices": [
8 -20,
9 0
10 ],
11 "modality_keys": [
12 "ego_view"
13 ]
14 },
15 "state": {
16 "delta_indices": [
17 0
18 ],
19 "modality_keys": [
20 "left_wrist_eef_9d",
21 "right_wrist_eef_9d",
22 "left_hand",
23 "right_hand",
24 "left_arm",
25 "right_arm",
26 "waist"
27 ]
28 },
29 "action": {
30 "delta_indices": [
31 0,
32 1,
33 2,
34 3,
35 4,
36 5,
37 6,
38 7,
39 8,
40 9,
41 10,
42 11,
43 12,
44 13,
45 14,
46 15,
47 16,
48 17,
49 18,
50 19,
51 20,
52 21,
53 22,
54 23,
55 24,
56 25,
57 26,
58 27,
59 28,
60 29,
61 30,
62 31,
63 32,
64 33,
65 34,
66 35,
67 36,
68 37,
69 38,
70 39
71 ],
72 "modality_keys": [
73 "left_wrist_eef_9d",
74 "right_wrist_eef_9d",
75 "left_hand",
76 "right_hand",
77 "left_arm",
78 "right_arm",
79 "waist",
80 "base_height_command",
81 "navigate_command"
82 ],
83 "action_configs": [
84 {
85 "rep": "RELATIVE",
86 "type": "EEF",
87 "format": "XYZ_ROT6D",
88 "state_key": "left_wrist_eef_9d"
89 },
90 {
91 "rep": "RELATIVE",
92 "type": "EEF",
93 "format": "XYZ_ROT6D",
94 "state_key": "right_wrist_eef_9d"
95 },
96 {
97 "rep": "ABSOLUTE",
98 "type": "NON_EEF",
99 "format": "DEFAULT",
100 "state_key": "left_hand"
101 },
102 {
103 "rep": "ABSOLUTE",
104 "type": "NON_EEF",
105 "format": "DEFAULT",
106 "state_key": "right_hand"
107 },
108 {
109 "rep": "RELATIVE",
110 "type": "NON_EEF",
111 "format": "DEFAULT",
112 "state_key": "left_arm"
113 },
114 {
115 "rep": "RELATIVE",
116 "type": "NON_EEF",
117 "format": "DEFAULT",
118 "state_key": "right_arm"
119 },
120 {
121 "rep": "ABSOLUTE",
122 "type": "NON_EEF",
123 "format": "DEFAULT",
124 "state_key": "waist"
125 },
126 {
127 "rep": "ABSOLUTE",
128 "type": "NON_EEF",
129 "format": "DEFAULT",
130 "state_key": "base_height_command"
131 },
132 {
133 "rep": "ABSOLUTE",
134 "type": "NON_EEF",
135 "format": "DEFAULT",
136 "state_key": "navigate_command"
137 }
138 ]
139 },
140 "language": {
141 "delta_indices": [
142 0
143 ],
144 "modality_keys": [
145 "annotation.human.task_description"
146 ]
147 }
148 },
149 "real_r1_pro_sharpa_relative_eef_mecka": {
150 "video": {
151 "delta_indices": [
152 -30,
153 0
154 ],
155 "modality_keys": [
156 "ego_view_cropratio_res320x240_freq30"
157 ]
158 },
159 "state": {
160 "delta_indices": [
161 0
162 ],
163 "modality_keys": [
164 "left_wrist_eef",
165 "right_wrist_eef",
166 "left_hand_joints",
167 "right_hand_joints"
168 ]
169 },
170 "action": {
171 "delta_indices": [
172 0,
173 1,
174 2,
175 3,
176 4,
177 5,
178 6,
179 7,
180 8,
181 9,
182 10,
183 11,
184 12,
185 13,
186 14,
187 15,
188 16,
189 17,
190 18,
191 19,
192 20,
193 21,
194 22,
195 23,
196 24,
197 25,
198 26,
199 27,
200 28,
201 29,
202 30,
203 31,
204 32,
205 33,
206 34,
207 35,
208 36,
209 37,
210 38,
211 39
212 ],
213 "modality_keys": [
214 "left_wrist_eef",
215 "right_wrist_eef",
216 "left_hand_joints",
217 "right_hand_joints"
218 ],
219 "action_configs": [
220 {
221 "rep": "RELATIVE",
222 "type": "EEF",
223 "format": "XYZ_ROT6D",
224 "state_key": "left_wrist_eef"
225 },
226 {
227 "rep": "RELATIVE",
228 "type": "EEF",
229 "format": "XYZ_ROT6D",
230 "state_key": "right_wrist_eef"
231 },
232 {
233 "rep": "ABSOLUTE",
234 "type": "NON_EEF",
235 "format": "DEFAULT",
236 "state_key": "left_hand_joints"
237 },
238 {
239 "rep": "ABSOLUTE",
240 "type": "NON_EEF",
241 "format": "DEFAULT",
242 "state_key": "right_hand_joints"
243 }
244 ]
245 },
246 "language": {
247 "delta_indices": [
248 0
249 ],
250 "modality_keys": [
251 "annotation.human.coarse_action"
252 ]
253 }
254 },
255 "real_r1_pro_sharpa_relative_eef_human": {
256 "video": {
257 "delta_indices": [
258 -20,
259 0
260 ],
261 "modality_keys": [
262 "ego_view_res320x240_freq20",
263 "left_wrist_view_res320x240_freq20",
264 "right_wrist_view_res320x240_freq20"
265 ]
266 },
267 "state": {
268 "delta_indices": [
269 0
270 ],
271 "modality_keys": [
272 "left_wrist_eef",
273 "right_wrist_eef",
274 "left_hand_joints",
275 "right_hand_joints"
276 ]
277 },
278 "action": {
279 "delta_indices": [
280 0,
281 1,
282 2,
283 3,
284 4,
285 5,
286 6,
287 7,
288 8,
289 9,
290 10,
291 11,
292 12,
293 13,
294 14,
295 15,
296 16,
297 17,
298 18,
299 19,
300 20,
301 21,
302 22,
303 23,
304 24,
305 25,
306 26,
307 27,
308 28,
309 29,
310 30,
311 31,
312 32,
313 33,
314 34,
315 35,
316 36,
317 37,
318 38,
319 39
320 ],
321 "modality_keys": [
322 "left_wrist_eef",
323 "right_wrist_eef",
324 "left_hand_joints",
325 "right_hand_joints"
326 ],
327 "action_configs": [
328 {
329 "rep": "RELATIVE",
330 "type": "EEF",
331 "format": "XYZ_ROT6D",
332 "state_key": "left_wrist_eef"
333 },
334 {
335 "rep": "RELATIVE",
336 "type": "EEF",
337 "format": "XYZ_ROT6D",
338 "state_key": "right_wrist_eef"
339 },
340 {
341 "rep": "ABSOLUTE",
342 "type": "NON_EEF",
343 "format": "DEFAULT",
344 "state_key": "left_hand_joints"
345 },
346 {
347 "rep": "ABSOLUTE",
348 "type": "NON_EEF",
349 "format": "DEFAULT",
350 "state_key": "right_hand_joints"
351 }
352 ]
353 },
354 "language": {
355 "delta_indices": [
356 0
357 ],
358 "modality_keys": [
359 "annotation.human.coarse_action"
360 ]
361 }
362 },
363 "real_r1_pro_sharpa_relative_eef": {
364 "video": {
365 "delta_indices": [
366 -20,
367 0
368 ],
369 "modality_keys": [
370 "ego_view_res320x240_freq20",
371 "left_wrist_view_res320x240_freq20",
372 "right_wrist_view_res320x240_freq20"
373 ]
374 },
375 "state": {
376 "delta_indices": [
377 0
378 ],
379 "modality_keys": [
380 "left_wrist_eef",
381 "right_wrist_eef",
382 "left_hand_joints",
383 "right_hand_joints"
384 ]
385 },
386 "action": {
387 "delta_indices": [
388 0,
389 1,
390 2,
391 3,
392 4,
393 5,
394 6,
395 7,
396 8,
397 9,
398 10,
399 11,
400 12,
401 13,
402 14,
403 15,
404 16,
405 17,
406 18,
407 19,
408 20,
409 21,
410 22,
411 23,
412 24,
413 25,
414 26,
415 27,
416 28,
417 29,
418 30,
419 31,
420 32,
421 33,
422 34,
423 35,
424 36,
425 37,
426 38,
427 39
428 ],
429 "modality_keys": [
430 "left_wrist_eef",
431 "right_wrist_eef",
432 "left_hand_joints",
433 "right_hand_joints"
434 ],
435 "action_configs": [
436 {
437 "rep": "RELATIVE",
438 "type": "EEF",
439 "format": "XYZ_ROT6D",
440 "state_key": "left_wrist_eef"
441 },
442 {
443 "rep": "RELATIVE",
444 "type": "EEF",
445 "format": "XYZ_ROT6D",
446 "state_key": "right_wrist_eef"
447 },
448 {
449 "rep": "ABSOLUTE",
450 "type": "NON_EEF",
451 "format": "DEFAULT",
452 "state_key": "left_hand_joints"
453 },
454 {
455 "rep": "ABSOLUTE",
456 "type": "NON_EEF",
457 "format": "DEFAULT",
458 "state_key": "right_hand_joints"
459 }
460 ]
461 },
462 "language": {
463 "delta_indices": [
464 0
465 ],
466 "modality_keys": [
467 "annotation.human.coarse_action"
468 ]
469 }
470 },
471 "xdof_relative_eef_relative_joint": {
472 "video": {
473 "delta_indices": [
474 -30,
475 0
476 ],
477 "modality_keys": [
478 "top_camera-images-rgb_320_240",
479 "left_camera-images-rgb_320_240",
480 "right_camera-images-rgb_320_240"
481 ]
482 },
483 "state": {
484 "delta_indices": [
485 0
486 ],
487 "modality_keys": [
488 "left_wrist_eef",
489 "right_wrist_eef",
490 "left_gripper_pos",
491 "right_gripper_pos",
492 "left_joint_pos",
493 "right_joint_pos"
494 ]
495 },
496 "action": {
497 "delta_indices": [
498 0,
499 1,
500 2,
501 3,
502 4,
503 5,
504 6,
505 7,
506 8,
507 9,
508 10,
509 11,
510 12,
511 13,
512 14,
513 15,
514 16,
515 17,
516 18,
517 19,
518 20,
519 21,
520 22,
521 23,
522 24,
523 25,
524 26,
525 27,
526 28,
527 29,
528 30,
529 31,
530 32,
531 33,
532 34,
533 35,
534 36,
535 37,
536 38,
537 39
538 ],
539 "modality_keys": [
540 "left_wrist_eef",
541 "right_wrist_eef",
542 "left_gripper_pos",
543 "right_gripper_pos",
544 "left_joint_pos",
545 "right_joint_pos"
546 ],
547 "action_configs": [
548 {
549 "rep": "RELATIVE",
550 "type": "EEF",
551 "format": "XYZ_ROT6D",
552 "state_key": "left_wrist_eef"
553 },
554 {
555 "rep": "RELATIVE",
556 "type": "EEF",
557 "format": "XYZ_ROT6D",
558 "state_key": "right_wrist_eef"
559 },
560 {
561 "rep": "ABSOLUTE",
562 "type": "NON_EEF",
563 "format": "DEFAULT",
564 "state_key": "left_gripper_pos"
565 },
566 {
567 "rep": "ABSOLUTE",
568 "type": "NON_EEF",
569 "format": "DEFAULT",
570 "state_key": "right_gripper_pos"
571 },
572 {
573 "rep": "RELATIVE",
574 "type": "NON_EEF",
575 "format": "DEFAULT",
576 "state_key": "left_joint_pos"
577 },
578 {
579 "rep": "RELATIVE",
580 "type": "NON_EEF",
581 "format": "DEFAULT",
582 "state_key": "right_joint_pos"
583 }
584 ]
585 },
586 "language": {
587 "delta_indices": [
588 0
589 ],
590 "modality_keys": [
591 "annotation.task"
592 ]
593 }
594 },
595 "real_r1_pro_sharpa_relative_eef_maxinsights": {
596 "video": {
597 "delta_indices": [
598 -30,
599 0
600 ],
601 "modality_keys": [
602 "ego_view_cropratio_res320x240_freq30"
603 ]
604 },
605 "state": {
606 "delta_indices": [
607 0
608 ],
609 "modality_keys": [
610 "left_wrist_eef",
611 "right_wrist_eef",
612 "left_hand_joints",
613 "right_hand_joints"
614 ]
615 },
616 "action": {
617 "delta_indices": [
618 0,
619 1,
620 2,
621 3,
622 4,
623 5,
624 6,
625 7,
626 8,
627 9,
628 10,
629 11,
630 12,
631 13,
632 14,
633 15,
634 16,
635 17,
636 18,
637 19,
638 20,
639 21,
640 22,
641 23,
642 24,
643 25,
644 26,
645 27,
646 28,
647 29,
648 30,
649 31,
650 32,
651 33,
652 34,
653 35,
654 36,
655 37,
656 38,
657 39
658 ],
659 "modality_keys": [
660 "left_wrist_eef",
661 "right_wrist_eef",
662 "left_hand_joints",
663 "right_hand_joints"
664 ],
665 "action_configs": [
666 {
667 "rep": "RELATIVE",
668 "type": "EEF",
669 "format": "XYZ_ROT6D",
670 "state_key": "left_wrist_eef"
671 },
672 {
673 "rep": "RELATIVE",
674 "type": "EEF",
675 "format": "XYZ_ROT6D",
676 "state_key": "right_wrist_eef"
677 },
678 {
679 "rep": "ABSOLUTE",
680 "type": "NON_EEF",
681 "format": "DEFAULT",
682 "state_key": "left_hand_joints"
683 },
684 {
685 "rep": "ABSOLUTE",
686 "type": "NON_EEF",
687 "format": "DEFAULT",
688 "state_key": "right_hand_joints"
689 }
690 ]
691 },
692 "language": {
693 "delta_indices": [
694 0
695 ],
696 "modality_keys": [
697 "annotation.human.coarse_action"
698 ]
699 }
700 },
701 "xdof_relative_eef_relative_joint_subtask": {
702 "video": {
703 "delta_indices": [
704 -30,
705 0
706 ],
707 "modality_keys": [
708 "top_camera-images-rgb_320_240",
709 "left_camera-images-rgb_320_240",
710 "right_camera-images-rgb_320_240"
711 ]
712 },
713 "state": {
714 "delta_indices": [
715 0
716 ],
717 "modality_keys": [
718 "left_wrist_eef",
719 "right_wrist_eef",
720 "left_gripper_pos",
721 "right_gripper_pos",
722 "left_joint_pos",
723 "right_joint_pos"
724 ]
725 },
726 "action": {
727 "delta_indices": [
728 0,
729 1,
730 2,
731 3,
732 4,
733 5,
734 6,
735 7,
736 8,
737 9,
738 10,
739 11,
740 12,
741 13,
742 14,
743 15,
744 16,
745 17,
746 18,
747 19,
748 20,
749 21,
750 22,
751 23,
752 24,
753 25,
754 26,
755 27,
756 28,
757 29,
758 30,
759 31,
760 32,
761 33,
762 34,
763 35,
764 36,
765 37,
766 38,
767 39
768 ],
769 "modality_keys": [
770 "left_wrist_eef",
771 "right_wrist_eef",
772 "left_gripper_pos",
773 "right_gripper_pos",
774 "left_joint_pos",
775 "right_joint_pos"
776 ],
777 "action_configs": [
778 {
779 "rep": "RELATIVE",
780 "type": "EEF",
781 "format": "XYZ_ROT6D",
782 "state_key": "left_wrist_eef"
783 },
784 {
785 "rep": "RELATIVE",
786 "type": "EEF",
787 "format": "XYZ_ROT6D",
788 "state_key": "right_wrist_eef"
789 },
790 {
791 "rep": "ABSOLUTE",
792 "type": "NON_EEF",
793 "format": "DEFAULT",
794 "state_key": "left_gripper_pos"
795 },
796 {
797 "rep": "ABSOLUTE",
798 "type": "NON_EEF",
799 "format": "DEFAULT",
800 "state_key": "right_gripper_pos"
801 },
802 {
803 "rep": "RELATIVE",
804 "type": "NON_EEF",
805 "format": "DEFAULT",
806 "state_key": "left_joint_pos"
807 },
808 {
809 "rep": "RELATIVE",
810 "type": "NON_EEF",
811 "format": "DEFAULT",
812 "state_key": "right_joint_pos"
813 }
814 ]
815 },
816 "language": {
817 "delta_indices": [
818 0
819 ],
820 "modality_keys": [
821 "annotation.sub_task"
822 ]
823 }
824 },
825 "oxe_droid_relative_eef_relative_joint": {
826 "video": {
827 "delta_indices": [
828 -15,
829 0
830 ],
831 "modality_keys": [
832 "exterior_image_1_left",
833 "wrist_image_left"
834 ]
835 },
836 "state": {
837 "delta_indices": [
838 0
839 ],
840 "modality_keys": [
841 "eef_9d",
842 "gripper_position",
843 "joint_position"
844 ]
845 },
846 "action": {
847 "delta_indices": [
848 0,
849 1,
850 2,
851 3,
852 4,
853 5,
854 6,
855 7,
856 8,
857 9,
858 10,
859 11,
860 12,
861 13,
862 14,
863 15,
864 16,
865 17,
866 18,
867 19,
868 20,
869 21,
870 22,
871 23,
872 24,
873 25,
874 26,
875 27,
876 28,
877 29,
878 30,
879 31,
880 32,
881 33,
882 34,
883 35,
884 36,
885 37,
886 38,
887 39
888 ],
889 "modality_keys": [
890 "eef_9d",
891 "gripper_position",
892 "joint_position"
893 ],
894 "action_configs": [
895 {
896 "rep": "RELATIVE",
897 "type": "EEF",
898 "format": "XYZ_ROT6D",
899 "state_key": "eef_9d"
900 },
901 {
902 "rep": "ABSOLUTE",
903 "type": "NON_EEF",
904 "format": "DEFAULT",
905 "state_key": "gripper_position"
906 },
907 {
908 "rep": "RELATIVE",
909 "type": "NON_EEF",
910 "format": "DEFAULT",
911 "state_key": "joint_position"
912 }
913 ]
914 },
915 "language": {
916 "delta_indices": [
917 0
918 ],
919 "modality_keys": [
920 "annotation.language.language_instruction"
921 ]
922 }
923 }
924 },
925 "use_percentiles": true,
926 "use_mean_std": false,
927 "image_crop_size": [
928 230,
929 230
930 ],
931 "image_target_size": [
932 256,
933 256
934 ],
935 "formalize_language": true,
936 "max_state_dim": 132,
937 "max_action_dim": 132,
938 "apply_sincos_state_encoding": false,
939 "color_jitter_params": {
940 "brightness": 0.3,
941 "contrast": 0.4,
942 "saturation": 0.5,
943 "hue": 0.08
944 },
945 "random_rotation_angle": 0,
946 "letter_box_transform": false,
947 "exclude_state": false,
948 "state_dropout_prob": 0.2,
949 "use_albumentations": true,
950 "shortest_image_edge": 256,
951 "crop_fraction": 0.95,
952 "max_action_horizon": 40,
953 "use_relative_action": true
954 }
955 }
956