ChaoyiPan commited on
Commit
ecb9a9e
·
verified ·
1 Parent(s): 68b85c6

Remove rejected EgoVerse hand-only example and previews

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/README.md +0 -19
  2. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/SHA256SUMS +0 -53
  3. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/config.input.json +0 -87
  4. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/0/config.yaml +0 -129
  5. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/0/sanity_check_drift.mp4 +0 -3
  6. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/0/trajectory_kinematic.npz +0 -3
  7. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/0/trajectory_mjwp_fast.npz +0 -3
  8. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/0/visualization_mjwp_fast.mp4 +0 -3
  9. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/elastomer.STL +0 -3
  10. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/elastomer_surface.STL +0 -3
  11. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_DP.STL +0 -3
  12. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_DP_visual.STL +0 -3
  13. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_MCP_VL.STL +0 -0
  14. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_MCP_VL_visual.STL +0 -3
  15. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_MP.STL +0 -0
  16. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_MP_visual.STL +0 -3
  17. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_PP.STL +0 -0
  18. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_PP_visual.STL +0 -3
  19. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_hand_C_MC.STL +0 -3
  20. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_hand_C_MC_visual_.STL +0 -3
  21. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_pinky_MC.STL +0 -0
  22. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_pinky_MC_visual.STL +0 -0
  23. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_thumb_CMC_VL.STL +0 -3
  24. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_thumb_DP.STL +0 -0
  25. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_thumb_DP_visual.STL +0 -3
  26. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_thumb_MC.STL +0 -3
  27. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_thumb_MCP_VL_visual.STL +0 -3
  28. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_thumb_MC_visual.STL +0 -3
  29. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_thumb_PP.STL +0 -0
  30. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_thumb_PP_visual.STL +0 -3
  31. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/thumb_elastomer.STL +0 -3
  32. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/thumb_elastomer_surface.STL +0 -3
  33. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/scene.xml +0 -304
  34. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/task_info.json +0 -19
  35. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/evidence/control_replay.json +0 -76
  36. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/evidence/evaluation_control_replay.json +0 -116
  37. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/evidence/native/comparison.mp4 +0 -3
  38. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/evidence/native/evaluation.json +0 -971
  39. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/evidence/native/thumbnail.jpg +0 -3
  40. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/evidence/replay_saved_controls.py +0 -63
  41. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/evidence/trajectory_control_replay.npz +0 -3
  42. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/hydra/run_mjwp_fast.log +0 -0
  43. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/manifest.json +0 -251
  44. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/metrics.json +0 -27
  45. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/parameters.json +0 -79
  46. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/run.log +0 -43
  47. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/runtime_source/examples/run_mjwp_fast.py +0 -852
  48. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/runtime_source/sha256.json +0 -9
  49. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/runtime_source/spider/config.py +0 -578
  50. agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/runtime_source/spider/io.py +0 -202
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/README.md DELETED
@@ -1,19 +0,0 @@
1
- # EgoVerse: successful fresh full Sharpa hand tracking
2
-
3
- **PASS: full native EgoVerse hand trajectory and independent control replay.** This example tracks 120 real left-hand observations over 3.967 seconds. It has no object or cloth simulation and does not claim cloth-folding success.
4
-
5
- The optimizer executed 400 MuJoCo Warp steps at 100 Hz; 397 source steps were scored, excluding the final hold padding. Palm/fingertip error is 1.19 mm mean, 3.12 mm at the 95th percentile, and 14.96 mm maximum. Maximum palm orientation error is 0.00587 rad. Source motion excursion is preserved at 99.93%. The unchanged evaluator limits are 20 mm / 0.2 rad.
6
-
7
- Independent CPU MuJoCo integration of all 400 saved controls reproduced the GPU hand sites within 7.99e-8 metres, with no state corrections after initialization. Controls stayed within their enabled actuator ranges.
8
-
9
- - [Reference versus executed motion](evidence/native/comparison.mp4)
10
- - [Independent trajectory evaluation](evidence/native/evaluation.json)
11
- - [Control replay evidence](evidence/control_replay.json)
12
- - [Replay script](evidence/replay_saved_controls.py)
13
- - [Run settings](config.input.json)
14
- - [Executed trajectory](dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/0/trajectory_mjwp_fast.npz)
15
- - [Runtime source snapshot and hashes](runtime_source/)
16
-
17
- This tuned simulation uses `hand_control_mode=inverse_dynamics`, `hand_gain_scale=10`, 128 samples, 8 iterations, a 0.8-second horizon and 0.1-second control/knot intervals. Gain scaling changes hand servo gains only; exact effective gains are recorded in the control-replay report. Replaying controls requires these recorded settings in addition to the portable scene.
18
-
19
- Run `evidence/replay_saved_controls.py` from the SPIDER environment to regenerate the independent replay. The script resolves its dataset relative to its own location. The original source observations, conversion provenance and Sharpa assets are also supplied in the earlier 2026-10-06 example release.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/SHA256SUMS DELETED
@@ -1,53 +0,0 @@
1
- 214664cdd4e2b1b7736a54b5f2d8e07f376936bf8ad42329f9a34f259e042f77 README.md
2
- a109177c78bb4b2c269f71f78f159e6ae8e39be78d72f59ee2aa031b60d49c32 config.input.json
3
- 3691028fb886ffd880661144622123398f60bbf529e10760dd032089e9bb47c8 dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/0/config.yaml
4
- 54357a3bb83005ad00c711e362b300157c4b0527dc4f22ad68f2fc0b3d97aeca dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/0/sanity_check_drift.mp4
5
- fb8aafe61241e8efc9866e9c82ecb0306394c0cc497f3a0b6ffbb3d7b3c9c20a dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/0/trajectory_kinematic.npz
6
- d8c6d098f4ee03f7f05e876222098a3cd855874cab3e3a0ee43c00bd72466695 dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/0/trajectory_mjwp_fast.npz
7
- 0342d322489a06277a78e26cabc86d447c6b2e547ae9f61cb969bbf6abd971ab dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/0/visualization_mjwp_fast.mp4
8
- 1611f42070a8fb34226508417bb930af53bbe9e3b619113136b60cb0595ca4ed dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/elastomer.STL
9
- 7f88a20e862895e3bf97d35c600b9eb5c2f2e49f8918ef9fe07bd82e6572fae8 dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/elastomer_surface.STL
10
- 09ae5e485d42c651c70d025ef6bcd000f727dd1535836d7d8fa9de0686db4583 dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_DP.STL
11
- f21c6a8d12a21b29ff05127ec1987135cc966a283707165d93a30f9e74d4a0ed dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_DP_visual.STL
12
- 587e34505de3b4f5b7131e6554e0ab92a18a3c28d250ea46eaf728c4539bf77d dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_MCP_VL.STL
13
- 0e8b15c4d8426a26cb222b8c1592f5e913c079aebdc0abaecda273b7d8ed0f3a dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_MCP_VL_visual.STL
14
- 1f0a96877af9ec35f20941ae696e9e22cedc6e341703657b0c6d0ebc982897a9 dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_MP.STL
15
- 0e9ac7a8208ab2478cdff8ae40036b93e040a05f89ae0c43ece83a62fdf15293 dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_MP_visual.STL
16
- 6e4434712a5c668dbeda169e727629e95d4a9492fca431126543109451e2484f dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_PP.STL
17
- 881e6eeeaecbf0a8bd6b11fa7f41a5aa139900635fe82b765586460c9a9c1580 dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_PP_visual.STL
18
- 01e17ed8392d99e76d1c00a849749610c3f7715f18ea12b107ba56eba3541fee dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_hand_C_MC.STL
19
- 0e7c42e037b4bc94e35afd2f1c1d3c569d7b14554911bb380fa631691b5d2550 dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_hand_C_MC_visual_.STL
20
- f7a7192ae246a6b88f9096332216013c8f6be507ee0dc9bfd4444518453b0c13 dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_pinky_MC.STL
21
- 98b520aa40d6f1fff0f433cfc6ef2baddb550a6452a8753301934812d111bbf1 dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_pinky_MC_visual.STL
22
- 9c862ce10c4958ef967ed46fed837fafb20dd2c84a23ef9ee4f793576ef1fb93 dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_thumb_CMC_VL.STL
23
- 7bd0a9187d9cca7c3136d471784db39a3593b5264e16586002f73db42292dc6e dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_thumb_DP.STL
24
- 7bac46dd5699d6b2fed60fdb384199cdab7407343d61db3fc8737866096fa440 dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_thumb_DP_visual.STL
25
- 3f15b81f22feec350ae2f60a7c69ca36028bd580d6cf4722d4f759d401ea02f9 dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_thumb_MC.STL
26
- 47f990c04f09174666773687f9e4e1b4031dd150481c1d26719bbba874ae0709 dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_thumb_MCP_VL_visual.STL
27
- da4fb47f6126ca3b653689aea6d958e08dde7d0c232a3043773e18f447aa121a dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_thumb_MC_visual.STL
28
- 2847673ad531a6d1fe3d12b3433d080358727def9acade9f0319c18579676b13 dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_thumb_PP.STL
29
- 7f4a561b611af82e27bb5c8d6838d574e71e2ede42d25edf1f57c2b54ebde072 dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_thumb_PP_visual.STL
30
- 8fb5b2036eda97fe4c2ce50ad8e021775185961c478c0d6f25f5a776588d9b14 dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/thumb_elastomer.STL
31
- 1b25efc390f9db9eca87ed167e7baa2e3220a417bbd49924beb077da458c6241 dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/thumb_elastomer_surface.STL
32
- 1219685ac913e7954a0785a2befba6ecb40eb0c30024192ec7c64c9f2dfd50b0 dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/scene.xml
33
- c63a6e6d5edba47ca7ce397b993f7d75fcf916a3661aea471959c71f685cc3c1 dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/task_info.json
34
- debd9c9dd9a820fd50de01f3e03601d21bbb59f571a9655e48e86ae23b8b522b evidence/control_replay.json
35
- 986aecd10e2c0faef161f19cbbf179d51eb9b1d7e45de17d371ea2f7b5e91a58 evidence/evaluation_control_replay.json
36
- e945e7dcdf5589380321f2335054b8dc907794f5013ef99e9fbd06ac9543b4db evidence/native/comparison.mp4
37
- 49a0ab3e2a377fca2bf3eb43b38df5245354fbaf3b8a6a32ff0439593d33d7db evidence/native/evaluation.json
38
- cafeba69360c2cd179fb7b80d29bd3108e5561124da1407edde7cc75a673eb87 evidence/native/thumbnail.jpg
39
- 366b795cde33883fe5da8ac7ce3b24fc5f4d746d66a5c54891424a652886e03a evidence/replay_saved_controls.py
40
- 158424c01bbccfff10b46f53b22190bd21c26583808c3966169cf90beedb25b5 evidence/trajectory_control_replay.npz
41
- e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855 hydra/run_mjwp_fast.log
42
- 683f712cc705a4c66592161489bf7ef1b80e7b2f61db0f1ef666066ebcde9496 manifest.json
43
- b16a0dd4e0b387fa622ced501c449f362d28e9829fd684cad1e62eb776104530 metrics.json
44
- 3c019f440e300a6034c3eae6418765b5b4f6fab1c7819f3c3621920377730a80 parameters.json
45
- 339dea8fccf5feaf27fede8174e8c499e967dd566f1b70494cc99870b35b63bc run.log
46
- e016907910699d80997c69f9f72eeedd9c420e540bd6945afd5a9a486f09960a runtime_source/examples/run_mjwp_fast.py
47
- 4f12de01e1d1faff3608856be53a10d444d7091dbd448c74ab1c2a032c5a0f72 runtime_source/sha256.json
48
- dc555763920efb095625791cfe7b85867ff61477a77ac5fb2b0f8334bdaae09b runtime_source/spider/config.py
49
- 94bd7204c0310825ba12609860f87988b2981bc360ef69a917b6412f8d2c39c1 runtime_source/spider/io.py
50
- 7dbe311531dde912c586d8626838f78850e75510835bdabcd5e35279f974d1cb runtime_source/spider/optimizers/sampling.py
51
- 6c9249fa9024287fdf115d33a0c8ccec4ecd74cfbb4e978770113937645c07b2 runtime_source/spider/optimizers/sampling_fast.py
52
- 78e5a0afc537929d3172a676f0c54d40fe1eca9b68af290561e666ed001f99f3 runtime_source/spider/pipeline.py
53
- 4ea554adbf2ece1bc5cb4a27d5f81e37182a5f3f449db5ed23dbcde17a71e330 runtime_source/spider/simulators/mjwp.py
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/config.input.json DELETED
@@ -1,87 +0,0 @@
1
- {
2
- "seed": 0,
3
- "sanity_check_seconds": 3.0,
4
- "sanity_check_pos_thresh": 0.01,
5
- "sanity_check_rot_thresh": 0.3,
6
- "simulator": "mjwp",
7
- "device": "cuda:0",
8
- "sim_dt": 0.01,
9
- "ctrl_dt": 0.1,
10
- "render_dt": 0.02,
11
- "horizon": 0.8,
12
- "knot_dt": 0.1,
13
- "nconmax_per_env": 100,
14
- "njmax_per_env": 350,
15
- "num_dyn": 1,
16
- "num_dr": 1,
17
- "pair_margin_range": [
18
- -0.005,
19
- 0.005
20
- ],
21
- "xy_offset_range": [
22
- -0.005,
23
- 0.005
24
- ],
25
- "perturb_force": 0.0,
26
- "perturb_torque": 0.0,
27
- "contact_guidance": false,
28
- "hand_control_mode": "inverse_dynamics",
29
- "hand_gain_scale": 10,
30
- "init_pos_actuator_gain": 10.0,
31
- "init_pos_actuator_bias": 10.0,
32
- "init_rot_actuator_gain": 0.1,
33
- "init_rot_actuator_bias": 0.1,
34
- "guidance_decay_ratio": 0.5,
35
- "gibbs_sampling": false,
36
- "num_samples": 128,
37
- "temperature": 0.3,
38
- "max_num_iterations": 8,
39
- "improvement_threshold": 0.01,
40
- "improvement_check_steps": 1,
41
- "terminate_resample": false,
42
- "object_pos_threshold": 0.1,
43
- "object_rot_threshold": 0.3,
44
- "max_local_retries": 1,
45
- "max_revert_depth": 1,
46
- "base_pos_threshold": 0.02,
47
- "base_rot_threshold": 0.2,
48
- "use_torch_compile": false,
49
- "first_ctrl_noise_scale": 0.5,
50
- "last_ctrl_noise_scale": 1.0,
51
- "final_noise_scale": 0.1,
52
- "exploit_ratio": 0.01,
53
- "exploit_noise_scale": 0.01,
54
- "joint_noise_scale": 0.02,
55
- "pos_noise_scale": 0.003,
56
- "rot_noise_scale": 0.005,
57
- "use_rl_reward": false,
58
- "base_pos_rew_scale": 1.0,
59
- "base_rot_rew_scale": 0.3,
60
- "joint_rew_scale": 0.03,
61
- "pos_rew_scale": 1.0,
62
- "rot_rew_scale": 0.3,
63
- "vel_rew_scale": 0.0001,
64
- "terminal_rew_scale": 1.0,
65
- "contact_rew_scale": 0.0,
66
- "show_viewer": false,
67
- "viewer": "mujoco",
68
- "wait_on_finish": false,
69
- "rerun_spawn": false,
70
- "save_video": true,
71
- "save_info": true,
72
- "save_rerun": false,
73
- "save_viser": false,
74
- "save_metrics": true,
75
- "save_config": true,
76
- "trace_dt": 0.02,
77
- "num_trace_uniform_samples": 4,
78
- "num_trace_topk_samples": 2,
79
- "dataset_dir": "/home/ubuntu/sky_workdir/example_datasets/full_retarget_campaign/runs/egoverse_trial2/dataset",
80
- "dataset_name": "egoverse_hand",
81
- "task": "fold_clothes_left",
82
- "embodiment_type": "left",
83
- "robot_type": "sharpa",
84
- "data_id": 0,
85
- "ref_dt": 0.03333333333333333,
86
- "max_sim_steps": 400
87
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/0/config.yaml DELETED
@@ -1,129 +0,0 @@
1
- robot_type: sharpa
2
- embodiment_type: left
3
- task: fold_clothes_left
4
- seed: 0
5
- hand_only: true
6
- sanity_check_seconds: 3.0
7
- sanity_check_pos_thresh: 0.01
8
- sanity_check_rot_thresh: 0.3
9
- dataset_dir: /home/ubuntu/sky_workdir/example_datasets/full_retarget_campaign/runs/egoverse_trial2/dataset
10
- dataset_name: egoverse_hand
11
- data_id: 0
12
- model_path: /home/ubuntu/sky_workdir/example_datasets/full_retarget_campaign/runs/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/0/../scene.xml
13
- data_path: /home/ubuntu/sky_workdir/example_datasets/full_retarget_campaign/runs/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/0/trajectory_kinematic.npz
14
- load_config_path: /home/ubuntu/sky_workdir/example_datasets/full_retarget_campaign/runs/egoverse_trial2/config.input.json
15
- simulator: mjwp
16
- device: cuda:0
17
- sim_dt: 0.01
18
- ctrl_dt: 0.1
19
- ref_dt: 0.03333333333333333
20
- render_dt: 0.02
21
- horizon: 0.8
22
- knot_dt: 0.1
23
- max_sim_steps: 400
24
- nconmax_per_env: 100
25
- njmax_per_env: 350
26
- num_dyn: 1
27
- num_dr: 1
28
- pair_margin_range:
29
- - -0.005
30
- - 0.005
31
- xy_offset_range:
32
- - -0.005
33
- - 0.005
34
- perturb_force: 0.0
35
- perturb_torque: 0.0
36
- contact_guidance: false
37
- hand_control_mode: inverse_dynamics
38
- hand_gain_scale: 10
39
- object_pos_actuator_names:
40
- - right_object_pos_x
41
- - right_object_pos_y
42
- - right_object_pos_z
43
- - left_object_pos_x
44
- - left_object_pos_y
45
- - left_object_pos_z
46
- object_rot_actuator_names:
47
- - right_object_rot_x
48
- - right_object_rot_y
49
- - right_object_rot_z
50
- - left_object_rot_x
51
- - left_object_rot_y
52
- - left_object_rot_z
53
- object_action_dims: 0
54
- object_actuator_ids: []
55
- object_actuator_names: []
56
- init_pos_actuator_gain: 10.0
57
- init_pos_actuator_bias: 10.0
58
- init_rot_actuator_gain: 0.1
59
- init_rot_actuator_bias: 0.1
60
- guidance_decay_ratio: 0.5
61
- gibbs_sampling: false
62
- num_samples: 128
63
- temperature: 0.3
64
- max_num_iterations: 8
65
- improvement_threshold: 0.01
66
- improvement_check_steps: 1
67
- terminate_resample: false
68
- object_pos_threshold: 0.1
69
- object_rot_threshold: 0.3
70
- max_local_retries: 1
71
- max_revert_depth: 1
72
- base_pos_threshold: 0.02
73
- base_rot_threshold: 0.2
74
- use_torch_compile: false
75
- first_ctrl_noise_scale: 0.5
76
- last_ctrl_noise_scale: 1.0
77
- final_noise_scale: 0.1
78
- exploit_ratio: 0.01
79
- exploit_noise_scale: 0.01
80
- joint_noise_scale: 0.02
81
- pos_noise_scale: 0.003
82
- rot_noise_scale: 0.005
83
- use_rl_reward: false
84
- base_pos_rew_scale: 1.0
85
- base_rot_rew_scale: 0.3
86
- joint_rew_scale: 0.03
87
- pos_rew_scale: 1.0
88
- rot_rew_scale: 0.3
89
- vel_rew_scale: 0.0001
90
- terminal_rew_scale: 1.0
91
- contact_rew_scale: 0.0
92
- show_viewer: false
93
- viewer: mujoco
94
- wait_on_finish: false
95
- rerun_spawn: false
96
- save_video: true
97
- save_info: true
98
- save_rerun: false
99
- save_viser: false
100
- save_metrics: true
101
- save_config: true
102
- trace_dt: 0.02
103
- num_trace_uniform_samples: 4
104
- num_trace_topk_samples: 2
105
- trace_site_ids:
106
- - 32
107
- - 6
108
- - 12
109
- - 18
110
- - 24
111
- - 30
112
- contact_order: []
113
- hand_contact_site_ids: []
114
- right_contact_indices: []
115
- left_contact_indices: []
116
- right_pos_ctrl_ids: []
117
- left_pos_ctrl_ids: []
118
- contact_len: 0
119
- horizon_steps: 80
120
- knot_steps: 10
121
- ref_steps: 3
122
- ctrl_steps: 10
123
- nq_obj: 7
124
- nq: 35
125
- nv: 34
126
- nu: 28
127
- npair: 0
128
- beta_traj: 0.7498942093324559
129
- output_dir: /home/ubuntu/sky_workdir/example_datasets/full_retarget_campaign/runs/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/0/sanity_check_drift.mp4 DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:54357a3bb83005ad00c711e362b300157c4b0527dc4f22ad68f2fc0b3d97aeca
3
- size 7386
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/0/trajectory_kinematic.npz DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:fb8aafe61241e8efc9866e9c82ecb0306394c0cc497f3a0b6ffbb3d7b3c9c20a
3
- size 67000
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/0/trajectory_mjwp_fast.npz DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:d8c6d098f4ee03f7f05e876222098a3cd855874cab3e3a0ee43c00bd72466695
3
- size 1033010
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/0/visualization_mjwp_fast.mp4 DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:0342d322489a06277a78e26cabc86d447c6b2e547ae9f61cb969bbf6abd971ab
3
- size 305932
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/elastomer.STL DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:1611f42070a8fb34226508417bb930af53bbe9e3b619113136b60cb0595ca4ed
3
- size 271584
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/elastomer_surface.STL DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:7f88a20e862895e3bf97d35c600b9eb5c2f2e49f8918ef9fe07bd82e6572fae8
3
- size 362534
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_DP.STL DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:09ae5e485d42c651c70d025ef6bcd000f727dd1535836d7d8fa9de0686db4583
3
- size 101684
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_DP_visual.STL DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:f21c6a8d12a21b29ff05127ec1987135cc966a283707165d93a30f9e74d4a0ed
3
- size 320284
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_MCP_VL.STL DELETED
Binary file (75.2 kB)
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_MCP_VL_visual.STL DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:0e8b15c4d8426a26cb222b8c1592f5e913c079aebdc0abaecda273b7d8ed0f3a
3
- size 322584
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_MP.STL DELETED
Binary file (37.1 kB)
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_MP_visual.STL DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:0e9ac7a8208ab2478cdff8ae40036b93e040a05f89ae0c43ece83a62fdf15293
3
- size 369384
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_PP.STL DELETED
Binary file (82.5 kB)
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_PP_visual.STL DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:881e6eeeaecbf0a8bd6b11fa7f41a5aa139900635fe82b765586460c9a9c1580
3
- size 559934
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_hand_C_MC.STL DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:01e17ed8392d99e76d1c00a849749610c3f7715f18ea12b107ba56eba3541fee
3
- size 174184
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_hand_C_MC_visual_.STL DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:0e7c42e037b4bc94e35afd2f1c1d3c569d7b14554911bb380fa631691b5d2550
3
- size 214484
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_pinky_MC.STL DELETED
Binary file (42 kB)
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_pinky_MC_visual.STL DELETED
Binary file (66.2 kB)
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_thumb_CMC_VL.STL DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:9c862ce10c4958ef967ed46fed837fafb20dd2c84a23ef9ee4f793576ef1fb93
3
- size 317984
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_thumb_DP.STL DELETED
Binary file (66.5 kB)
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_thumb_DP_visual.STL DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:7bac46dd5699d6b2fed60fdb384199cdab7407343d61db3fc8737866096fa440
3
- size 335484
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_thumb_MC.STL DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:3f15b81f22feec350ae2f60a7c69ca36028bd580d6cf4722d4f759d401ea02f9
3
- size 171084
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_thumb_MCP_VL_visual.STL DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:47f990c04f09174666773687f9e4e1b4031dd150481c1d26719bbba874ae0709
3
- size 300784
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_thumb_MC_visual.STL DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:da4fb47f6126ca3b653689aea6d958e08dde7d0c232a3043773e18f447aa121a
3
- size 580084
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_thumb_PP.STL DELETED
Binary file (83.8 kB)
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/left_thumb_PP_visual.STL DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:7f4a561b611af82e27bb5c8d6838d574e71e2ede42d25edf1f57c2b54ebde072
3
- size 772184
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/thumb_elastomer.STL DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:8fb5b2036eda97fe4c2ce50ad8e021775185961c478c0d6f25f5a776588d9b14
3
- size 270684
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/_assets/f5579dfee924f2d5/thumb_elastomer_surface.STL DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:1b25efc390f9db9eca87ed167e7baa2e3220a417bbd49924beb077da458c6241
3
- size 506484
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/scene.xml DELETED
@@ -1,304 +0,0 @@
1
- <mujoco model="sharpa_left">
2
- <compiler angle="radian" strippath="false" />
3
- <default>
4
- <joint armature="1" />
5
- <geom contype="0" conaffinity="0" condim="1" density="800" />
6
- <site group="3" rgba="1 0 0 1" />
7
- <general biastype="affine" gainprm="300" biasprm="0 -300 1" />
8
- </default>
9
- <asset>
10
- <texture type="skybox" colorspace="auto" builtin="gradient" rgb1="1 1 1" rgb2="1 1 1" width="512" height="3072" />
11
- <texture type="2d" colorspace="auto" name="left_groundplane" builtin="checker" mark="edge" rgb1="0.2 0.3 0.4" rgb2="0.1 0.2 0.3" markrgb="0.8 0.8 0.8" width="300" height="300" />
12
- <material name="left_groundplane" texture="left_groundplane" texuniform="true" texrepeat="5 5" reflectance="0.2" />
13
- <mesh name="left_hand_C_MC_visual" file="_assets/f5579dfee924f2d5/left_hand_C_MC_visual_.STL" />
14
- <mesh name="left_hand_C_MC" file="_assets/f5579dfee924f2d5/left_hand_C_MC.STL" />
15
- <mesh name="left_thumb_CMC_VL_visual" file="_assets/f5579dfee924f2d5/left_thumb_CMC_VL.STL" />
16
- <mesh name="left_thumb_MC_visual" file="_assets/f5579dfee924f2d5/left_thumb_MC_visual.STL" />
17
- <mesh name="left_thumb_MC" file="_assets/f5579dfee924f2d5/left_thumb_MC.STL" />
18
- <mesh name="left_thumb_MCP_VL_visual" file="_assets/f5579dfee924f2d5/left_thumb_MCP_VL_visual.STL" />
19
- <mesh name="left_thumb_PP_visual" file="_assets/f5579dfee924f2d5/left_thumb_PP_visual.STL" />
20
- <mesh name="left_thumb_PP" file="_assets/f5579dfee924f2d5/left_thumb_PP.STL" />
21
- <mesh name="left_thumb_DP_visual" file="_assets/f5579dfee924f2d5/left_thumb_DP_visual.STL" />
22
- <mesh name="left_thumb_DP" file="_assets/f5579dfee924f2d5/left_thumb_DP.STL" />
23
- <mesh name="left_thumb_elastomer_surface" file="_assets/f5579dfee924f2d5/thumb_elastomer_surface.STL" />
24
- <mesh name="left_thumb_elastomer" file="_assets/f5579dfee924f2d5/thumb_elastomer.STL" />
25
- <mesh name="left_MCP_VL_visual" file="_assets/f5579dfee924f2d5/left_MCP_VL_visual.STL" />
26
- <mesh name="left_MCP_VL" file="_assets/f5579dfee924f2d5/left_MCP_VL.STL" />
27
- <mesh name="left_PP_visual" file="_assets/f5579dfee924f2d5/left_PP_visual.STL" />
28
- <mesh name="left_PP" file="_assets/f5579dfee924f2d5/left_PP.STL" />
29
- <mesh name="left_MP_visual" file="_assets/f5579dfee924f2d5/left_MP_visual.STL" />
30
- <mesh name="left_MP" file="_assets/f5579dfee924f2d5/left_MP.STL" />
31
- <mesh name="left_DP_visual" file="_assets/f5579dfee924f2d5/left_DP_visual.STL" />
32
- <mesh name="left_DP" file="_assets/f5579dfee924f2d5/left_DP.STL" />
33
- <mesh name="left_elastomer_surface" file="_assets/f5579dfee924f2d5/elastomer_surface.STL" />
34
- <mesh name="left_elastomer" file="_assets/f5579dfee924f2d5/elastomer.STL" />
35
- <mesh name="left_pinky_MC_visual" file="_assets/f5579dfee924f2d5/left_pinky_MC_visual.STL" />
36
- <mesh name="left_pinky_MC" file="_assets/f5579dfee924f2d5/left_pinky_MC.STL" />
37
- </asset>
38
- <worldbody>
39
- <geom name="floor" size="0 0 0.05" pos="0 0 -0.001" type="plane" material="left_groundplane" />
40
- <camera name="front" pos="0.031 0.941 0.844" quat="0.0150738 0.00673439 0.407848 0.912901" mode="trackcom" />
41
- <body name="left_base_tx">
42
- <inertial pos="0 0 0" mass="0.1" diaginertia="0.01 0.01 0.01" />
43
- <joint name="left_pos_x" pos="0 0 0" axis="1 0 0" type="slide" range="-5 5" />
44
- <camera name="left_track" pos="0.868 -0.348 -0.175" quat="0.506363 0.51196 0.493349 0.487956" />
45
- <body name="left_base_ty">
46
- <inertial pos="0 0 0" mass="0.1" diaginertia="0.01 0.01 0.01" />
47
- <joint name="left_pos_y" pos="0 0 0" axis="0 1 0" type="slide" range="-5 5" />
48
- <body name="left_base_tz">
49
- <inertial pos="0 0 0" mass="0.1" diaginertia="0.01 0.01 0.01" />
50
- <joint name="left_pos_z" pos="0 0 0" axis="0 0 1" type="slide" range="-5 5" />
51
- <body name="left_base_roll">
52
- <inertial pos="0 0 0" mass="0.1" diaginertia="0.01 0.01 0.01" />
53
- <joint name="left_rot_x" pos="0 0 0" axis="1 0 0" range="-6.28 6.28" />
54
- <body name="left_base_pitch">
55
- <inertial pos="0 0 0" mass="0.1" diaginertia="0.01 0.01 0.01" />
56
- <joint name="left_rot_y" pos="0 0 0" axis="0 1 0" range="-6.28 6.28" />
57
- <body name="left_base_yaw">
58
- <inertial pos="0 0 0" mass="0.1" diaginertia="0.01 0.01 0.01" />
59
- <joint name="left_rot_z" pos="0 0 0" axis="0 0 1" range="-6.28 6.28" />
60
- <body name="left_hand_C_MC">
61
- <inertial pos="0.0060282 0.00091891 0.038914" mass="0.72045" diaginertia="0.00071328 0.00081371 0.00020795" />
62
- <geom name="left_hand_C_MC_visual" type="mesh" group="1" density="0" rgba="0.79216 0.81961 0.93333 1" mesh="left_hand_C_MC_visual" />
63
- <geom name="collision_hand_left_palm_0" size="0.016 0.035 0.035" pos="0 0 0.05" quat="0 0 1 0" type="box" group="3" rgba="0 1 0 1" />
64
- <site name="left_palm" pos="0 0 0.05" size="0.01 0.02 0.03" type="box" />
65
- <body name="left_thumb_CMC_VL" pos="0.01 -0.026 0.0212" quat="0.707105 0 0 -0.707108">
66
- <inertial pos="0 0 0" mass="1e-06" diaginertia="2e-12 2e-12 2e-12" />
67
- <joint name="left_thumb_CMC_FE" pos="0 0 0" axis="0 0 1" range="-0.1745 1.9199" />
68
- <geom type="mesh" group="1" density="0" rgba="0.79216 0.81961 0.93333 1" mesh="left_thumb_CMC_VL_visual" />
69
- <site name="left_thumb_mcp" pos="0 0 0" />
70
- <body name="left_thumb_MC" pos="0 -0.005 0" quat="0.65328 0.653282 -0.270598 0.270599">
71
- <inertial pos="0.0385056 -0.0105846 -0.00991873" quat="0.158428 0.667584 0.222963 0.692474" mass="0.116" diaginertia="3.83518e-05 3.34032e-05 1.641e-05" />
72
- <joint name="left_thumb_CMC_AA" pos="0 0 0" axis="0 0 1" range="-0.3491 0.3491" />
73
- <geom name="left_thumb_MC_visual" type="mesh" group="1" density="0" rgba="0.79216 0.81961 0.93333 1" mesh="left_thumb_MC_visual" />
74
- <geom name="collision_hand_left_thumb_2" size="0.015 0.0330189" pos="0.0325 -0.003 -0.005" quat="0.651372 0.0697427 0.755546 0" type="capsule" group="3" rgba="0 1 0 1" />
75
- <body name="left_thumb_MCP_VL" pos="0.065 -0.006 -0.010392" quat="0.965926 -0.25882 0 0">
76
- <inertial pos="4.8e-09 0.000869809 -3.6e-09" mass="0.001" diaginertia="2e-12 2e-12 2e-12" />
77
- <joint name="left_thumb_MCP_FE" pos="0 0 0" axis="0 0 1" range="-0.5236 1.3963" />
78
- <geom type="mesh" group="1" density="0" rgba="0.79216 0.81961 0.93333 1" mesh="left_thumb_MCP_VL_visual" />
79
- <site name="left_thumb_pip" pos="0 0 0" />
80
- <body name="left_thumb_PP" quat="0.707105 0.707108 0 0">
81
- <inertial pos="0.0219483 -0.00015581 0.00149446" quat="0.526585 0.473016 0.473458 0.524215" mass="0.04955" diaginertia="4.60001e-06 4.35675e-06 2.57733e-06" />
82
- <joint name="left_thumb_MCP_AA" pos="0 0 0" axis="0 0 1" range="-0.3491 0.3491" />
83
- <geom name="left_thumb_PP_visual" type="mesh" group="1" density="0" rgba="0.79216 0.81961 0.93333 1" mesh="left_thumb_PP_visual" />
84
- <geom name="collision_hand_left_thumb_1" size="0.012 0.0195" pos="0.0195 0 0" quat="0.707107 0 0.707107 0" type="capsule" group="3" rgba="0 1 0 1" />
85
- <body name="left_thumb_DP" pos="0.039 0 0" quat="0.707105 -0.707108 0 0">
86
- <inertial pos="0.00985829 -0.0101254 0.00044771" quat="0.649916 0.20258 -0.647399 0.342703" mass="0.006986" diaginertia="8.79989e-07 8.08573e-07 2.50588e-07" />
87
- <joint name="left_thumb_IP" pos="0 0 0" axis="0 0 1" range="0 1.7453" />
88
- <geom name="left_thumb_DP_visual" type="mesh" group="1" density="0" rgba="0.79216 0.81961 0.93333 1" mesh="left_thumb_DP_visual" />
89
- <geom quat="0.5 0.5 -0.5 0.5" type="mesh" group="1" density="0" rgba="0.2 1 0.2 1" mesh="left_thumb_elastomer" />
90
- <geom name="collision_hand_left_thumb_0" size="0.007 0.0112805" pos="0.011 -0.0005 0" quat="0.707107 0.15671 0.689523 0" type="capsule" group="3" rgba="0 1 0 1" />
91
- <site name="left_thumb_dip" pos="0 0 0" />
92
- <site name="left_thumb_tip" pos="0.015 0 0" />
93
- <site name="track_hand_left_thumb_tip" pos="0.015 0 0" />
94
- <site name="trace_hand_left_thumb_tip" pos="0.015 0 0" />
95
- </body>
96
- </body>
97
- </body>
98
- </body>
99
- </body>
100
- <body name="left_index_MCP_VL" pos="0.001 -0.0303 0.0957" quat="0.499998 -0.500002 -0.5 -0.5">
101
- <inertial pos="4.8e-09 0.00086981 -3.6e-09" mass="0.001" diaginertia="2e-12 2e-12 2e-12" />
102
- <joint name="left_index_MCP_FE" pos="0 0 0" axis="0 0 1" range="-0.174533 1.5708" />
103
- <geom name="left_index_MCP_VL_visual" type="mesh" group="1" density="0" rgba="0.79216 0.81961 0.93333 1" mesh="left_MCP_VL_visual" />
104
- <site name="left_index_mcp" pos="0 0 0" />
105
- <body name="left_index_PP" quat="0.707105 0.707108 0 0">
106
- <inertial pos="0.0243772 -1.00263e-06 0.000755901" quat="0.519877 0.479286 0.479204 0.519977" mass="0.04138" diaginertia="6.82371e-06 6.54097e-06 2.16753e-06" />
107
- <joint name="left_index_MCP_AA" pos="0 0 0" axis="0 0 1" range="-0.3491 0.3491" />
108
- <geom name="left_index_PP_visual" type="mesh" group="1" density="0" rgba="0.79216 0.81961 0.93333 1" mesh="left_PP_visual" />
109
- <geom name="collision_hand_left_index_2" size="0.01 0.0235" pos="0.0235 0 0" quat="0.707107 0 0.707107 0" type="capsule" group="3" rgba="0 1 0 1" />
110
- <body name="left_index_MP" pos="0.047 0 0" quat="0.707105 -0.707108 0 0">
111
- <inertial pos="0.015316 -0.00138561 6.27987e-08" quat="-0.0196259 0.706839 0.0195186 0.706832" mass="0.02509" diaginertia="2.28516e-06 2.26758e-06 9.99868e-07" />
112
- <joint name="left_index_PIP" pos="0 0 0" axis="0 0 1" range="0 1.7453" />
113
- <geom name="left_index_MP_visual" type="mesh" group="1" density="0" rgba="0.79216 0.81961 0.93333 1" mesh="left_MP_visual" />
114
- <geom name="collision_hand_left_index_1" size="0.009 0.01575" pos="0.01575 0 0" quat="0.707107 0 0.707107 0" type="capsule" group="3" rgba="0 1 0 1" />
115
- <site name="left_index_pip" pos="0 0 0" />
116
- <body name="left_index_DP" pos="0.0315 0 0">
117
- <inertial pos="0.00931302 -0.00769378 0.000111984" quat="0.643869 0.245527 -0.649442 0.321517" mass="0.005102" diaginertia="4.23794e-07 3.8546e-07 1.37255e-07" />
118
- <joint name="left_index_DIP" pos="0 0 0" axis="0 0 1" range="0 1.3963" />
119
- <geom name="left_index_DP_visual" type="mesh" group="1" density="0" rgba="0.79216 0.81961 0.93333 1" mesh="left_DP_visual" />
120
- <geom name="left_index_elastomer_visual" quat="0.5 0.5 -0.5 0.5" type="mesh" group="1" density="0" rgba="0.2 1 0.2 1" mesh="left_elastomer" />
121
- <geom name="collision_hand_left_index_0" size="0.007 0.0100499" pos="0.01 -0.001 0" quat="0.707107 0.0703598 0.703598 0" type="capsule" group="3" rgba="0 1 0 1" />
122
- <site name="left_index_dip" pos="0 0 0" />
123
- <site name="left_index_tip" pos="0.014 0 0" />
124
- <site name="track_hand_left_index_tip" pos="0.014 0 0" />
125
- <site name="trace_hand_left_index_tip" pos="0.014 0 0" />
126
- </body>
127
- </body>
128
- </body>
129
- </body>
130
- <body name="left_middle_MCP_VL" pos="0 -0.01 0.0987" quat="0.499998 -0.500002 -0.5 -0.5">
131
- <inertial pos="4.8e-09 0.00086981 -3.6e-09" mass="1e-06" diaginertia="2e-12 2e-12 2e-12" />
132
- <joint name="left_middle_MCP_FE" pos="0 0 0" axis="0 0 1" range="-0.174533 1.5708" />
133
- <geom name="left_middle_MCP_VL_visual" type="mesh" group="1" density="0" rgba="0.79216 0.81961 0.93333 1" mesh="left_MCP_VL_visual" />
134
- <site name="left_middle_mcp" pos="0 0 0" />
135
- <body name="left_middle_PP" quat="0.707105 0.707108 0 0">
136
- <inertial pos="0.0243772 -1.00263e-06 0.000755901" quat="0.519877 0.479286 0.479204 0.519977" mass="0.04138" diaginertia="6.82371e-06 6.54097e-06 2.16753e-06" />
137
- <joint name="left_middle_MCP_AA" pos="0 0 0" axis="0 0 1" range="-0.3491 0.3491" />
138
- <geom name="left_middle_PP_visual" type="mesh" group="1" density="0" rgba="0.79216 0.81961 0.93333 1" mesh="left_PP_visual" />
139
- <geom name="collision_hand_left_middle_2" size="0.01 0.0235" pos="0.0235 0 0" quat="0.707107 0 0.707107 0" type="capsule" group="3" rgba="0 1 0 1" />
140
- <body name="left_middle_MP" pos="0.047 0 0" quat="0.707105 -0.707108 0 0">
141
- <inertial pos="0.015316 -0.00138561 6.27987e-08" quat="-0.0196259 0.706839 0.0195186 0.706832" mass="0.02509" diaginertia="2.28516e-06 2.26758e-06 9.99868e-07" />
142
- <joint name="left_middle_PIP" pos="0 0 0" axis="0 0 1" range="0 1.7453" />
143
- <geom name="left_middle_MP_visual" type="mesh" group="1" density="0" rgba="0.79216 0.81961 0.93333 1" mesh="left_MP_visual" />
144
- <geom name="collision_hand_left_middle_1" size="0.009 0.01575" pos="0.01575 0 0" quat="0.707107 0 0.707107 0" type="capsule" group="3" rgba="0 1 0 1" />
145
- <site name="left_middle_pip" pos="0 0 0" />
146
- <body name="left_middle_DP" pos="0.0315 0 0">
147
- <inertial pos="0.00931302 -0.00769378 0.000111984" quat="0.643869 0.245527 -0.649442 0.321517" mass="0.005102" diaginertia="4.23794e-07 3.8546e-07 1.37255e-07" />
148
- <joint name="left_middle_DIP" pos="0 0 0" axis="0 0 1" range="0 1.3963" />
149
- <geom name="left_middle_DP_visual" type="mesh" group="1" density="0" rgba="0.79216 0.81961 0.93333 1" mesh="left_DP_visual" />
150
- <geom name="left_middle_elastomer_visual" quat="0.5 0.5 -0.5 0.5" type="mesh" group="1" density="0" rgba="0.2 1 0.2 1" mesh="left_elastomer" />
151
- <geom name="collision_hand_left_middle_0" size="0.007 0.0100499" pos="0.01 -0.001 0" quat="0.707107 0.0703598 0.703598 0" type="capsule" group="3" rgba="0 1 0 1" />
152
- <site name="left_middle_dip" pos="0 0 0" />
153
- <site name="left_middle_tip" pos="0.014 0 0" />
154
- <site name="track_hand_left_middle_tip" pos="0.014 0 0" />
155
- <site name="trace_hand_left_middle_tip" pos="0.014 0 0" />
156
- </body>
157
- </body>
158
- </body>
159
- </body>
160
- <body name="left_ring_MCP_VL" pos="0.0015 0.0103 0.0927" quat="0.499998 -0.500002 -0.5 -0.5">
161
- <inertial pos="4.8e-09 0.00086981 3.6e-09" mass="1e-06" diaginertia="2e-12 2e-12 2e-12" />
162
- <joint name="left_ring_MCP_FE" pos="0 0 0" axis="0 0 1" range="-0.174533 1.5708" />
163
- <geom name="left_ring_MCP_VL_visual" type="mesh" group="1" density="0" rgba="0.79216 0.81961 0.93333 1" mesh="left_MCP_VL_visual" />
164
- <site name="left_ring_mcp" pos="0 0 0" />
165
- <body name="left_ring_PP" quat="0.707105 0.707108 0 0">
166
- <inertial pos="0.0243772 -1.00263e-06 0.000755901" quat="0.519877 0.479286 0.479204 0.519977" mass="0.04138" diaginertia="6.82371e-06 6.54097e-06 2.16753e-06" />
167
- <joint name="left_ring_MCP_AA" pos="0 0 0" axis="0 0 1" range="-0.3491 0.3491" />
168
- <geom name="left_ring_PP_visual" type="mesh" group="1" density="0" rgba="0.79216 0.81961 0.93333 1" mesh="left_PP_visual" />
169
- <geom name="collision_hand_left_ring_2" size="0.01 0.0235" pos="0.0235 0 0" quat="0.707107 0 0.707107 0" type="capsule" group="3" rgba="0 1 0 1" />
170
- <body name="left_ring_MP" pos="0.047 0 0" quat="0.707105 -0.707108 0 0">
171
- <inertial pos="0.015316 -0.00138561 6.27987e-08" quat="-0.0196259 0.706839 0.0195186 0.706832" mass="0.02509" diaginertia="2.28516e-06 2.26758e-06 9.99868e-07" />
172
- <joint name="left_ring_PIP" pos="0 0 0" axis="0 0 1" range="0 1.7453" />
173
- <geom name="left_ring_MP_visual" type="mesh" group="1" density="0" rgba="0.79216 0.81961 0.93333 1" mesh="left_MP_visual" />
174
- <geom name="collision_hand_left_ring_1" size="0.009 0.01575" pos="0.01575 0 0" quat="0.707107 0 0.707107 0" type="capsule" group="3" rgba="0 1 0 1" />
175
- <site name="left_ring_pip" pos="0 0 0" />
176
- <body name="left_ring_DP" pos="0.0315 0 0">
177
- <inertial pos="0.00931208 -0.00769378 0.000111984" quat="0.643953 0.245771 -0.649317 0.321417" mass="0.005102" diaginertia="4.23686e-07 3.85387e-07 1.37226e-07" />
178
- <joint name="left_ring_DIP" pos="0 0 0" axis="0 0 1" range="0 1.3963" />
179
- <geom name="left_ring_DP_visual" type="mesh" group="1" density="0" rgba="0.79216 0.81961 0.93333 1" mesh="left_DP_visual" />
180
- <geom name="left_ring_elastomer_visual" quat="0.499998 0.5 -0.5 0.500002" type="mesh" group="1" density="0" rgba="0.2 1 0.2 1" mesh="left_elastomer" />
181
- <geom name="collision_hand_left_ring_0" size="0.007 0.0100499" pos="0.01 -0.001 0" quat="0.707107 0.0703598 0.703598 0" type="capsule" group="3" rgba="0 1 0 1" />
182
- <site name="left_ring_dip" pos="0 0 0" />
183
- <site name="left_ring_tip" pos="0.014 0 0" />
184
- <site name="track_hand_left_ring_tip" pos="0.014 0 0" />
185
- <site name="trace_hand_left_ring_tip" pos="0.014 0 0" />
186
- </body>
187
- </body>
188
- </body>
189
- </body>
190
- <body name="left_pinky_MC" pos="0.01136 0.0263 0.0867" quat="1.34924e-11 3.67321e-06 -1 -3.67321e-06">
191
- <inertial pos="0.0107985 0.00598801 0.0335125" quat="0.709657 0.106152 0.0812795 0.691746" mass="0.12541" diaginertia="3.57685e-05 2.46485e-05 2.14424e-05" />
192
- <joint name="left_pinky_CMC" pos="0 0 0" axis="0 0 1" range="0 0.2618" />
193
- <geom name="left_pinky_MC_visual" type="mesh" group="1" density="0" rgba="0.79216 0.81961 0.93333 1" mesh="left_pinky_MC_visual" />
194
- <body name="left_pinky_MCP_VL" pos="0.00836 0.0048 0" quat="0.499998 -0.5 0.5 0.500002">
195
- <inertial pos="4.8e-09 0.00086981 -3.6e-09" mass="0.001" diaginertia="2e-12 2e-12 2e-12" />
196
- <joint name="left_pinky_MCP_FE" pos="0 0 0" axis="0 0 1" range="-0.174533 1.5708" />
197
- <geom name="left_pinky_MCP_VL_visual" type="mesh" group="1" density="0" rgba="0.79216 0.81961 0.93333 1" mesh="left_MCP_VL_visual" />
198
- <site name="left_pinky_mcp" pos="0 0 0" />
199
- <body name="left_pinky_PP" quat="0.707105 0.707108 0 0">
200
- <inertial pos="0.0243772 -1.00263e-06 0.000755901" quat="0.519877 0.479286 0.479204 0.519977" mass="0.04138" diaginertia="6.82371e-06 6.54097e-06 2.16753e-06" />
201
- <joint name="left_pinky_MCP_AA" pos="0 0 0" axis="0 0 1" range="-0.3491 0.3491" />
202
- <geom name="left_pinky_PP_visual" type="mesh" group="1" density="0" rgba="0.79216 0.81961 0.93333 1" mesh="left_PP_visual" />
203
- <geom name="collision_hand_left_pinky_2" size="0.009 0.0235" pos="0.0235 0 0" quat="0.707107 0 0.707107 0" type="capsule" group="3" rgba="0 1 0 1" />
204
- <body name="left_pinky_MP" pos="0.047 0 0" quat="0.707105 -0.707108 0 0">
205
- <inertial pos="0.015316 -0.00138561 6.27987e-08" quat="-0.0196259 0.706839 0.0195186 0.706832" mass="0.02509" diaginertia="2.28516e-06 2.26758e-06 9.99868e-07" />
206
- <joint name="left_pinky_PIP" pos="0 0 0" axis="0 0 1" range="0 1.7453" />
207
- <geom name="left_pinky_MP_visual" type="mesh" group="1" density="0" rgba="0.79216 0.81961 0.93333 1" mesh="left_MP_visual" />
208
- <geom name="collision_hand_left_pinky_1" size="0.008 0.01575" pos="0.01575 0 0" quat="0.707107 0 0.707107 0" type="capsule" group="3" rgba="0 1 0 1" />
209
- <site name="left_pinky_pip" pos="0 0 0" />
210
- <body name="left_pinky_DP" pos="0.0315 0 0">
211
- <inertial pos="0.00931208 -0.00769378 0.000111984" quat="0.643953 0.245771 -0.649317 0.321417" mass="0.005102" diaginertia="4.23686e-07 3.85387e-07 1.37226e-07" />
212
- <joint name="left_pinky_DIP" pos="0 0 0" axis="0 0 1" range="0 1.3963" />
213
- <geom name="left_pinky_DP_visual" type="mesh" group="1" density="0" rgba="0.79216 0.81961 0.93333 1" mesh="left_DP_visual" />
214
- <geom name="left_pinky_elastomer_visual" quat="0.5 0.5 -0.5 0.5" type="mesh" group="1" density="0" rgba="0.2 1 0.2 1" mesh="left_elastomer" />
215
- <geom name="collision_hand_left_pinky_0" size="0.007 0.0100499" pos="0.01 -0.001 0" quat="0.707107 0.0703598 0.703598 0" type="capsule" group="3" rgba="0 1 0 1" />
216
- <site name="left_pinky_dip" pos="0 0 0" />
217
- <site name="left_pinky_tip" pos="0.014 0 0" />
218
- <site name="track_hand_left_pinky_tip" pos="0.014 0 0" />
219
- <site name="trace_hand_left_pinky_tip" pos="0.014 0 0" />
220
- </body>
221
- </body>
222
- </body>
223
- </body>
224
- </body>
225
- </body>
226
- </body>
227
- </body>
228
- </body>
229
- </body>
230
- </body>
231
- </body>
232
- <body name="left_object" gravcomp="1">
233
- <inertial pos="0 0 0" mass="0.001" diaginertia="1e-05 1e-05 1e-05" />
234
- <joint name="left_object_joint" type="free" armature="0.0001" frictionloss="0.0001" />
235
- <site name="left_object" pos="0 0 0" size="0.001" rgba="0 0 0 0" />
236
- <site name="trace_left_object" pos="0 0 0" size="0.001" group="4" rgba="0 0 0 0" />
237
- <site name="track_object_left_thumb_tip" pos="0 0 0" size="0.001" group="4" rgba="0 0 0 0" />
238
- <site name="track_object_left_index_tip" pos="0 0 0" size="0.001" group="4" rgba="0 0 0 0" />
239
- <site name="track_object_left_middle_tip" pos="0 0 0" size="0.001" group="4" rgba="0 0 0 0" />
240
- <site name="track_object_left_ring_tip" pos="0 0 0" size="0.001" group="4" rgba="0 0 0 0" />
241
- <site name="track_object_left_pinky_tip" pos="0 0 0" size="0.001" group="4" rgba="0 0 0 0" />
242
- </body>
243
- <body name="ref_hand_left_thumb_tip" mocap="true">
244
- <site name="ref_hand_left_thumb_tip" pos="0 0 0" size="0.001" group="4" rgba="0 0 0 0" />
245
- </body>
246
- <body name="ref_object_left_thumb_tip" mocap="true">
247
- <site name="ref_object_left_thumb_tip" pos="0 0 0" size="0.001" group="4" rgba="0 0 0 0" />
248
- </body>
249
- <body name="ref_hand_left_index_tip" mocap="true">
250
- <site name="ref_hand_left_index_tip" pos="0 0 0" size="0.001" group="4" rgba="0 0 0 0" />
251
- </body>
252
- <body name="ref_object_left_index_tip" mocap="true">
253
- <site name="ref_object_left_index_tip" pos="0 0 0" size="0.001" group="4" rgba="0 0 0 0" />
254
- </body>
255
- <body name="ref_hand_left_middle_tip" mocap="true">
256
- <site name="ref_hand_left_middle_tip" pos="0 0 0" size="0.001" group="4" rgba="0 0 0 0" />
257
- </body>
258
- <body name="ref_object_left_middle_tip" mocap="true">
259
- <site name="ref_object_left_middle_tip" pos="0 0 0" size="0.001" group="4" rgba="0 0 0 0" />
260
- </body>
261
- <body name="ref_hand_left_ring_tip" mocap="true">
262
- <site name="ref_hand_left_ring_tip" pos="0 0 0" size="0.001" group="4" rgba="0 0 0 0" />
263
- </body>
264
- <body name="ref_object_left_ring_tip" mocap="true">
265
- <site name="ref_object_left_ring_tip" pos="0 0 0" size="0.001" group="4" rgba="0 0 0 0" />
266
- </body>
267
- <body name="ref_hand_left_pinky_tip" mocap="true">
268
- <site name="ref_hand_left_pinky_tip" pos="0 0 0" size="0.001" group="4" rgba="0 0 0 0" />
269
- </body>
270
- <body name="ref_object_left_pinky_tip" mocap="true">
271
- <site name="ref_object_left_pinky_tip" pos="0 0 0" size="0.001" group="4" rgba="0 0 0 0" />
272
- </body>
273
- </worldbody>
274
- <actuator>
275
- <general name="left_pos_x_position" joint="left_pos_x" ctrlrange="-5 5" gainprm="1000" biasprm="0 -1000" />
276
- <general name="left_pos_y_position" joint="left_pos_y" ctrlrange="-5 5" gainprm="1000" biasprm="0 -1000" />
277
- <general name="left_pos_z_position" joint="left_pos_z" ctrlrange="-5 5" gainprm="1000" biasprm="0 -1000" />
278
- <general name="left_rot_x_position" joint="left_rot_x" ctrlrange="-6.28 6.28" gainprm="1000" biasprm="0 -1000" />
279
- <general name="left_rot_y_position" joint="left_rot_y" ctrlrange="-6.28 6.28" gainprm="1000" biasprm="0 -1000" />
280
- <general name="left_rot_z_position" joint="left_rot_z" ctrlrange="-6.28 6.28" gainprm="1000" biasprm="0 -1000" />
281
- <general name="left_thumb_CMC_FE_position" joint="left_thumb_CMC_FE" ctrlrange="-0.1745 1.9199" />
282
- <general name="left_thumb_CMC_AA_position" joint="left_thumb_CMC_AA" ctrlrange="-0.3491 0.3491" />
283
- <general name="left_thumb_MCP_FE_position" joint="left_thumb_MCP_FE" ctrlrange="-0.5236 1.3963" />
284
- <general name="left_thumb_MCP_AA_position" joint="left_thumb_MCP_AA" ctrlrange="-0.3491 0.3491" />
285
- <general name="left_thumb_IP_position" joint="left_thumb_IP" ctrlrange="0 1.7453" />
286
- <general name="left_index_MCP_FE_position" joint="left_index_MCP_FE" ctrlrange="-0.174533 1.5708" />
287
- <general name="left_index_MCP_AA_position" joint="left_index_MCP_AA" ctrlrange="-0.3491 0.3491" />
288
- <general name="left_index_PIP_position" joint="left_index_PIP" ctrlrange="0 1.7453" />
289
- <general name="left_index_DIP_position" joint="left_index_DIP" ctrlrange="0 1.3963" />
290
- <general name="left_middle_MCP_FE_position" joint="left_middle_MCP_FE" ctrlrange="-0.174533 1.5708" />
291
- <general name="left_middle_MCP_AA_position" joint="left_middle_MCP_AA" ctrlrange="-0.3491 0.3491" />
292
- <general name="left_middle_PIP_position" joint="left_middle_PIP" ctrlrange="0 1.7453" />
293
- <general name="left_middle_DIP_position" joint="left_middle_DIP" ctrlrange="0 1.3963" />
294
- <general name="left_ring_MCP_FE_position" joint="left_ring_MCP_FE" ctrlrange="-0.174533 1.5708" />
295
- <general name="left_ring_MCP_AA_position" joint="left_ring_MCP_AA" ctrlrange="-0.3491 0.3491" />
296
- <general name="left_ring_PIP_position" joint="left_ring_PIP" ctrlrange="0 1.7453" />
297
- <general name="left_ring_DIP_position" joint="left_ring_DIP" ctrlrange="0 1.3963" />
298
- <general name="left_pinky_CMC_position" joint="left_pinky_CMC" ctrlrange="0 0.2618" />
299
- <general name="left_pinky_MCP_FE_position" joint="left_pinky_MCP_FE" ctrlrange="-0.174533 1.5708" />
300
- <general name="left_pinky_MCP_AA_position" joint="left_pinky_MCP_AA" ctrlrange="-0.3491 0.3491" />
301
- <general name="left_pinky_PIP_position" joint="left_pinky_PIP" ctrlrange="0 1.7453" />
302
- <general name="left_pinky_DIP_position" joint="left_pinky_DIP" ctrlrange="0 1.3963" />
303
- </actuator>
304
- </mujoco>
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/task_info.json DELETED
@@ -1,19 +0,0 @@
1
- {
2
- "task": "fold_clothes_left",
3
- "dataset_name": "egoverse_hand",
4
- "robot_type": "sharpa",
5
- "embodiment_type": "left",
6
- "data_id": 0,
7
- "ref_dt": 0.03333333333333333,
8
- "hand_only": true,
9
- "right_object_mesh_dir": null,
10
- "left_object_mesh_dir": null,
11
- "right_object_convex_dir": null,
12
- "left_object_convex_dir": null,
13
- "floor_height": -0.001,
14
- "hand_floor_collision": false,
15
- "object_floor_collision": false,
16
- "object_object_collision": false,
17
- "scene_builder": "examples.egoverse.hand_scene",
18
- "object_placeholder": "inertial_only; zero geometries; gravity compensated"
19
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/evidence/control_replay.json DELETED
@@ -1,76 +0,0 @@
1
- {
2
- "backend": "CPU MuJoCo 3.7 independent integration of native saved controls",
3
- "steps": 400,
4
- "final_sim_time": 3.9999999999999587,
5
- "no_state_overrides_after_initialization": true,
6
- "control_limits_enabled": true,
7
- "control_sha256": "cbe8e1502867315f187f62aa9379d213cc0e661c76cf56a5a16d7e3c0092642e",
8
- "hand_control_mode": "inverse_dynamics",
9
- "hand_gain_scale": 10,
10
- "effective_kp": [
11
- 10000.0,
12
- 10000.0,
13
- 10000.0,
14
- 10000.0,
15
- 10000.0,
16
- 10000.0,
17
- 3000.0,
18
- 3000.0,
19
- 3000.0,
20
- 3000.0,
21
- 3000.0,
22
- 3000.0,
23
- 3000.0,
24
- 3000.0,
25
- 3000.0,
26
- 3000.0,
27
- 3000.0,
28
- 3000.0,
29
- 3000.0,
30
- 3000.0,
31
- 3000.0,
32
- 3000.0,
33
- 3000.0,
34
- 3000.0,
35
- 3000.0,
36
- 3000.0,
37
- 3000.0,
38
- 3000.0
39
- ],
40
- "effective_kd": [
41
- 341.03882476926293,
42
- 335.1230818669464,
43
- 329.1010179261073,
44
- 203.90970903560145,
45
- 202.82110335833255,
46
- 201.15403259775766,
47
- 109.56926675998416,
48
- 109.58289286453568,
49
- 109.5470774114592,
50
- 109.54699912919904,
51
- 109.54463577048301,
52
- 109.55389195566207,
53
- 109.55384384269877,
54
- 109.54546685611147,
55
- 109.54457545025748,
56
- 109.55389191426805,
57
- 109.55384384269877,
58
- 109.54546685611147,
59
- 109.54457545025748,
60
- 109.55389186239185,
61
- 109.553843790843,
62
- 109.54546682892311,
63
- 109.54457543961777,
64
- 109.54742206992059,
65
- 109.55389190378588,
66
- 109.553843790843,
67
- 109.54546682892311,
68
- 109.54457543961777
69
- ],
70
- "max_abs_qpos_difference_vs_native": 4.37370158756778e-07,
71
- "max_hand_site_difference_vs_native_m": 7.981460739595656e-08,
72
- "replay_success": true,
73
- "full_source_covered": true,
74
- "max_site_error_m": 0.01496272811646004,
75
- "max_palm_rotation_error_rad": 0.0058743355631741614
76
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/evidence/evaluation_control_replay.json DELETED
@@ -1,116 +0,0 @@
1
- {
2
- "schema_version": 1,
3
- "status": "hand-only",
4
- "outcome": "success",
5
- "tracking_domain": "hand",
6
- "success": true,
7
- "manipulation_success": null,
8
- "failure_reasons": [],
9
- "coverage": {
10
- "source_frames": 120,
11
- "source_duration_seconds": 3.966666666666667,
12
- "source_coverage_fraction": 1.0,
13
- "full_source_covered": true,
14
- "source_coverage_passed": true,
15
- "minimum_source_coverage_fraction": 0.99,
16
- "saved_frames": 400,
17
- "evaluated_frames": 397,
18
- "reference_sim_steps": 397,
19
- "last_scored_source_time_seconds": 3.9699999999999593,
20
- "post_reference_hold_steps": 4,
21
- "scored_endpoint_hold_steps": 1,
22
- "excluded_end_padding_frames": 3,
23
- "warmup_seconds": 0.0,
24
- "warmup_saved_frames": 0,
25
- "saved_source_start_seconds": 0.01,
26
- "saved_source_end_seconds": 3.9999999999999587,
27
- "covered_source_start_seconds": 0.0,
28
- "covered_source_end_seconds": 3.966666666666667,
29
- "ref_dt": 0.03333333333333333,
30
- "sim_dt": 0.01,
31
- "definition": "Contiguous measured simulation intervals intersecting [0, last source timestamp]; warmup and post-source padding give no coverage."
32
- },
33
- "thresholds": {
34
- "position_m": 0.02,
35
- "rotation_rad": 0.3,
36
- "hand_position_m": 0.02,
37
- "hand_rotation_rad": 0.2,
38
- "minimum_motion_ratio": 0.8,
39
- "maximum_motion_ratio": 1.2,
40
- "meaningful_translation_m": 0.005,
41
- "meaningful_rotation_rad": 0.03
42
- },
43
- "threshold_definitions": {
44
- "position_m": "Maximum absolute world-frame object-body position error over every saved source-time sample.",
45
- "rotation_rad": "Maximum sign-invariant geodesic object orientation error in radians.",
46
- "hand_position_m": "Maximum world-frame palm/fingertip FK position error against the IK reference.",
47
- "hand_rotation_rad": "Maximum palm geodesic orientation error in radians against the IK reference.",
48
- "motion": "Excursion, positive world-Z lift and rotation relative to the first pose; ratios use the full source. Aligned displacement gain is sum(actual_delta dot reference_delta)/sum(reference_delta squared). Gates apply above the recorded meaningful-motion floors.",
49
- "success": "At least 99% of the full source duration from the start, valid controls, and every applicable object/hand/motion check passes. full_source_covered separately records exact endpoint coverage. Hand-only never establishes manipulation success."
50
- },
51
- "validation": {
52
- "nq": 35,
53
- "nv": 34,
54
- "nu": 28,
55
- "finite_arrays": true,
56
- "contiguous_timestamps": true,
57
- "control_range_violations": 0,
58
- "controls_replayed": false
59
- },
60
- "per_object": {},
61
- "per_hand": {
62
- "left": {
63
- "sites": [
64
- "left_palm",
65
- "left_thumb_tip",
66
- "left_index_tip",
67
- "left_middle_tip",
68
- "left_ring_tip",
69
- "left_pinky_tip"
70
- ],
71
- "site_position_error_m": {
72
- "mean": 0.001193288708427611,
73
- "p95": 0.003123275352742065,
74
- "max": 0.01496272811646004
75
- },
76
- "palm_orientation_error_rad": {
77
- "mean": 0.0014171518664664428,
78
- "p95": 0.00261039975359466,
79
- "max": 0.0058743355631741614
80
- },
81
- "within_tracking_thresholds": true,
82
- "motion": {
83
- "reference_site_excursion_m": 0.33187779581603055,
84
- "actual_site_excursion_m": 0.33163671399690736,
85
- "excursion_ratio": 0.9992735825591151,
86
- "aligned_displacement_gain": 0.9987219841390507,
87
- "within_motion_thresholds": true
88
- }
89
- }
90
- },
91
- "provenance": {
92
- "inputs": {
93
- "scene": {
94
- "path": "/local/home/zhenyuj/spider-data-pipeline-verification/example_datasets/full_retarget_campaign/runs/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/scene.xml",
95
- "sha256": "1219685ac913e7954a0785a2befba6ecb40eb0c30024192ec7c64c9f2dfd50b0"
96
- },
97
- "trajectory": {
98
- "path": "/local/home/zhenyuj/spider-data-pipeline-verification/example_datasets/full_retarget_campaign/runs/egoverse_trial2/evidence/trajectory_control_replay.npz",
99
- "sha256": "158424c01bbccfff10b46f53b22190bd21c26583808c3966169cf90beedb25b5"
100
- },
101
- "reference": {
102
- "path": "/local/home/zhenyuj/spider-data-pipeline-verification/example_datasets/full_retarget_campaign/runs/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/0/trajectory_kinematic.npz",
103
- "sha256": "fb8aafe61241e8efc9866e9c82ecb0306394c0cc497f3a0b6ffbb3d7b3c9c20a"
104
- }
105
- },
106
- "reference_timestep_source": "frequency",
107
- "declared_reference_timestep": 0.03333333333333333,
108
- "trajectory_timestep_source": "explicit/metadata",
109
- "trajectory_time_source": "saved time",
110
- "recorded_run_status": null,
111
- "warning": "Saved poses and controls are evaluated, not independently re-simulated."
112
- },
113
- "rendering": {
114
- "status": "not_requested"
115
- }
116
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/evidence/native/comparison.mp4 DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:e945e7dcdf5589380321f2335054b8dc907794f5013ef99e9fbd06ac9543b4db
3
- size 1583139
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/evidence/native/evaluation.json DELETED
@@ -1,971 +0,0 @@
1
- {
2
- "schema_version": 1,
3
- "status": "hand-only",
4
- "outcome": "success",
5
- "tracking_domain": "hand",
6
- "success": true,
7
- "manipulation_success": null,
8
- "failure_reasons": [],
9
- "coverage": {
10
- "source_frames": 120,
11
- "source_duration_seconds": 3.966666666666667,
12
- "source_coverage_fraction": 1.0,
13
- "full_source_covered": true,
14
- "source_coverage_passed": true,
15
- "minimum_source_coverage_fraction": 0.99,
16
- "saved_frames": 400,
17
- "evaluated_frames": 397,
18
- "reference_sim_steps": 397,
19
- "last_scored_source_time_seconds": 3.97,
20
- "post_reference_hold_steps": 4,
21
- "scored_endpoint_hold_steps": 1,
22
- "excluded_end_padding_frames": 3,
23
- "warmup_seconds": 0.0,
24
- "warmup_saved_frames": 0,
25
- "saved_source_start_seconds": 0.01,
26
- "saved_source_end_seconds": 4.0,
27
- "covered_source_start_seconds": 0.0,
28
- "covered_source_end_seconds": 3.966666666666667,
29
- "ref_dt": 0.03333333333333333,
30
- "sim_dt": 0.01,
31
- "definition": "Contiguous measured simulation intervals intersecting [0, last source timestamp]; warmup and post-source padding give no coverage."
32
- },
33
- "thresholds": {
34
- "position_m": 0.02,
35
- "rotation_rad": 0.3,
36
- "hand_position_m": 0.02,
37
- "hand_rotation_rad": 0.2,
38
- "minimum_motion_ratio": 0.8,
39
- "maximum_motion_ratio": 1.2,
40
- "meaningful_translation_m": 0.005,
41
- "meaningful_rotation_rad": 0.03
42
- },
43
- "threshold_definitions": {
44
- "position_m": "Maximum absolute world-frame object-body position error over every saved source-time sample.",
45
- "rotation_rad": "Maximum sign-invariant geodesic object orientation error in radians.",
46
- "hand_position_m": "Maximum world-frame palm/fingertip FK position error against the IK reference.",
47
- "hand_rotation_rad": "Maximum palm geodesic orientation error in radians against the IK reference.",
48
- "motion": "Excursion, positive world-Z lift and rotation relative to the first pose; ratios use the full source. Aligned displacement gain is sum(actual_delta dot reference_delta)/sum(reference_delta squared). Gates apply above the recorded meaningful-motion floors.",
49
- "success": "At least 99% of the full source duration from the start, valid controls, and every applicable object/hand/motion check passes. full_source_covered separately records exact endpoint coverage. Hand-only never establishes manipulation success."
50
- },
51
- "validation": {
52
- "nq": 35,
53
- "nv": 34,
54
- "nu": 28,
55
- "finite_arrays": true,
56
- "contiguous_timestamps": true,
57
- "control_range_violations": 0,
58
- "controls_replayed": false
59
- },
60
- "per_object": {},
61
- "per_hand": {
62
- "left": {
63
- "sites": [
64
- "left_palm",
65
- "left_thumb_tip",
66
- "left_index_tip",
67
- "left_middle_tip",
68
- "left_ring_tip",
69
- "left_pinky_tip"
70
- ],
71
- "site_position_error_m": {
72
- "mean": 0.0011932878521424845,
73
- "p95": 0.0031232839145199608,
74
- "max": 0.014962734140670902
75
- },
76
- "palm_orientation_error_rad": {
77
- "mean": 0.0014171413975547088,
78
- "p95": 0.002610409718036009,
79
- "max": 0.00587454305089409
80
- },
81
- "within_tracking_thresholds": true,
82
- "motion": {
83
- "reference_site_excursion_m": 0.33187779581603055,
84
- "actual_site_excursion_m": 0.33163671093011066,
85
- "excursion_ratio": 0.9992735733183743,
86
- "aligned_displacement_gain": 0.9987218943804519,
87
- "within_motion_thresholds": true
88
- }
89
- }
90
- },
91
- "provenance": {
92
- "inputs": {
93
- "scene": {
94
- "path": "/local/home/zhenyuj/spider-data-pipeline-verification/example_datasets/full_retarget_campaign/runs/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/scene.xml",
95
- "sha256": "1219685ac913e7954a0785a2befba6ecb40eb0c30024192ec7c64c9f2dfd50b0"
96
- },
97
- "trajectory": {
98
- "path": "/local/home/zhenyuj/spider-data-pipeline-verification/example_datasets/full_retarget_campaign/runs/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/0/trajectory_mjwp_fast.npz",
99
- "sha256": "d8c6d098f4ee03f7f05e876222098a3cd855874cab3e3a0ee43c00bd72466695"
100
- },
101
- "reference": {
102
- "path": "/local/home/zhenyuj/spider-data-pipeline-verification/example_datasets/full_retarget_campaign/runs/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/0/trajectory_kinematic.npz",
103
- "sha256": "fb8aafe61241e8efc9866e9c82ecb0306394c0cc497f3a0b6ffbb3d7b3c9c20a"
104
- }
105
- },
106
- "reference_timestep_source": "frequency",
107
- "declared_reference_timestep": 0.03333333333333333,
108
- "trajectory_timestep_source": "explicit/metadata",
109
- "trajectory_time_source": "saved time",
110
- "recorded_run_status": "succeeded",
111
- "warning": "Saved poses and controls are evaluated, not independently re-simulated."
112
- },
113
- "rendering": {
114
- "status": "succeeded",
115
- "fps": 30,
116
- "frames": 120,
117
- "video": "comparison.mp4",
118
- "thumbnail": "thumbnail.jpg",
119
- "video_sha256": "e945e7dcdf5589380321f2335054b8dc907794f5013ef99e9fbd06ac9543b4db",
120
- "thumbnail_sha256": "cafeba69360c2cd179fb7b80d29bd3108e5561124da1407edde7cc75a673eb87",
121
- "encoded_duration_seconds": 4.0,
122
- "last_reference_time_seconds": 3.966666666666667,
123
- "unavailable_actual_video_frames": 1,
124
- "actual_sampling": "Nearest measured saved pose to the 30 Hz video clock. The reference is rendered at that same saved time (clamped only for the endpoint boundary step). Missing intervals stay blank; no actual-pose interpolation, extension or post-rollout freezing.",
125
- "camera": "Shared camera follows reference geometry only; actual tracking drift stays visible.",
126
- "source_video_alignment": null,
127
- "frame_timeline": [
128
- {
129
- "video_frame": 0,
130
- "video_time_seconds": 0.0,
131
- "reference_time_seconds": 0.0,
132
- "saved_frame": null
133
- },
134
- {
135
- "video_frame": 1,
136
- "video_time_seconds": 0.03333333333333333,
137
- "reference_time_seconds": 0.03,
138
- "saved_frame": 2,
139
- "saved_source_time_seconds": 0.03
140
- },
141
- {
142
- "video_frame": 2,
143
- "video_time_seconds": 0.06666666666666667,
144
- "reference_time_seconds": 0.07,
145
- "saved_frame": 6,
146
- "saved_source_time_seconds": 0.07
147
- },
148
- {
149
- "video_frame": 3,
150
- "video_time_seconds": 0.1,
151
- "reference_time_seconds": 0.1,
152
- "saved_frame": 9,
153
- "saved_source_time_seconds": 0.1
154
- },
155
- {
156
- "video_frame": 4,
157
- "video_time_seconds": 0.13333333333333333,
158
- "reference_time_seconds": 0.13,
159
- "saved_frame": 12,
160
- "saved_source_time_seconds": 0.13
161
- },
162
- {
163
- "video_frame": 5,
164
- "video_time_seconds": 0.16666666666666666,
165
- "reference_time_seconds": 0.17,
166
- "saved_frame": 16,
167
- "saved_source_time_seconds": 0.17
168
- },
169
- {
170
- "video_frame": 6,
171
- "video_time_seconds": 0.2,
172
- "reference_time_seconds": 0.2,
173
- "saved_frame": 19,
174
- "saved_source_time_seconds": 0.2
175
- },
176
- {
177
- "video_frame": 7,
178
- "video_time_seconds": 0.23333333333333334,
179
- "reference_time_seconds": 0.23,
180
- "saved_frame": 22,
181
- "saved_source_time_seconds": 0.23
182
- },
183
- {
184
- "video_frame": 8,
185
- "video_time_seconds": 0.26666666666666666,
186
- "reference_time_seconds": 0.27,
187
- "saved_frame": 26,
188
- "saved_source_time_seconds": 0.27
189
- },
190
- {
191
- "video_frame": 9,
192
- "video_time_seconds": 0.3,
193
- "reference_time_seconds": 0.3,
194
- "saved_frame": 29,
195
- "saved_source_time_seconds": 0.3
196
- },
197
- {
198
- "video_frame": 10,
199
- "video_time_seconds": 0.3333333333333333,
200
- "reference_time_seconds": 0.33,
201
- "saved_frame": 32,
202
- "saved_source_time_seconds": 0.33
203
- },
204
- {
205
- "video_frame": 11,
206
- "video_time_seconds": 0.36666666666666664,
207
- "reference_time_seconds": 0.37,
208
- "saved_frame": 36,
209
- "saved_source_time_seconds": 0.37
210
- },
211
- {
212
- "video_frame": 12,
213
- "video_time_seconds": 0.4,
214
- "reference_time_seconds": 0.4,
215
- "saved_frame": 39,
216
- "saved_source_time_seconds": 0.4
217
- },
218
- {
219
- "video_frame": 13,
220
- "video_time_seconds": 0.43333333333333335,
221
- "reference_time_seconds": 0.43,
222
- "saved_frame": 42,
223
- "saved_source_time_seconds": 0.43
224
- },
225
- {
226
- "video_frame": 14,
227
- "video_time_seconds": 0.4666666666666667,
228
- "reference_time_seconds": 0.47000000000000003,
229
- "saved_frame": 46,
230
- "saved_source_time_seconds": 0.47000000000000003
231
- },
232
- {
233
- "video_frame": 15,
234
- "video_time_seconds": 0.5,
235
- "reference_time_seconds": 0.5,
236
- "saved_frame": 49,
237
- "saved_source_time_seconds": 0.5
238
- },
239
- {
240
- "video_frame": 16,
241
- "video_time_seconds": 0.5333333333333333,
242
- "reference_time_seconds": 0.53,
243
- "saved_frame": 52,
244
- "saved_source_time_seconds": 0.53
245
- },
246
- {
247
- "video_frame": 17,
248
- "video_time_seconds": 0.5666666666666667,
249
- "reference_time_seconds": 0.5700000000000001,
250
- "saved_frame": 56,
251
- "saved_source_time_seconds": 0.5700000000000001
252
- },
253
- {
254
- "video_frame": 18,
255
- "video_time_seconds": 0.6,
256
- "reference_time_seconds": 0.6,
257
- "saved_frame": 59,
258
- "saved_source_time_seconds": 0.6
259
- },
260
- {
261
- "video_frame": 19,
262
- "video_time_seconds": 0.6333333333333333,
263
- "reference_time_seconds": 0.63,
264
- "saved_frame": 62,
265
- "saved_source_time_seconds": 0.63
266
- },
267
- {
268
- "video_frame": 20,
269
- "video_time_seconds": 0.6666666666666666,
270
- "reference_time_seconds": 0.67,
271
- "saved_frame": 66,
272
- "saved_source_time_seconds": 0.67
273
- },
274
- {
275
- "video_frame": 21,
276
- "video_time_seconds": 0.7,
277
- "reference_time_seconds": 0.7000000000000001,
278
- "saved_frame": 69,
279
- "saved_source_time_seconds": 0.7000000000000001
280
- },
281
- {
282
- "video_frame": 22,
283
- "video_time_seconds": 0.7333333333333333,
284
- "reference_time_seconds": 0.73,
285
- "saved_frame": 72,
286
- "saved_source_time_seconds": 0.73
287
- },
288
- {
289
- "video_frame": 23,
290
- "video_time_seconds": 0.7666666666666667,
291
- "reference_time_seconds": 0.77,
292
- "saved_frame": 76,
293
- "saved_source_time_seconds": 0.77
294
- },
295
- {
296
- "video_frame": 24,
297
- "video_time_seconds": 0.8,
298
- "reference_time_seconds": 0.8,
299
- "saved_frame": 79,
300
- "saved_source_time_seconds": 0.8
301
- },
302
- {
303
- "video_frame": 25,
304
- "video_time_seconds": 0.8333333333333334,
305
- "reference_time_seconds": 0.8300000000000001,
306
- "saved_frame": 82,
307
- "saved_source_time_seconds": 0.8300000000000001
308
- },
309
- {
310
- "video_frame": 26,
311
- "video_time_seconds": 0.8666666666666667,
312
- "reference_time_seconds": 0.87,
313
- "saved_frame": 86,
314
- "saved_source_time_seconds": 0.87
315
- },
316
- {
317
- "video_frame": 27,
318
- "video_time_seconds": 0.9,
319
- "reference_time_seconds": 0.9,
320
- "saved_frame": 89,
321
- "saved_source_time_seconds": 0.9
322
- },
323
- {
324
- "video_frame": 28,
325
- "video_time_seconds": 0.9333333333333333,
326
- "reference_time_seconds": 0.93,
327
- "saved_frame": 92,
328
- "saved_source_time_seconds": 0.93
329
- },
330
- {
331
- "video_frame": 29,
332
- "video_time_seconds": 0.9666666666666667,
333
- "reference_time_seconds": 0.97,
334
- "saved_frame": 96,
335
- "saved_source_time_seconds": 0.97
336
- },
337
- {
338
- "video_frame": 30,
339
- "video_time_seconds": 1.0,
340
- "reference_time_seconds": 1.0,
341
- "saved_frame": 99,
342
- "saved_source_time_seconds": 1.0
343
- },
344
- {
345
- "video_frame": 31,
346
- "video_time_seconds": 1.0333333333333334,
347
- "reference_time_seconds": 1.03,
348
- "saved_frame": 102,
349
- "saved_source_time_seconds": 1.03
350
- },
351
- {
352
- "video_frame": 32,
353
- "video_time_seconds": 1.0666666666666667,
354
- "reference_time_seconds": 1.07,
355
- "saved_frame": 106,
356
- "saved_source_time_seconds": 1.07
357
- },
358
- {
359
- "video_frame": 33,
360
- "video_time_seconds": 1.1,
361
- "reference_time_seconds": 1.1,
362
- "saved_frame": 109,
363
- "saved_source_time_seconds": 1.1
364
- },
365
- {
366
- "video_frame": 34,
367
- "video_time_seconds": 1.1333333333333333,
368
- "reference_time_seconds": 1.1300000000000001,
369
- "saved_frame": 112,
370
- "saved_source_time_seconds": 1.1300000000000001
371
- },
372
- {
373
- "video_frame": 35,
374
- "video_time_seconds": 1.1666666666666667,
375
- "reference_time_seconds": 1.17,
376
- "saved_frame": 116,
377
- "saved_source_time_seconds": 1.17
378
- },
379
- {
380
- "video_frame": 36,
381
- "video_time_seconds": 1.2,
382
- "reference_time_seconds": 1.2,
383
- "saved_frame": 119,
384
- "saved_source_time_seconds": 1.2
385
- },
386
- {
387
- "video_frame": 37,
388
- "video_time_seconds": 1.2333333333333334,
389
- "reference_time_seconds": 1.23,
390
- "saved_frame": 122,
391
- "saved_source_time_seconds": 1.23
392
- },
393
- {
394
- "video_frame": 38,
395
- "video_time_seconds": 1.2666666666666666,
396
- "reference_time_seconds": 1.27,
397
- "saved_frame": 126,
398
- "saved_source_time_seconds": 1.27
399
- },
400
- {
401
- "video_frame": 39,
402
- "video_time_seconds": 1.3,
403
- "reference_time_seconds": 1.3,
404
- "saved_frame": 129,
405
- "saved_source_time_seconds": 1.3
406
- },
407
- {
408
- "video_frame": 40,
409
- "video_time_seconds": 1.3333333333333333,
410
- "reference_time_seconds": 1.33,
411
- "saved_frame": 132,
412
- "saved_source_time_seconds": 1.33
413
- },
414
- {
415
- "video_frame": 41,
416
- "video_time_seconds": 1.3666666666666667,
417
- "reference_time_seconds": 1.37,
418
- "saved_frame": 136,
419
- "saved_source_time_seconds": 1.37
420
- },
421
- {
422
- "video_frame": 42,
423
- "video_time_seconds": 1.4,
424
- "reference_time_seconds": 1.4000000000000001,
425
- "saved_frame": 139,
426
- "saved_source_time_seconds": 1.4000000000000001
427
- },
428
- {
429
- "video_frame": 43,
430
- "video_time_seconds": 1.4333333333333333,
431
- "reference_time_seconds": 1.43,
432
- "saved_frame": 142,
433
- "saved_source_time_seconds": 1.43
434
- },
435
- {
436
- "video_frame": 44,
437
- "video_time_seconds": 1.4666666666666666,
438
- "reference_time_seconds": 1.47,
439
- "saved_frame": 146,
440
- "saved_source_time_seconds": 1.47
441
- },
442
- {
443
- "video_frame": 45,
444
- "video_time_seconds": 1.5,
445
- "reference_time_seconds": 1.5,
446
- "saved_frame": 149,
447
- "saved_source_time_seconds": 1.5
448
- },
449
- {
450
- "video_frame": 46,
451
- "video_time_seconds": 1.5333333333333334,
452
- "reference_time_seconds": 1.53,
453
- "saved_frame": 152,
454
- "saved_source_time_seconds": 1.53
455
- },
456
- {
457
- "video_frame": 47,
458
- "video_time_seconds": 1.5666666666666667,
459
- "reference_time_seconds": 1.57,
460
- "saved_frame": 156,
461
- "saved_source_time_seconds": 1.57
462
- },
463
- {
464
- "video_frame": 48,
465
- "video_time_seconds": 1.6,
466
- "reference_time_seconds": 1.6,
467
- "saved_frame": 159,
468
- "saved_source_time_seconds": 1.6
469
- },
470
- {
471
- "video_frame": 49,
472
- "video_time_seconds": 1.6333333333333333,
473
- "reference_time_seconds": 1.6300000000000001,
474
- "saved_frame": 162,
475
- "saved_source_time_seconds": 1.6300000000000001
476
- },
477
- {
478
- "video_frame": 50,
479
- "video_time_seconds": 1.6666666666666667,
480
- "reference_time_seconds": 1.67,
481
- "saved_frame": 166,
482
- "saved_source_time_seconds": 1.67
483
- },
484
- {
485
- "video_frame": 51,
486
- "video_time_seconds": 1.7,
487
- "reference_time_seconds": 1.7,
488
- "saved_frame": 169,
489
- "saved_source_time_seconds": 1.7
490
- },
491
- {
492
- "video_frame": 52,
493
- "video_time_seconds": 1.7333333333333334,
494
- "reference_time_seconds": 1.73,
495
- "saved_frame": 172,
496
- "saved_source_time_seconds": 1.73
497
- },
498
- {
499
- "video_frame": 53,
500
- "video_time_seconds": 1.7666666666666666,
501
- "reference_time_seconds": 1.77,
502
- "saved_frame": 176,
503
- "saved_source_time_seconds": 1.77
504
- },
505
- {
506
- "video_frame": 54,
507
- "video_time_seconds": 1.8,
508
- "reference_time_seconds": 1.8,
509
- "saved_frame": 179,
510
- "saved_source_time_seconds": 1.8
511
- },
512
- {
513
- "video_frame": 55,
514
- "video_time_seconds": 1.8333333333333333,
515
- "reference_time_seconds": 1.83,
516
- "saved_frame": 182,
517
- "saved_source_time_seconds": 1.83
518
- },
519
- {
520
- "video_frame": 56,
521
- "video_time_seconds": 1.8666666666666667,
522
- "reference_time_seconds": 1.87,
523
- "saved_frame": 186,
524
- "saved_source_time_seconds": 1.87
525
- },
526
- {
527
- "video_frame": 57,
528
- "video_time_seconds": 1.9,
529
- "reference_time_seconds": 1.9000000000000001,
530
- "saved_frame": 189,
531
- "saved_source_time_seconds": 1.9000000000000001
532
- },
533
- {
534
- "video_frame": 58,
535
- "video_time_seconds": 1.9333333333333333,
536
- "reference_time_seconds": 1.93,
537
- "saved_frame": 192,
538
- "saved_source_time_seconds": 1.93
539
- },
540
- {
541
- "video_frame": 59,
542
- "video_time_seconds": 1.9666666666666666,
543
- "reference_time_seconds": 1.97,
544
- "saved_frame": 196,
545
- "saved_source_time_seconds": 1.97
546
- },
547
- {
548
- "video_frame": 60,
549
- "video_time_seconds": 2.0,
550
- "reference_time_seconds": 2.0,
551
- "saved_frame": 199,
552
- "saved_source_time_seconds": 2.0
553
- },
554
- {
555
- "video_frame": 61,
556
- "video_time_seconds": 2.033333333333333,
557
- "reference_time_seconds": 2.0300000000000002,
558
- "saved_frame": 202,
559
- "saved_source_time_seconds": 2.0300000000000002
560
- },
561
- {
562
- "video_frame": 62,
563
- "video_time_seconds": 2.066666666666667,
564
- "reference_time_seconds": 2.07,
565
- "saved_frame": 206,
566
- "saved_source_time_seconds": 2.07
567
- },
568
- {
569
- "video_frame": 63,
570
- "video_time_seconds": 2.1,
571
- "reference_time_seconds": 2.1,
572
- "saved_frame": 209,
573
- "saved_source_time_seconds": 2.1
574
- },
575
- {
576
- "video_frame": 64,
577
- "video_time_seconds": 2.1333333333333333,
578
- "reference_time_seconds": 2.13,
579
- "saved_frame": 212,
580
- "saved_source_time_seconds": 2.13
581
- },
582
- {
583
- "video_frame": 65,
584
- "video_time_seconds": 2.1666666666666665,
585
- "reference_time_seconds": 2.17,
586
- "saved_frame": 216,
587
- "saved_source_time_seconds": 2.17
588
- },
589
- {
590
- "video_frame": 66,
591
- "video_time_seconds": 2.2,
592
- "reference_time_seconds": 2.2,
593
- "saved_frame": 219,
594
- "saved_source_time_seconds": 2.2
595
- },
596
- {
597
- "video_frame": 67,
598
- "video_time_seconds": 2.2333333333333334,
599
- "reference_time_seconds": 2.23,
600
- "saved_frame": 222,
601
- "saved_source_time_seconds": 2.23
602
- },
603
- {
604
- "video_frame": 68,
605
- "video_time_seconds": 2.2666666666666666,
606
- "reference_time_seconds": 2.27,
607
- "saved_frame": 226,
608
- "saved_source_time_seconds": 2.27
609
- },
610
- {
611
- "video_frame": 69,
612
- "video_time_seconds": 2.3,
613
- "reference_time_seconds": 2.3000000000000003,
614
- "saved_frame": 229,
615
- "saved_source_time_seconds": 2.3000000000000003
616
- },
617
- {
618
- "video_frame": 70,
619
- "video_time_seconds": 2.3333333333333335,
620
- "reference_time_seconds": 2.33,
621
- "saved_frame": 232,
622
- "saved_source_time_seconds": 2.33
623
- },
624
- {
625
- "video_frame": 71,
626
- "video_time_seconds": 2.3666666666666667,
627
- "reference_time_seconds": 2.37,
628
- "saved_frame": 236,
629
- "saved_source_time_seconds": 2.37
630
- },
631
- {
632
- "video_frame": 72,
633
- "video_time_seconds": 2.4,
634
- "reference_time_seconds": 2.4,
635
- "saved_frame": 239,
636
- "saved_source_time_seconds": 2.4
637
- },
638
- {
639
- "video_frame": 73,
640
- "video_time_seconds": 2.433333333333333,
641
- "reference_time_seconds": 2.43,
642
- "saved_frame": 242,
643
- "saved_source_time_seconds": 2.43
644
- },
645
- {
646
- "video_frame": 74,
647
- "video_time_seconds": 2.466666666666667,
648
- "reference_time_seconds": 2.47,
649
- "saved_frame": 246,
650
- "saved_source_time_seconds": 2.47
651
- },
652
- {
653
- "video_frame": 75,
654
- "video_time_seconds": 2.5,
655
- "reference_time_seconds": 2.5,
656
- "saved_frame": 249,
657
- "saved_source_time_seconds": 2.5
658
- },
659
- {
660
- "video_frame": 76,
661
- "video_time_seconds": 2.533333333333333,
662
- "reference_time_seconds": 2.5300000000000002,
663
- "saved_frame": 252,
664
- "saved_source_time_seconds": 2.5300000000000002
665
- },
666
- {
667
- "video_frame": 77,
668
- "video_time_seconds": 2.566666666666667,
669
- "reference_time_seconds": 2.57,
670
- "saved_frame": 256,
671
- "saved_source_time_seconds": 2.57
672
- },
673
- {
674
- "video_frame": 78,
675
- "video_time_seconds": 2.6,
676
- "reference_time_seconds": 2.6,
677
- "saved_frame": 259,
678
- "saved_source_time_seconds": 2.6
679
- },
680
- {
681
- "video_frame": 79,
682
- "video_time_seconds": 2.6333333333333333,
683
- "reference_time_seconds": 2.63,
684
- "saved_frame": 262,
685
- "saved_source_time_seconds": 2.63
686
- },
687
- {
688
- "video_frame": 80,
689
- "video_time_seconds": 2.6666666666666665,
690
- "reference_time_seconds": 2.67,
691
- "saved_frame": 266,
692
- "saved_source_time_seconds": 2.67
693
- },
694
- {
695
- "video_frame": 81,
696
- "video_time_seconds": 2.7,
697
- "reference_time_seconds": 2.7,
698
- "saved_frame": 269,
699
- "saved_source_time_seconds": 2.7
700
- },
701
- {
702
- "video_frame": 82,
703
- "video_time_seconds": 2.7333333333333334,
704
- "reference_time_seconds": 2.73,
705
- "saved_frame": 272,
706
- "saved_source_time_seconds": 2.73
707
- },
708
- {
709
- "video_frame": 83,
710
- "video_time_seconds": 2.7666666666666666,
711
- "reference_time_seconds": 2.77,
712
- "saved_frame": 276,
713
- "saved_source_time_seconds": 2.77
714
- },
715
- {
716
- "video_frame": 84,
717
- "video_time_seconds": 2.8,
718
- "reference_time_seconds": 2.8000000000000003,
719
- "saved_frame": 279,
720
- "saved_source_time_seconds": 2.8000000000000003
721
- },
722
- {
723
- "video_frame": 85,
724
- "video_time_seconds": 2.8333333333333335,
725
- "reference_time_seconds": 2.83,
726
- "saved_frame": 282,
727
- "saved_source_time_seconds": 2.83
728
- },
729
- {
730
- "video_frame": 86,
731
- "video_time_seconds": 2.8666666666666667,
732
- "reference_time_seconds": 2.87,
733
- "saved_frame": 286,
734
- "saved_source_time_seconds": 2.87
735
- },
736
- {
737
- "video_frame": 87,
738
- "video_time_seconds": 2.9,
739
- "reference_time_seconds": 2.9,
740
- "saved_frame": 289,
741
- "saved_source_time_seconds": 2.9
742
- },
743
- {
744
- "video_frame": 88,
745
- "video_time_seconds": 2.933333333333333,
746
- "reference_time_seconds": 2.93,
747
- "saved_frame": 292,
748
- "saved_source_time_seconds": 2.93
749
- },
750
- {
751
- "video_frame": 89,
752
- "video_time_seconds": 2.966666666666667,
753
- "reference_time_seconds": 2.97,
754
- "saved_frame": 296,
755
- "saved_source_time_seconds": 2.97
756
- },
757
- {
758
- "video_frame": 90,
759
- "video_time_seconds": 3.0,
760
- "reference_time_seconds": 3.0,
761
- "saved_frame": 299,
762
- "saved_source_time_seconds": 3.0
763
- },
764
- {
765
- "video_frame": 91,
766
- "video_time_seconds": 3.033333333333333,
767
- "reference_time_seconds": 3.0300000000000002,
768
- "saved_frame": 302,
769
- "saved_source_time_seconds": 3.0300000000000002
770
- },
771
- {
772
- "video_frame": 92,
773
- "video_time_seconds": 3.066666666666667,
774
- "reference_time_seconds": 3.0700000000000003,
775
- "saved_frame": 306,
776
- "saved_source_time_seconds": 3.0700000000000003
777
- },
778
- {
779
- "video_frame": 93,
780
- "video_time_seconds": 3.1,
781
- "reference_time_seconds": 3.1,
782
- "saved_frame": 309,
783
- "saved_source_time_seconds": 3.1
784
- },
785
- {
786
- "video_frame": 94,
787
- "video_time_seconds": 3.1333333333333333,
788
- "reference_time_seconds": 3.13,
789
- "saved_frame": 312,
790
- "saved_source_time_seconds": 3.13
791
- },
792
- {
793
- "video_frame": 95,
794
- "video_time_seconds": 3.1666666666666665,
795
- "reference_time_seconds": 3.17,
796
- "saved_frame": 316,
797
- "saved_source_time_seconds": 3.17
798
- },
799
- {
800
- "video_frame": 96,
801
- "video_time_seconds": 3.2,
802
- "reference_time_seconds": 3.2,
803
- "saved_frame": 319,
804
- "saved_source_time_seconds": 3.2
805
- },
806
- {
807
- "video_frame": 97,
808
- "video_time_seconds": 3.2333333333333334,
809
- "reference_time_seconds": 3.23,
810
- "saved_frame": 322,
811
- "saved_source_time_seconds": 3.23
812
- },
813
- {
814
- "video_frame": 98,
815
- "video_time_seconds": 3.2666666666666666,
816
- "reference_time_seconds": 3.27,
817
- "saved_frame": 326,
818
- "saved_source_time_seconds": 3.27
819
- },
820
- {
821
- "video_frame": 99,
822
- "video_time_seconds": 3.3,
823
- "reference_time_seconds": 3.3000000000000003,
824
- "saved_frame": 329,
825
- "saved_source_time_seconds": 3.3000000000000003
826
- },
827
- {
828
- "video_frame": 100,
829
- "video_time_seconds": 3.3333333333333335,
830
- "reference_time_seconds": 3.33,
831
- "saved_frame": 332,
832
- "saved_source_time_seconds": 3.33
833
- },
834
- {
835
- "video_frame": 101,
836
- "video_time_seconds": 3.3666666666666667,
837
- "reference_time_seconds": 3.37,
838
- "saved_frame": 336,
839
- "saved_source_time_seconds": 3.37
840
- },
841
- {
842
- "video_frame": 102,
843
- "video_time_seconds": 3.4,
844
- "reference_time_seconds": 3.4,
845
- "saved_frame": 339,
846
- "saved_source_time_seconds": 3.4
847
- },
848
- {
849
- "video_frame": 103,
850
- "video_time_seconds": 3.433333333333333,
851
- "reference_time_seconds": 3.43,
852
- "saved_frame": 342,
853
- "saved_source_time_seconds": 3.43
854
- },
855
- {
856
- "video_frame": 104,
857
- "video_time_seconds": 3.466666666666667,
858
- "reference_time_seconds": 3.47,
859
- "saved_frame": 346,
860
- "saved_source_time_seconds": 3.47
861
- },
862
- {
863
- "video_frame": 105,
864
- "video_time_seconds": 3.5,
865
- "reference_time_seconds": 3.5,
866
- "saved_frame": 349,
867
- "saved_source_time_seconds": 3.5
868
- },
869
- {
870
- "video_frame": 106,
871
- "video_time_seconds": 3.533333333333333,
872
- "reference_time_seconds": 3.5300000000000002,
873
- "saved_frame": 352,
874
- "saved_source_time_seconds": 3.5300000000000002
875
- },
876
- {
877
- "video_frame": 107,
878
- "video_time_seconds": 3.566666666666667,
879
- "reference_time_seconds": 3.5700000000000003,
880
- "saved_frame": 356,
881
- "saved_source_time_seconds": 3.5700000000000003
882
- },
883
- {
884
- "video_frame": 108,
885
- "video_time_seconds": 3.6,
886
- "reference_time_seconds": 3.6,
887
- "saved_frame": 359,
888
- "saved_source_time_seconds": 3.6
889
- },
890
- {
891
- "video_frame": 109,
892
- "video_time_seconds": 3.6333333333333333,
893
- "reference_time_seconds": 3.63,
894
- "saved_frame": 362,
895
- "saved_source_time_seconds": 3.63
896
- },
897
- {
898
- "video_frame": 110,
899
- "video_time_seconds": 3.6666666666666665,
900
- "reference_time_seconds": 3.67,
901
- "saved_frame": 366,
902
- "saved_source_time_seconds": 3.67
903
- },
904
- {
905
- "video_frame": 111,
906
- "video_time_seconds": 3.7,
907
- "reference_time_seconds": 3.7,
908
- "saved_frame": 369,
909
- "saved_source_time_seconds": 3.7
910
- },
911
- {
912
- "video_frame": 112,
913
- "video_time_seconds": 3.7333333333333334,
914
- "reference_time_seconds": 3.73,
915
- "saved_frame": 372,
916
- "saved_source_time_seconds": 3.73
917
- },
918
- {
919
- "video_frame": 113,
920
- "video_time_seconds": 3.7666666666666666,
921
- "reference_time_seconds": 3.77,
922
- "saved_frame": 376,
923
- "saved_source_time_seconds": 3.77
924
- },
925
- {
926
- "video_frame": 114,
927
- "video_time_seconds": 3.8,
928
- "reference_time_seconds": 3.8000000000000003,
929
- "saved_frame": 379,
930
- "saved_source_time_seconds": 3.8000000000000003
931
- },
932
- {
933
- "video_frame": 115,
934
- "video_time_seconds": 3.8333333333333335,
935
- "reference_time_seconds": 3.83,
936
- "saved_frame": 382,
937
- "saved_source_time_seconds": 3.83
938
- },
939
- {
940
- "video_frame": 116,
941
- "video_time_seconds": 3.8666666666666667,
942
- "reference_time_seconds": 3.87,
943
- "saved_frame": 386,
944
- "saved_source_time_seconds": 3.87
945
- },
946
- {
947
- "video_frame": 117,
948
- "video_time_seconds": 3.9,
949
- "reference_time_seconds": 3.9,
950
- "saved_frame": 389,
951
- "saved_source_time_seconds": 3.9
952
- },
953
- {
954
- "video_frame": 118,
955
- "video_time_seconds": 3.933333333333333,
956
- "reference_time_seconds": 3.93,
957
- "saved_frame": 392,
958
- "saved_source_time_seconds": 3.93
959
- },
960
- {
961
- "video_frame": 119,
962
- "video_time_seconds": 3.966666666666667,
963
- "reference_time_seconds": 3.966666666666667,
964
- "saved_frame": 396,
965
- "saved_source_time_seconds": 3.97
966
- }
967
- ],
968
- "gl_backend": "egl"
969
- },
970
- "evidence_complete": true
971
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/evidence/native/thumbnail.jpg DELETED

Git LFS Details

  • SHA256: cafeba69360c2cd179fb7b80d29bd3108e5561124da1407edde7cc75a673eb87
  • Pointer size: 130 Bytes
  • Size of remote file: 98.6 kB
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/evidence/replay_saved_controls.py DELETED
@@ -1,63 +0,0 @@
1
- """Independently integrate saved native commands; never restore rollout states."""
2
- from pathlib import Path
3
- import hashlib
4
- import json
5
- import mujoco
6
- import numpy as np
7
- from spider.config import Config, process_config
8
- from spider.io import load_data
9
- from spider.simulators.mjwp import setup_mj_model, initialize_hand_controls
10
- from examples.evaluate_retarget import load_rollout, evaluate
11
- run = Path(__file__).resolve().parent.parent
12
- values = json.loads((run/'config.input.json').read_text())
13
- values.update(device='cpu', dataset_dir=str(run/'dataset'), num_samples=1, save_video=False)
14
- cfg = process_config(Config(**values))
15
- ref = load_data(cfg, cfg.data_path)
16
- model = setup_mj_model(cfg)
17
- with np.load(Path(cfg.output_dir)/'trajectory_mjwp_fast.npz') as archive:
18
- saved_q = archive['qpos'].reshape(-1,model.nq)
19
- saved_ctrl = archive['ctrl'].reshape(-1,model.nu)
20
- saved_times = archive['time'].reshape(-1)
21
- controls_ref = initialize_hand_controls(cfg,model,*ref[:3])
22
- live = mujoco.MjData(model)
23
- live.qpos[:] = ref[0][0].numpy()
24
- live.qvel[:] = ref[1][0].numpy()
25
- live.ctrl[:] = controls_ref[0].numpy()
26
- mujoco.mj_forward(model,live)
27
- qpos=[]; qvel=[]; timestamps=[]
28
- for ctrl in saved_ctrl:
29
- live.ctrl[:]=ctrl
30
- mujoco.mj_step(model,live)
31
- qpos.append(live.qpos.copy()); qvel.append(live.qvel.copy()); timestamps.append(live.time)
32
- qpos=np.asarray(qpos)
33
- np.testing.assert_allclose(timestamps,saved_times,atol=1e-10,rtol=0)
34
- out=run/'evidence'; out.mkdir(exist_ok=True)
35
- replay_path=out/'trajectory_control_replay.npz'
36
- np.savez(replay_path,qpos=qpos,qvel=qvel,ctrl=saved_ctrl,time=timestamps)
37
- report,_=evaluate(load_rollout(scene=Path(cfg.model_path),trajectory=replay_path,reference=Path(cfg.data_path),metadata_path=run/'config.input.json',sim_dt=cfg.sim_dt))
38
- (out/'evaluation_control_replay.json').write_text(json.dumps(report,indent=2)+'\n')
39
- actual,target=mujoco.MjData(model),mujoco.MjData(model)
40
- sites=[model.site('left_'+name).id for name in ('palm','thumb_tip','index_tip','middle_tip','ring_tip','pinky_tip')]
41
- max_site_delta=0.
42
- for qa,qb in zip(qpos,saved_q,strict=True):
43
- actual.qpos[:]=qa; target.qpos[:]=qb
44
- mujoco.mj_kinematics(model,actual);mujoco.mj_kinematics(model,target)
45
- max_site_delta=max(max_site_delta,float(np.linalg.norm(actual.site_xpos[sites]-target.site_xpos[sites],axis=-1).max()))
46
- summary={
47
- 'backend':'CPU MuJoCo 3.7 independent integration of native saved controls',
48
- 'steps':len(saved_ctrl),'final_sim_time':live.time,
49
- 'no_state_overrides_after_initialization':True,
50
- 'control_limits_enabled':True,
51
- 'control_sha256':hashlib.sha256(saved_ctrl.tobytes()).hexdigest(),
52
- 'hand_control_mode':cfg.hand_control_mode,'hand_gain_scale':cfg.hand_gain_scale,
53
- 'effective_kp':model.actuator_gainprm[:,0].tolist(),
54
- 'effective_kd':(-model.actuator_biasprm[:,2]).tolist(),
55
- 'max_abs_qpos_difference_vs_native':float(np.max(np.abs(qpos-saved_q))),
56
- 'max_hand_site_difference_vs_native_m':max_site_delta,
57
- 'replay_success':report['success'],
58
- 'full_source_covered':report['coverage']['full_source_covered'],
59
- 'max_site_error_m':report['per_hand']['left']['site_position_error_m']['max'],
60
- 'max_palm_rotation_error_rad':report['per_hand']['left']['palm_orientation_error_rad']['max'],
61
- }
62
- (out/'control_replay.json').write_text(json.dumps(summary,indent=2)+'\n')
63
- print(json.dumps(summary,indent=2))
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/evidence/trajectory_control_replay.npz DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:158424c01bbccfff10b46f53b22190bd21c26583808c3966169cf90beedb25b5
3
- size 314582
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/hydra/run_mjwp_fast.log DELETED
File without changes
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/manifest.json DELETED
@@ -1,251 +0,0 @@
1
- {
2
- "kind": "retarget",
3
- "source": {
4
- "dataset_dir": "/home/ubuntu/sky_workdir/example_datasets/egoverse",
5
- "dataset_name": "egoverse_hand",
6
- "task": "fold_clothes_left",
7
- "embodiment_type": "left",
8
- "robot_type": "sharpa",
9
- "data_id": 0
10
- },
11
- "task_info": {
12
- "task": "fold_clothes_left",
13
- "dataset_name": "egoverse_hand",
14
- "robot_type": "sharpa",
15
- "embodiment_type": "left",
16
- "data_id": 0,
17
- "ref_dt": 0.03333333333333333,
18
- "hand_only": true,
19
- "right_object_mesh_dir": null,
20
- "left_object_mesh_dir": null,
21
- "right_object_convex_dir": null,
22
- "left_object_convex_dir": null,
23
- "floor_height": -0.001,
24
- "hand_floor_collision": false,
25
- "object_floor_collision": false,
26
- "object_object_collision": false,
27
- "scene_builder": "examples.egoverse.hand_scene",
28
- "object_placeholder": "inertial_only; zero geometries; gravity compensated"
29
- },
30
- "output_dir": "/home/ubuntu/sky_workdir/example_datasets/full_retarget_campaign/runs/egoverse_trial2",
31
- "artifact_dir": "/home/ubuntu/sky_workdir/example_datasets/full_retarget_campaign/runs/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/0",
32
- "config_path": "/home/ubuntu/sky_workdir/example_datasets/full_retarget_campaign/runs/egoverse_trial2/config.input.json",
33
- "config": {
34
- "seed": 0,
35
- "sanity_check_seconds": 3.0,
36
- "sanity_check_pos_thresh": 0.01,
37
- "sanity_check_rot_thresh": 0.3,
38
- "simulator": "mjwp",
39
- "device": "cuda:0",
40
- "sim_dt": 0.01,
41
- "ctrl_dt": 0.1,
42
- "render_dt": 0.02,
43
- "horizon": 0.8,
44
- "knot_dt": 0.1,
45
- "nconmax_per_env": 100,
46
- "njmax_per_env": 350,
47
- "num_dyn": 1,
48
- "num_dr": 1,
49
- "pair_margin_range": [
50
- -0.005,
51
- 0.005
52
- ],
53
- "xy_offset_range": [
54
- -0.005,
55
- 0.005
56
- ],
57
- "perturb_force": 0.0,
58
- "perturb_torque": 0.0,
59
- "contact_guidance": false,
60
- "hand_control_mode": "inverse_dynamics",
61
- "hand_gain_scale": 10,
62
- "init_pos_actuator_gain": 10.0,
63
- "init_pos_actuator_bias": 10.0,
64
- "init_rot_actuator_gain": 0.1,
65
- "init_rot_actuator_bias": 0.1,
66
- "guidance_decay_ratio": 0.5,
67
- "gibbs_sampling": false,
68
- "num_samples": 128,
69
- "temperature": 0.3,
70
- "max_num_iterations": 8,
71
- "improvement_threshold": 0.01,
72
- "improvement_check_steps": 1,
73
- "terminate_resample": false,
74
- "object_pos_threshold": 0.1,
75
- "object_rot_threshold": 0.3,
76
- "max_local_retries": 1,
77
- "max_revert_depth": 1,
78
- "base_pos_threshold": 0.02,
79
- "base_rot_threshold": 0.2,
80
- "use_torch_compile": false,
81
- "first_ctrl_noise_scale": 0.5,
82
- "last_ctrl_noise_scale": 1.0,
83
- "final_noise_scale": 0.1,
84
- "exploit_ratio": 0.01,
85
- "exploit_noise_scale": 0.01,
86
- "joint_noise_scale": 0.02,
87
- "pos_noise_scale": 0.003,
88
- "rot_noise_scale": 0.005,
89
- "use_rl_reward": false,
90
- "base_pos_rew_scale": 1.0,
91
- "base_rot_rew_scale": 0.3,
92
- "joint_rew_scale": 0.03,
93
- "pos_rew_scale": 1.0,
94
- "rot_rew_scale": 0.3,
95
- "vel_rew_scale": 0.0001,
96
- "terminal_rew_scale": 1.0,
97
- "contact_rew_scale": 0.0,
98
- "show_viewer": false,
99
- "viewer": "mujoco",
100
- "wait_on_finish": false,
101
- "rerun_spawn": false,
102
- "save_video": true,
103
- "save_info": true,
104
- "save_rerun": false,
105
- "save_viser": false,
106
- "save_metrics": true,
107
- "save_config": true,
108
- "trace_dt": 0.02,
109
- "num_trace_uniform_samples": 4,
110
- "num_trace_topk_samples": 2,
111
- "dataset_dir": "/home/ubuntu/sky_workdir/example_datasets/full_retarget_campaign/runs/egoverse_trial2/dataset",
112
- "dataset_name": "egoverse_hand",
113
- "task": "fold_clothes_left",
114
- "embodiment_type": "left",
115
- "robot_type": "sharpa",
116
- "data_id": 0,
117
- "ref_dt": 0.03333333333333333,
118
- "max_sim_steps": 400
119
- },
120
- "artifacts": {
121
- "scene": "dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/scene.xml",
122
- "task_info": "dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/task_info.json",
123
- "reference": "dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/0/trajectory_kinematic.npz",
124
- "trajectory": "dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/0/trajectory_mjwp_fast.npz",
125
- "config": "config.input.json",
126
- "parameters": "parameters.json",
127
- "metrics": "metrics.json",
128
- "log": "run.log"
129
- },
130
- "timing": {
131
- "ref_dt": 0.03333333333333333,
132
- "sim_dt": 0.01,
133
- "reference_frames": 120,
134
- "reference_duration_seconds": 3.966666666666667,
135
- "requested_max_sim_steps": 400,
136
- "effective_max_sim_steps": 400,
137
- "reference_sim_steps": 397,
138
- "post_reference_hold_steps": 3,
139
- "timeout_seconds": 1200.0
140
- },
141
- "staging": {
142
- "scene": "/home/ubuntu/sky_workdir/example_datasets/egoverse/processed/egoverse_hand/sharpa/left/fold_clothes_left/scene.xml",
143
- "trajectory": "/home/ubuntu/sky_workdir/example_datasets/egoverse/processed/egoverse_hand/sharpa/left/fold_clothes_left/0/trajectory_kinematic.npz"
144
- },
145
- "argv": [
146
- "/home/ubuntu/sky_workdir/.venv/bin/python",
147
- "-m",
148
- "examples.run_mjwp_fast",
149
- "++load_config_path=\"/home/ubuntu/sky_workdir/example_datasets/full_retarget_campaign/runs/egoverse_trial2/config.input.json\"",
150
- "hydra.run.dir=\"/home/ubuntu/sky_workdir/example_datasets/full_retarget_campaign/runs/egoverse_trial2/hydra\"",
151
- "hydra.output_subdir=null",
152
- "hydra.job.chdir=false"
153
- ],
154
- "command": "/home/ubuntu/sky_workdir/.venv/bin/python -m examples.run_mjwp_fast '++load_config_path=\"/home/ubuntu/sky_workdir/example_datasets/full_retarget_campaign/runs/egoverse_trial2/config.input.json\"' 'hydra.run.dir=\"/home/ubuntu/sky_workdir/example_datasets/full_retarget_campaign/runs/egoverse_trial2/hydra\"' hydra.output_subdir=null hydra.job.chdir=false",
155
- "cwd": "/home/ubuntu/sky_workdir",
156
- "execute_argv": [
157
- "/home/ubuntu/sky_workdir/.venv/bin/python",
158
- "-m",
159
- "spider.pipeline",
160
- "retarget",
161
- "--dataset-dir",
162
- "example_datasets/egoverse",
163
- "--dataset-name",
164
- "egoverse_hand",
165
- "--task",
166
- "fold_clothes_left",
167
- "--embodiment-type",
168
- "left",
169
- "--output-dir",
170
- "example_datasets/full_retarget_campaign/runs/egoverse_trial2",
171
- "--max-sim-steps",
172
- "400",
173
- "--timeout-seconds",
174
- "1200",
175
- "--set",
176
- "num_samples=128",
177
- "--set",
178
- "max_num_iterations=8",
179
- "--set",
180
- "horizon=0.8",
181
- "--set",
182
- "ctrl_dt=0.1",
183
- "--set",
184
- "knot_dt=0.1",
185
- "--set",
186
- "joint_rew_scale=0.03",
187
- "--set",
188
- "base_pos_threshold=0.02",
189
- "--set",
190
- "base_rot_threshold=0.2",
191
- "--set",
192
- "hand_control_mode=inverse_dynamics",
193
- "--set",
194
- "hand_gain_scale=10",
195
- "--set",
196
- "pos_noise_scale=0.003",
197
- "--set",
198
- "rot_noise_scale=0.005",
199
- "--set",
200
- "joint_noise_scale=0.02",
201
- "--set",
202
- "max_local_retries=1",
203
- "--set",
204
- "max_revert_depth=1",
205
- "--set",
206
- "use_torch_compile=false",
207
- "--set",
208
- "save_video=true"
209
- ],
210
- "status": "succeeded",
211
- "started_at": "2026-10-06T22:08:28.790908+00:00",
212
- "metrics": {
213
- "tracking_domain": "hand",
214
- "saved_frames": 400,
215
- "scored_frames": 397,
216
- "completed_sim_steps": 400,
217
- "completion_fraction": 1.0,
218
- "source_completion_fraction": 1.0,
219
- "tracking_pos_err": 0.0011932878521424854,
220
- "tracking_rot_err": 0.0014171413975127898,
221
- "hand_pos_err": 0.0011932878521424854,
222
- "hand_rot_err": 0.0014171413975127898,
223
- "max_pos_error_m": 0.014962734140670845,
224
- "max_rot_error_rad": 0.005874543051005981,
225
- "tracking_within_thresholds": true,
226
- "position_threshold_m": 0.02,
227
- "rotation_threshold_rad": 0.2,
228
- "per_hand": {
229
- "left": {
230
- "mean_site_error_m": 0.0011932878521424854,
231
- "max_site_error_m": 0.014962734140670845,
232
- "mean_wrist_orientation_error_rad": 0.0014171413975127898
233
- }
234
- },
235
- "per_object": {},
236
- "manipulation_success": null,
237
- "definition": "Robot palm/fingertip FK position and wrist orientation versus the IK reference; no object/manipulation score."
238
- },
239
- "input_sha256": {
240
- "/home/ubuntu/sky_workdir/example_datasets/egoverse/processed/egoverse_hand/sharpa/left/fold_clothes_left/0/trajectory_kinematic.npz": "fb8aafe61241e8efc9866e9c82ecb0306394c0cc497f3a0b6ffbb3d7b3c9c20a",
241
- "/home/ubuntu/sky_workdir/example_datasets/egoverse/processed/egoverse_hand/sharpa/left/fold_clothes_left/scene.xml": "e2c1b74d188ec3e2c1a133457e9aa067155ad7fa12deef77c6ee9b4069d8f7b5",
242
- "/home/ubuntu/sky_workdir/example_datasets/egoverse/processed/egoverse_hand/sharpa/left/fold_clothes_left/task_info.json": "c63a6e6d5edba47ca7ce397b993f7d75fcf916a3661aea471959c71f685cc3c1"
243
- },
244
- "process": {
245
- "returncode": 0,
246
- "timed_out": false,
247
- "elapsed_seconds": 163.31359765700017
248
- },
249
- "finished_at": "2026-10-06T22:11:12.648843+00:00",
250
- "elapsed_seconds": 163.85792771100023
251
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/metrics.json DELETED
@@ -1,27 +0,0 @@
1
- {
2
- "tracking_domain": "hand",
3
- "saved_frames": 400,
4
- "scored_frames": 397,
5
- "completed_sim_steps": 400,
6
- "completion_fraction": 1.0,
7
- "source_completion_fraction": 1.0,
8
- "tracking_pos_err": 0.0011932878521424854,
9
- "tracking_rot_err": 0.0014171413975127898,
10
- "hand_pos_err": 0.0011932878521424854,
11
- "hand_rot_err": 0.0014171413975127898,
12
- "max_pos_error_m": 0.014962734140670845,
13
- "max_rot_error_rad": 0.005874543051005981,
14
- "tracking_within_thresholds": true,
15
- "position_threshold_m": 0.02,
16
- "rotation_threshold_rad": 0.2,
17
- "per_hand": {
18
- "left": {
19
- "mean_site_error_m": 0.0011932878521424854,
20
- "max_site_error_m": 0.014962734140670845,
21
- "mean_wrist_orientation_error_rad": 0.0014171413975127898
22
- }
23
- },
24
- "per_object": {},
25
- "manipulation_success": null,
26
- "definition": "Robot palm/fingertip FK position and wrist orientation versus the IK reference; no object/manipulation score."
27
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/parameters.json DELETED
@@ -1,79 +0,0 @@
1
- {
2
- "seed": 0,
3
- "sanity_check_seconds": 3.0,
4
- "sanity_check_pos_thresh": 0.01,
5
- "sanity_check_rot_thresh": 0.3,
6
- "simulator": "mjwp",
7
- "device": "cuda:0",
8
- "sim_dt": 0.01,
9
- "ctrl_dt": 0.1,
10
- "render_dt": 0.02,
11
- "horizon": 0.8,
12
- "knot_dt": 0.1,
13
- "nconmax_per_env": 100,
14
- "njmax_per_env": 350,
15
- "num_dyn": 1,
16
- "num_dr": 1,
17
- "pair_margin_range": [
18
- -0.005,
19
- 0.005
20
- ],
21
- "xy_offset_range": [
22
- -0.005,
23
- 0.005
24
- ],
25
- "perturb_force": 0.0,
26
- "perturb_torque": 0.0,
27
- "contact_guidance": false,
28
- "hand_control_mode": "inverse_dynamics",
29
- "hand_gain_scale": 10,
30
- "init_pos_actuator_gain": 10.0,
31
- "init_pos_actuator_bias": 10.0,
32
- "init_rot_actuator_gain": 0.1,
33
- "init_rot_actuator_bias": 0.1,
34
- "guidance_decay_ratio": 0.5,
35
- "gibbs_sampling": false,
36
- "num_samples": 128,
37
- "temperature": 0.3,
38
- "max_num_iterations": 8,
39
- "improvement_threshold": 0.01,
40
- "improvement_check_steps": 1,
41
- "terminate_resample": false,
42
- "object_pos_threshold": 0.1,
43
- "object_rot_threshold": 0.3,
44
- "max_local_retries": 1,
45
- "max_revert_depth": 1,
46
- "base_pos_threshold": 0.02,
47
- "base_rot_threshold": 0.2,
48
- "use_torch_compile": false,
49
- "first_ctrl_noise_scale": 0.5,
50
- "last_ctrl_noise_scale": 1.0,
51
- "final_noise_scale": 0.1,
52
- "exploit_ratio": 0.01,
53
- "exploit_noise_scale": 0.01,
54
- "joint_noise_scale": 0.02,
55
- "pos_noise_scale": 0.003,
56
- "rot_noise_scale": 0.005,
57
- "use_rl_reward": false,
58
- "base_pos_rew_scale": 1.0,
59
- "base_rot_rew_scale": 0.3,
60
- "joint_rew_scale": 0.03,
61
- "pos_rew_scale": 1.0,
62
- "rot_rew_scale": 0.3,
63
- "vel_rew_scale": 0.0001,
64
- "terminal_rew_scale": 1.0,
65
- "contact_rew_scale": 0.0,
66
- "show_viewer": false,
67
- "viewer": "mujoco",
68
- "wait_on_finish": false,
69
- "rerun_spawn": false,
70
- "save_video": true,
71
- "save_info": true,
72
- "save_rerun": false,
73
- "save_viser": false,
74
- "save_metrics": true,
75
- "save_config": true,
76
- "trace_dt": 0.02,
77
- "num_trace_uniform_samples": 4,
78
- "num_trace_topk_samples": 2
79
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/run.log DELETED
@@ -1,43 +0,0 @@
1
- Warp 1.12.1 initialized:
2
- CUDA Toolkit 12.9, Driver 13.0
3
- Devices:
4
- "cpu" : "x86_64"
5
- "cuda:0" : "NVIDIA L40S" (44 GiB, sm_89, mempool enabled)
6
- Kernel cache:
7
- /home/ubuntu/.cache/warp/1.12.1
8
- 2026-10-06 22:08:35.137 | INFO | spider.config:process_config:557 - overriding ref_dt: 0.03333333333333333 from task_info.json
9
- 2026-10-06 22:08:35.142 | WARNING | spider.io:load_data:55 - contact data not found
10
- 2026-10-06 22:08:35.142 | WARNING | spider.io:load_data:60 - contact_pos data not found
11
- 2026-10-06 22:08:35.142 | WARNING | spider.io:load_data:66 - ctrl data not found, using 'qpos' as a initial guess for control.
12
- 2026-10-06 22:08:35.389 | INFO | spider.simulators.mjwp:initialize_hand_controls:198 - Initialized 28 hand servos with inverse_dynamics feedforward; gain scale=10.
13
- Module mujoco_warp._src.smooth 26bbc02 load on device 'cuda:0' took 3.26 ms (cached)
14
- Module mujoco_warp._src.collision_driver 88f7770 load on device 'cuda:0' took 0.57 ms (cached)
15
- Module _primitive_narrowphase__locals__primitive_narrowphase_e32dc1af e32dc1a load on device 'cuda:0' took 0.43 ms (cached)
16
- Module mujoco_warp._src.constraint c567367 load on device 'cuda:0' took 5.12 ms (cached)
17
- Module mujoco_warp._src.forward 16fbac8 load on device 'cuda:0' took 1.18 ms (cached)
18
- Module mujoco_warp._src.passive 8bfe11c load on device 'cuda:0' took 1.08 ms (cached)
19
- Module mujoco_warp._src.forward 8f5355c load on device 'cuda:0' took 2.74 ms (cached)
20
- Module mujoco_warp._src.support dd3f924 load on device 'cuda:0' took 0.55 ms (cached)
21
- Module _solve_LD_sparse_fused__locals__kernel_2d0d6a46 dc0f5bc load on device 'cuda:0' took 0.90 ms (cached)
22
- Module mujoco_warp._src.solver b116efb load on device 'cuda:0' took 1.90 ms (cached)
23
- Module solve_init_jaref__locals__kernel_84f69353 84f6935 load on device 'cuda:0' took 1.62 ms (cached)
24
- Module mul_m_sparse__locals___mul_m_sparse_aa423e43 aa423e4 load on device 'cuda:0' took 0.60 ms (cached)
25
- Module update_constraint_efc__locals__kernel_81015163 8101516 load on device 'cuda:0' took 2.39 ms (cached)
26
- Module update_constraint_gauss_cost__locals__kernel_0eaf1ba5 0eaf1ba load on device 'cuda:0' took 0.59 ms (cached)
27
- Module update_gradient_cholesky_blocked__locals__kernel_475d8c4a 33bb335 load on device 'cuda:0' took 234.06 ms (cached)
28
- Module linesearch_jv_fused__locals__kernel_56b0bf37 56b0bf3 load on device 'cuda:0' took 2.44 ms (cached)
29
- Module linesearch_iterative__locals__kernel_6b81acbf 10af5cd load on device 'cuda:0' took 1.14 ms (cached)
30
- Module update_constraint_efc__locals__kernel_25737258 2573725 load on device 'cuda:0' took 2.90 ms (cached)
31
- Module mujoco_warp._src.sensor 18a78e5 load on device 'cuda:0' took 2.44 ms (cached)
32
- Module mujoco_warp._src.derivative 7ba27fe load on device 'cuda:0' took 1.63 ms (cached)
33
- 2026-10-06 22:08:36.237 | INFO | spider.simulators.mjwp:_run_sanity_check_body:678 - Running initial-state sanity check: holding ctrl_ref[0] for 3.00s (1500 steps).
34
- 2026-10-06 22:08:36.715 | INFO | spider.simulators.mjwp:_run_sanity_check_body:791 - Saved sanity-check video to /home/ubuntu/sky_workdir/example_datasets/full_retarget_campaign/runs/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/0/sanity_check_drift.mp4
35
- 2026-10-06 22:08:36.715 | INFO | spider.simulators.mjwp:_run_sanity_check_body:802 - Sanity check passed (1500 steps held at ctrl_ref[0]). hand palm/fingertip max pos drift=0.0893cm; hand palm max rot drift=0.0034rad
36
- 2026-10-06 22:08:36.740 | INFO | __main__:_save_config_yaml:136 - Saved config to /home/ubuntu/sky_workdir/example_datasets/full_retarget_campaign/runs/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/0/config.yaml
37
- 2026-10-06 22:08:36.740 | INFO | spider.viewers:setup_viewer:129 - viewer is disabled, launch dummy viewer
38
- Realtime rate: 0.03, plan time: 3.9667s, sim_steps: 10/400, opt_steps: 8, attempts: 1, state: OPTIMIZE, boost: 0
39
- 2026-10-06 22:11:04.819 | INFO | __main__:main:722 - Saved info to /home/ubuntu/sky_workdir/example_datasets/full_retarget_campaign/runs/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/0/trajectory_mjwp_fast.npz
40
- 2026-10-06 22:11:07.699 | INFO | __main__:main:733 - Saved video to /home/ubuntu/sky_workdir/example_datasets/full_retarget_campaign/runs/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/0/visualization_mjwp_fast.mp4
41
- 2026-10-06 22:11:07.702 | INFO | __main__:main:752 - Final hand tracking error: base pos=0.0004, rot=0.0014, max finger joint=0.2236
42
- 2026-10-06 22:11:07.703 | INFO | __main__:main:773 - Smoothness: max frame-to-frame qpos diff=0.583773 at frame 133/400
43
- 2026-10-06 22:11:10.124 | WARNING | __main__:_emergency_flush:341 - [emergency flush] Saved partial video (200 frames) to /home/ubuntu/sky_workdir/example_datasets/full_retarget_campaign/runs/egoverse_trial2/dataset/processed/egoverse_hand/sharpa/left/fold_clothes_left/0/visualization_mjwp_fast.mp4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/runtime_source/examples/run_mjwp_fast.py DELETED
@@ -1,852 +0,0 @@
1
- # Copyright (c) Meta Platforms, Inc. and affiliates.
2
- # All rights reserved.
3
-
4
- # This source code is licensed under the license found in the
5
- # LICENSE file in the root directory of this source tree.
6
-
7
- """Deterministic optimization variant of run_mjwp.py.
8
-
9
- Instead of re-rolling out the optimized controls, this script replays the
10
- pre-computed best trajectory states and restores the full simulation state
11
- from the optimizer's snapshot.
12
-
13
- Retry/revert is modeled as an explicit state machine — see `_StepState`
14
- and the transition table at the top of `main()`.
15
-
16
- Author: Chaoyi Pan
17
- Date: 2025-04-16
18
- """
19
-
20
- from __future__ import annotations
21
-
22
- import collections
23
- import math
24
- import sys
25
- import time
26
- from dataclasses import dataclass, field, fields
27
- from pathlib import Path
28
-
29
- import hydra
30
- import imageio
31
- import loguru
32
- import mujoco
33
- import numpy as np
34
- import torch
35
- from omegaconf import DictConfig, OmegaConf
36
-
37
- from spider.config import (
38
- Config,
39
- filter_config_fields,
40
- load_config_yaml,
41
- process_config,
42
- )
43
- from spider.interp import get_slice
44
- from spider.io import load_data
45
- from spider.optimizers.sampling_fast import (
46
- _restore_best_state_to_env,
47
- make_optimize_fn_fast,
48
- make_optimize_once_fn_fast,
49
- make_rollout_fn_fast,
50
- )
51
- from spider.postprocess.get_success_rate import compute_object_tracking_error
52
- from spider.simulators.mjwp import (
53
- _initial_state_sanity_check,
54
- copy_sample_state,
55
- get_qpos,
56
- get_qvel,
57
- get_reward,
58
- get_terminal_reward,
59
- get_terminate,
60
- get_trace,
61
- hand_tracking_error,
62
- initialize_hand_controls,
63
- load_env_params,
64
- load_state,
65
- save_env_params,
66
- save_state,
67
- setup_env,
68
- setup_mj_model,
69
- step_env,
70
- sync_env_mujoco,
71
- )
72
- from spider.viewers import (
73
- log_frame,
74
- render_image,
75
- setup_renderer,
76
- setup_viewer,
77
- update_viewer,
78
- )
79
-
80
- _CONFIG_SKIP_FIELDS = {
81
- "noise_scale",
82
- "env_params_list",
83
- "viewer_body_entity_and_ids",
84
- }
85
-
86
-
87
- def _parse_override_tokens(tokens: list[str]) -> dict:
88
- allowed = {field.name for field in fields(Config)}
89
- override_dict: dict = {}
90
- for item in tokens:
91
- if "=" not in item:
92
- continue
93
- key, value = item.split("=", 1)
94
- key = key.lstrip("+")
95
- if key not in allowed:
96
- continue
97
- parsed = OmegaConf.to_container(
98
- OmegaConf.from_dotlist([f"{key}={value}"]), resolve=True
99
- )
100
- if isinstance(parsed, dict) and key in parsed:
101
- override_dict[key] = parsed[key]
102
- return override_dict
103
-
104
-
105
- def _extract_cli_overrides(cfg: DictConfig) -> dict:
106
- """Extract CLI overrides so they can be applied on top of a loaded config."""
107
- overrides = OmegaConf.select(cfg, "hydra.overrides.task") or []
108
- override_dict = _parse_override_tokens(overrides)
109
- if override_dict:
110
- return override_dict
111
- return _parse_override_tokens(sys.argv[1:])
112
-
113
-
114
- def _normalize_yaml_value(value):
115
- if isinstance(value, torch.Tensor):
116
- return value.detach().cpu().tolist()
117
- if isinstance(value, np.ndarray):
118
- return value.tolist()
119
- if isinstance(value, np.generic):
120
- return value.item()
121
- if isinstance(value, tuple):
122
- return list(value)
123
- return value
124
-
125
-
126
- def _save_config_yaml(config: Config) -> None:
127
- if not config.save_config:
128
- return
129
- config_dict = {}
130
- for fld in fields(config):
131
- if fld.name in _CONFIG_SKIP_FIELDS:
132
- continue
133
- config_dict[fld.name] = _normalize_yaml_value(getattr(config, fld.name))
134
- output_path = Path(config.output_dir) / "config.yaml"
135
- OmegaConf.save(config=OmegaConf.create(config_dict), f=str(output_path))
136
- loguru.logger.info(f"Saved config to {output_path}")
137
-
138
-
139
- # --- Progressive retry boost ------------------------------------------------
140
- #
141
- # On each retry/revert pop, boost_level increments. Iterations double per
142
- # level, noise scales by sqrt(2) per level — both fixed-base (relative to
143
- # the captured originals), so level=2 is orig*4 iterations, orig*2 noise.
144
-
145
- _BOOST_NOISE_FACTOR = math.sqrt(2.0)
146
-
147
-
148
- def _apply_boost(
149
- config: Config,
150
- boost_level: int,
151
- orig_max_iters: int,
152
- orig_env_params_list: list,
153
- orig_noise_scale: torch.Tensor,
154
- ) -> None:
155
- iters = orig_max_iters * (2**boost_level)
156
- config.max_num_iterations = iters
157
- config.beta_traj = config.final_noise_scale ** (1.0 / iters)
158
- extended = list(orig_env_params_list)
159
- while len(extended) < iters:
160
- extended.append(extended[-1])
161
- config.env_params_list = extended
162
- config.noise_scale = orig_noise_scale * (_BOOST_NOISE_FACTOR**boost_level)
163
-
164
-
165
- # --- Retry/revert state machine ---------------------------------------------
166
- #
167
- # States and transitions (event -> action):
168
- #
169
- # OPTIMIZE succeed -> OPTIMIZE
170
- # fail, local<cap -> LOCAL_RETRY
171
- # fail, local>=cap, history -> REVERT (first)
172
- # fail, local>=cap, no history -> DONE
173
- #
174
- # LOCAL_RETRY (transient) reload curr_cp, boost -> (unchanged state)
175
- #
176
- # REVERT (first, from OPTIMIZE) pop once, save anchor -> RECOVERING
177
- #
178
- # RECOVERING succeed, sim_step > anchor -> OPTIMIZE (full recovery)
179
- # succeed, sim_step <= anchor -> RECOVERING (no push)
180
- # fail, local<cap -> LOCAL_RETRY
181
- # fail, local>=cap, attempts<max -> RE-REVERT (same anchor)
182
- # fail, local>=cap, attempts>=max -> DONE
183
- #
184
- # "Recovery anchor" = sim_step right after the first REVERT pop. On subsequent
185
- # failures during recovery, we reload the *same* anchor checkpoint instead of
186
- # popping deeper into history. Recovery is declared when a step succeeds at
187
- # sim_step > anchor.
188
-
189
-
190
- class _StepState:
191
- OPTIMIZE = "OPTIMIZE"
192
- RECOVERING = "RECOVERING"
193
- DONE = "DONE"
194
-
195
-
196
- @dataclass
197
- class _LoopCtx:
198
- state: str = _StepState.OPTIMIZE
199
- history: collections.deque = field(default_factory=collections.deque)
200
- local_retries: int = 0
201
- boost_level: int = 0
202
- revert_anchor_step: int | None = None
203
- revert_anchor_checkpoint: dict | None = None
204
- revert_attempts_from_anchor: int = 0
205
-
206
-
207
- def main(config: Config):
208
- """Run SPIDER using MuJoCo Warp backend with deterministic best-trajectory replay."""
209
- config = process_config(config)
210
-
211
- if config.contact_guidance:
212
- raise NotImplementedError(
213
- "contact_guidance is not supported in run_mjwp_fast.py. "
214
- "Use examples/run_mjwp.py for contact-guided runs."
215
- )
216
- if config.gibbs_sampling:
217
- raise NotImplementedError(
218
- "gibbs_sampling is not supported in run_mjwp_fast.py. "
219
- "Use examples/run_mjwp.py for Gibbs sampling."
220
- )
221
-
222
- qpos_ref, qvel_ref, ctrl_ref, contact, contact_pos = load_data(
223
- config, config.data_path
224
- )
225
- mj_model = setup_mj_model(config)
226
- ctrl_ref = initialize_hand_controls(
227
- config, mj_model, qpos_ref, qvel_ref, ctrl_ref
228
- )
229
-
230
- ref_data = (qpos_ref, qvel_ref, ctrl_ref, contact, contact_pos)
231
- config.max_sim_steps = (
232
- config.max_sim_steps
233
- if config.max_sim_steps > 0
234
- else qpos_ref.shape[0] - config.horizon_steps - config.ctrl_steps
235
- )
236
-
237
- env = setup_env(config, ref_data)
238
-
239
- mj_data = mujoco.MjData(mj_model)
240
- mj_data_ref = mujoco.MjData(mj_model)
241
- mj_data.qpos[:] = qpos_ref[0].detach().cpu().numpy()
242
- mj_data.qvel[:] = qvel_ref[0].detach().cpu().numpy()
243
- mj_data.ctrl[:] = ctrl_ref[0].detach().cpu().numpy()
244
- mujoco.mj_forward(mj_model, mj_data)
245
- mj_data.time = 0.0
246
-
247
- # Initial-state stability sanity check (CPU MuJoCo only). Disabled when
248
- # config.sanity_check_seconds <= 0.0.
249
- if config.sanity_check_seconds > 0.0:
250
- _initial_state_sanity_check(
251
- mj_model,
252
- mj_data,
253
- qpos_ref,
254
- qvel_ref,
255
- ctrl_ref,
256
- config,
257
- save_video_dir=config.output_dir,
258
- )
259
- # Restore initial scene state after the check.
260
- mj_data.qpos[:] = qpos_ref[0].detach().cpu().numpy()
261
- mj_data.qvel[:] = qvel_ref[0].detach().cpu().numpy()
262
- mj_data.ctrl[:] = ctrl_ref[0].detach().cpu().numpy()
263
- mujoco.mj_forward(mj_model, mj_data)
264
- mj_data.time = 0.0
265
-
266
- images = []
267
- object_trace_site_ids = []
268
- robot_trace_site_ids = []
269
- for sid in range(mj_model.nsite):
270
- name = mujoco.mj_id2name(mj_model, mujoco.mjtObj.mjOBJ_SITE, sid)
271
- if name is not None and name.startswith("trace"):
272
- if "object" in name:
273
- object_trace_site_ids.append(sid)
274
- else:
275
- robot_trace_site_ids.append(sid)
276
- config.trace_site_ids = object_trace_site_ids + robot_trace_site_ids
277
-
278
- # setup env params (no DR-schedule dependency on contact guidance)
279
- if config.num_dr == 0:
280
- xy_offset_list = [0.0]
281
- pair_margin_list = [0.0]
282
- else:
283
- xy_offset_list = np.linspace(
284
- config.xy_offset_range[0], config.xy_offset_range[1], config.num_dr
285
- )
286
- pair_margin_list = np.linspace(
287
- config.pair_margin_range[0], config.pair_margin_range[1], config.num_dr
288
- )
289
- env_params_list = []
290
- for _ in range(config.max_num_iterations):
291
- env_params = [
292
- {"xy_offset": xy_offset_list[j], "pair_margin": pair_margin_list[j]}
293
- for j in range(config.num_dr)
294
- ]
295
- env_params_list.append(env_params)
296
- config.env_params_list = env_params_list
297
- _save_config_yaml(config)
298
-
299
- # setup viewer and renderer
300
- run_viewer = setup_viewer(config, mj_model, mj_data)
301
- renderer = setup_renderer(config, mj_model)
302
-
303
- # If viser is enabled and the user wants the recording saved, start
304
- # recording NOW so every per-frame transform update from log_frame is
305
- # captured to a serializable timeline (a single .viser file).
306
- _viser_recording_active = False
307
- if "viser" in (config.viewer or "").lower() and config.save_viser:
308
- try:
309
- from spider.viewers.viser_viewer import start_recording
310
-
311
- start_recording(dt=float(config.render_dt))
312
- _viser_recording_active = True
313
- except Exception as e:
314
- loguru.logger.warning(f"Failed to start viser recording: {e}")
315
-
316
- # Register an atexit + SIGTERM handler so timeout-killed runs still
317
- # flush whatever optimization video / viser timeline was captured.
318
- # Without this, mp4 + .viser are only written if main() reaches its
319
- # end-of-loop save block, which fails-to-converge runs never do.
320
- import atexit
321
- import os as _os
322
- import signal as _signal
323
-
324
- _flushed = {"done": False}
325
-
326
- def _emergency_flush() -> None:
327
- if _flushed["done"]:
328
- return
329
- _flushed["done"] = True
330
- # Save partial mp4 if any frames are queued
331
- try:
332
- if config.save_video and len(images) > 0:
333
- video_path = (
334
- f"{config.output_dir}/visualization_mjwp_fast.mp4"
335
- )
336
- imageio.mimsave(
337
- video_path,
338
- images,
339
- fps=int(1 / config.render_dt),
340
- )
341
- loguru.logger.warning(
342
- f"[emergency flush] Saved partial video ({len(images)} frames) to {video_path}"
343
- )
344
- except Exception as e: # noqa: BLE001
345
- loguru.logger.warning(f"[emergency flush] mp4 save failed: {e}")
346
- # Save viser timeline if recording was started
347
- try:
348
- if _viser_recording_active:
349
- from spider.viewers.viser_viewer import end_recording_and_save
350
-
351
- viser_path = (
352
- f"{config.output_dir}/visualization_mjwp_fast.viser"
353
- )
354
- if end_recording_and_save(viser_path):
355
- loguru.logger.warning(
356
- f"[emergency flush] Saved partial viser timeline to {viser_path}"
357
- )
358
- except Exception as e: # noqa: BLE001
359
- loguru.logger.warning(f"[emergency flush] viser save failed: {e}")
360
-
361
- atexit.register(_emergency_flush)
362
-
363
- def _on_sigterm(_sig, _frame) -> None:
364
- loguru.logger.warning("Received SIGTERM; flushing artifacts.")
365
- _emergency_flush()
366
- # Re-raise default handler so the process actually exits.
367
- _signal.signal(_signal.SIGTERM, _signal.SIG_DFL)
368
- _os.kill(_os.getpid(), _signal.SIGTERM)
369
-
370
- try:
371
- _signal.signal(_signal.SIGTERM, _on_sigterm)
372
- except Exception: # noqa: BLE001
373
- pass
374
-
375
- # setup optimizer — use fast variants
376
- rollout = make_rollout_fn_fast(
377
- step_env,
378
- save_state,
379
- load_state,
380
- get_reward,
381
- get_terminal_reward,
382
- get_terminate,
383
- get_trace,
384
- save_env_params,
385
- load_env_params,
386
- copy_sample_state,
387
- get_qpos,
388
- get_qvel,
389
- )
390
- optimize_once = make_optimize_once_fn_fast(rollout)
391
- optimize = make_optimize_fn_fast(optimize_once)
392
-
393
- ctrls = ctrl_ref[: config.horizon_steps]
394
- info_list = []
395
-
396
- orig_max_iters = config.max_num_iterations
397
- orig_env_params_list = list(config.env_params_list)
398
- orig_noise_scale = config.noise_scale.clone()
399
-
400
- last_snap = None
401
- last_snap_idx = None
402
-
403
- def _save_checkpoint():
404
- return {
405
- "qpos": mj_data.qpos.copy(),
406
- "qvel": mj_data.qvel.copy(),
407
- "ctrl": mj_data.ctrl.copy(),
408
- "time": mj_data.time,
409
- "ctrls": ctrls.clone(),
410
- "env_snap": last_snap,
411
- "env_idx": last_snap_idx,
412
- "images_len": len(images),
413
- "info_list_len": len(info_list),
414
- }
415
-
416
- def _load_checkpoint(cp):
417
- nonlocal ctrls, last_snap, last_snap_idx
418
- mj_data.qpos[:] = cp["qpos"]
419
- mj_data.qvel[:] = cp["qvel"]
420
- mj_data.ctrl[:] = cp["ctrl"]
421
- mj_data.time = cp["time"]
422
- mujoco.mj_forward(mj_model, mj_data)
423
- ctrls = cp["ctrls"].clone()
424
- last_snap = cp["env_snap"]
425
- last_snap_idx = cp["env_idx"]
426
- if last_snap is not None:
427
- _restore_best_state_to_env(env, last_snap, last_snap_idx)
428
- else:
429
- sync_env_mujoco(config, env, mj_data)
430
- del images[cp["images_len"] :]
431
- del info_list[cp["info_list_len"] :]
432
-
433
- def _run_optimize():
434
- nonlocal ctrls
435
- s = int(np.round(mj_data.time / config.sim_dt))
436
- rs = get_slice(ref_data, s + 1, s + config.horizon_steps + 1)
437
- ctrls, inf = optimize(config, env, ctrls, rs)
438
- if len(config.trace_site_ids) > 0:
439
- tr = []
440
- qrh = rs[0]
441
- for h in range(config.horizon_steps):
442
- mj_data_ref.qpos[:] = qrh[h].detach().cpu().numpy()
443
- mujoco.mj_kinematics(mj_model, mj_data_ref)
444
- tr.append(
445
- np.array(
446
- [mj_data_ref.site_xpos[sid] for sid in config.trace_site_ids]
447
- )
448
- )
449
- inf["trace_ref"] = np.stack(tr, axis=0)[None, None, :, :, :]
450
- return s, rs, inf
451
-
452
- def _replay_and_advance(sim_step_val, inf):
453
- nonlocal ctrls
454
- bqp = torch.from_numpy(inf["best_qpos"])
455
- bqv = torch.from_numpy(inf["best_qvel"])
456
- si = {"qpos": [], "qvel": [], "time": [], "ctrl": []}
457
- for i in range(config.ctrl_steps):
458
- mj_data.qpos[:] = bqp[i].numpy()
459
- mj_data.qvel[:] = bqv[i].numpy()
460
- mj_data.ctrl[:] = ctrls[i].detach().cpu().numpy()
461
- mj_data.time = (sim_step_val + i + 1) * config.sim_dt
462
- mujoco.mj_forward(mj_model, mj_data)
463
- if (
464
- config.save_video
465
- and renderer is not None
466
- and i % int(np.round(config.render_dt / config.sim_dt)) == 0
467
- ):
468
- mj_data_ref.qpos[:] = qpos_ref[sim_step_val + i + 1].detach().cpu().numpy()
469
- img = render_image(config, renderer, mj_model, mj_data, mj_data_ref)
470
- images.append(img)
471
- if "rerun" in config.viewer or "viser" in config.viewer:
472
- mj_data_ref.qpos[:] = qpos_ref[sim_step_val + i + 1].detach().cpu().numpy()
473
- mujoco.mj_kinematics(mj_model, mj_data_ref)
474
- log_frame(
475
- mj_data,
476
- sim_time=mj_data.time,
477
- viewer_body_entity_and_ids=config.viewer_body_entity_and_ids,
478
- data_ref=mj_data_ref,
479
- )
480
- # Tick the optimization-process recording so this frame is
481
- # placed at its own timestamp on the timeline.
482
- if "viser" in config.viewer and config.save_viser:
483
- try:
484
- from spider.viewers.viser_viewer import tick_recording
485
-
486
- tick_recording()
487
- except Exception:
488
- pass
489
- si["qpos"].append(mj_data.qpos.copy())
490
- si["qvel"].append(mj_data.qvel.copy())
491
- si["time"].append(mj_data.time)
492
- si["ctrl"].append(mj_data.ctrl.copy())
493
- for k in si:
494
- si[k] = np.stack(si[k], axis=0)
495
- inf.update(si)
496
- _restore_best_state_to_env(env, inf["state_snapshot"], inf["best_idx"])
497
- new_sim_step = int(np.round(mj_data.time / config.sim_dt))
498
- prev_c = ctrls[config.ctrl_steps :]
499
- new_c = ctrl_ref[
500
- new_sim_step + prev_c.shape[0] : new_sim_step
501
- + prev_c.shape[0]
502
- + config.ctrl_steps
503
- ]
504
- ctrls = torch.cat([prev_c, new_c], dim=0)
505
- mj_data_ref.qpos[:] = qpos_ref[new_sim_step].detach().cpu().numpy()
506
- update_viewer(config, viewer, mj_model, mj_data, mj_data_ref, inf)
507
- info_list.append(
508
- {
509
- k: v
510
- for k, v in inf.items()
511
- if k not in ["trace_sample", "state_snapshot"]
512
- }
513
- )
514
- return new_sim_step
515
-
516
- def _full_reset():
517
- ctx.boost_level = 0
518
- ctx.revert_anchor_step = None
519
- ctx.revert_anchor_checkpoint = None
520
- ctx.revert_attempts_from_anchor = 0
521
- ctx.local_retries = 0
522
- _apply_boost(config, 0, orig_max_iters, orig_env_params_list, orig_noise_scale)
523
-
524
- ctx = _LoopCtx()
525
- # sim_step is defined in the loop; keep a module-visible default for the
526
- # `sim_step >= max_sim_steps` terminal check.
527
- sim_step = 0
528
-
529
- t_start = time.perf_counter()
530
- with run_viewer() as viewer:
531
- while ctx.state != _StepState.DONE and viewer.is_running():
532
- t0 = time.perf_counter()
533
- curr_checkpoint = _save_checkpoint()
534
- sim_step, ref_slice, infos = _run_optimize()
535
-
536
- # --- Per-step tracking error for diagnostics ---
537
- bqp_diag = infos.get("best_qpos")
538
- err_info = {"max_pos_error": float("nan"), "max_rot_error": float("nan")}
539
- if bqp_diag is not None:
540
- from spider.optimizers.sampling_fast import check_tracking_error
541
-
542
- bqp_t = (
543
- torch.from_numpy(bqp_diag)
544
- if isinstance(bqp_diag, np.ndarray)
545
- else bqp_diag
546
- ).to(config.device)
547
- # Same alignment as in sampling_fast.optimize: recorded_qpos[i]
548
- # lives at ref index sim_step+1+i = ref_slice[0][i] (NOT [i+1]).
549
- avail = min(config.ctrl_steps, ref_slice[0].shape[0])
550
- ref_qpos_diag = ref_slice[0][:avail].to(config.device)
551
- _, err_info = check_tracking_error(config, bqp_t[:avail], ref_qpos_diag)
552
- infos["step_pos_error"] = np.array([err_info["max_pos_error"]])
553
- infos["step_rot_error"] = np.array([err_info["max_rot_error"]])
554
-
555
- succeeded = bool(infos["succeeded"][0])
556
-
557
- if succeeded:
558
- # --- SUCCESS ---
559
- full_recovery = (
560
- ctx.state == _StepState.RECOVERING
561
- and ctx.revert_anchor_step is not None
562
- and sim_step > ctx.revert_anchor_step
563
- )
564
- if full_recovery:
565
- loguru.logger.info(
566
- "Recovered after {} revert attempts at sim_step={}.",
567
- ctx.revert_attempts_from_anchor,
568
- sim_step,
569
- )
570
- _full_reset()
571
- ctx.state = _StepState.OPTIMIZE
572
-
573
- ctx.local_retries = 0
574
- if ctx.state == _StepState.OPTIMIZE:
575
- ctx.history.append(curr_checkpoint)
576
- last_snap = infos["state_snapshot"]
577
- last_snap_idx = infos["best_idx"]
578
- sim_step = _replay_and_advance(sim_step, infos)
579
-
580
- elif ctx.local_retries < config.max_local_retries:
581
- # --- LOCAL_RETRY ---
582
- ctx.local_retries += 1
583
- loguru.logger.warning(
584
- "Local retry {}/{} at sim_step={} (pos={:.4f}, rot={:.4f}).",
585
- ctx.local_retries,
586
- config.max_local_retries,
587
- sim_step,
588
- err_info["max_pos_error"],
589
- err_info["max_rot_error"],
590
- )
591
- _load_checkpoint(curr_checkpoint)
592
- ctx.boost_level += 1
593
- _apply_boost(
594
- config,
595
- ctx.boost_level,
596
- orig_max_iters,
597
- orig_env_params_list,
598
- orig_noise_scale,
599
- )
600
- continue
601
-
602
- elif (
603
- ctx.state == _StepState.RECOVERING
604
- and ctx.revert_anchor_checkpoint is not None
605
- ):
606
- # --- RE-REVERT to same anchor ---
607
- ctx.revert_attempts_from_anchor += 1
608
- if ctx.revert_attempts_from_anchor > config.max_revert_depth:
609
- loguru.logger.error(
610
- "Exhausted {} revert attempts from anchor sim_step={}.",
611
- config.max_revert_depth,
612
- ctx.revert_anchor_step,
613
- )
614
- _apply_boost(
615
- config,
616
- 0,
617
- orig_max_iters,
618
- orig_env_params_list,
619
- orig_noise_scale,
620
- )
621
- # The accepted prefix already ends at curr_checkpoint.
622
- # Truncating to an older history entry and then appending
623
- # this failed block creates a time gap and a pose jump.
624
- _load_checkpoint(curr_checkpoint)
625
- ctx.state = _StepState.DONE
626
- break
627
- ctx.local_retries = 0
628
- _load_checkpoint(ctx.revert_anchor_checkpoint)
629
- # Reset boost so iters/noise don't compound across anchors —
630
- # each new anchor starts a fresh local-retry sweep at level 0.
631
- ctx.boost_level = 0
632
- _apply_boost(
633
- config,
634
- ctx.boost_level,
635
- orig_max_iters,
636
- orig_env_params_list,
637
- orig_noise_scale,
638
- )
639
- loguru.logger.warning(
640
- "Re-revert attempt {}/{} back to anchor sim_step={}.",
641
- ctx.revert_attempts_from_anchor,
642
- config.max_revert_depth,
643
- ctx.revert_anchor_step,
644
- )
645
- continue
646
-
647
- elif ctx.history and config.max_revert_depth > 0:
648
- # --- FIRST REVERT (from OPTIMIZE state) ---
649
- ctx.local_retries = 0
650
- prev_cp = ctx.history.pop()
651
- ctx.revert_anchor_checkpoint = prev_cp
652
- _load_checkpoint(prev_cp)
653
- ctx.revert_anchor_step = int(np.round(mj_data.time / config.sim_dt))
654
- ctx.revert_attempts_from_anchor = 1
655
- # Reset boost so iters/noise don't compound across anchors —
656
- # each new anchor starts a fresh local-retry sweep at level 0.
657
- ctx.boost_level = 0
658
- _apply_boost(
659
- config,
660
- ctx.boost_level,
661
- orig_max_iters,
662
- orig_env_params_list,
663
- orig_noise_scale,
664
- )
665
- ctx.state = _StepState.RECOVERING
666
- loguru.logger.warning(
667
- "Revert attempt 1/{} from sim_step={} to anchor sim_step={}.",
668
- config.max_revert_depth,
669
- sim_step,
670
- ctx.revert_anchor_step,
671
- )
672
- continue
673
-
674
- else:
675
- # --- DONE (no history to revert to) ---
676
- loguru.logger.error(
677
- "Unrecoverable failure at sim_step={} (no history).",
678
- sim_step,
679
- )
680
- _apply_boost(
681
- config, 0, orig_max_iters, orig_env_params_list, orig_noise_scale
682
- )
683
- _load_checkpoint(curr_checkpoint)
684
- ctx.state = _StepState.DONE
685
- break
686
-
687
- t1 = time.perf_counter()
688
- rtr = config.ctrl_dt / (t1 - t0)
689
- n_att = infos.get("num_attempts", [1])
690
- n_att_v = n_att[0] if hasattr(n_att, "__len__") else n_att
691
- print(
692
- f"Realtime rate: {rtr:.2f}, plan time: {t1 - t0:.4f}s, "
693
- f"sim_steps: {sim_step}/{config.max_sim_steps}, "
694
- f"opt_steps: {infos['opt_steps'][0]}, attempts: {n_att_v}, "
695
- f"state: {ctx.state}, boost: {ctx.boost_level}",
696
- end="\r",
697
- )
698
-
699
- if sim_step >= config.max_sim_steps:
700
- ctx.state = _StepState.DONE
701
- break
702
-
703
- t_end = time.perf_counter()
704
- print(f"Total time: {t_end - t_start:.4f}s")
705
-
706
- if config.save_info and len(info_list) > 0:
707
- info_aggregated = {}
708
- for k in info_list[0]:
709
- if not isinstance(info_list[0][k], np.ndarray):
710
- continue
711
- vals = [info[k] for info in info_list if k in info]
712
- if len(vals) == 0:
713
- continue
714
- if all(v.shape == vals[0].shape for v in vals):
715
- info_aggregated[k] = np.stack(vals, axis=0)
716
- elif k in ("qpos", "qvel", "time", "ctrl"):
717
- info_aggregated[k] = np.concatenate(vals, axis=0)
718
- np.savez(
719
- f"{config.output_dir}/trajectory_mjwp_fast.npz",
720
- **info_aggregated,
721
- )
722
- loguru.logger.info(
723
- f"Saved info to {config.output_dir}/trajectory_mjwp_fast.npz"
724
- )
725
-
726
- if config.save_video and len(images) > 0:
727
- video_path = f"{config.output_dir}/visualization_mjwp_fast.mp4"
728
- imageio.mimsave(
729
- video_path,
730
- images,
731
- fps=int(1 / config.render_dt),
732
- )
733
- loguru.logger.info(f"Saved video to {video_path}")
734
-
735
- errors = None
736
- if info_list:
737
- qpos_traj = np.concatenate([info["qpos"] for info in info_list], axis=0)
738
- # Saved states are post-step, beginning at t=sim_dt.
739
- qpos_ref_np = qpos_ref[1 : qpos_traj.shape[0] + 1].detach().cpu().numpy()
740
- if config.hand_only:
741
- pos, rot, joint = hand_tracking_error(
742
- config, torch.from_numpy(qpos_traj), torch.from_numpy(qpos_ref_np)
743
- )
744
- errors = {
745
- "tracking_domain": "hand",
746
- "hand_base_pos_err": pos.mean().item(),
747
- "hand_base_rot_err": rot.mean().item(),
748
- "max_pos_error": pos.max().item(),
749
- "max_rot_error": rot.max().item(),
750
- "max_joint_error": joint.max().item(),
751
- }
752
- loguru.logger.info(
753
- "Final hand tracking error: base pos={:.4f}, rot={:.4f}, "
754
- "max finger joint={:.4f}",
755
- errors["hand_base_pos_err"],
756
- errors["hand_base_rot_err"],
757
- errors["max_joint_error"],
758
- )
759
- else:
760
- errors = compute_object_tracking_error(
761
- qpos_traj, qpos_ref_np, config.embodiment_type, "mjwp"
762
- )
763
- loguru.logger.info(
764
- "Final object tracking error: pos={:.4f}, quat={:.4f}",
765
- errors["obj_pos_err"],
766
- errors["obj_quat_err"],
767
- )
768
- # Smoothness check: detect jumps between control ticks
769
- if qpos_traj.shape[0] > 1:
770
- diffs = np.linalg.norm(np.diff(qpos_traj, axis=0), axis=1)
771
- max_jump = diffs.max()
772
- max_jump_frame = int(diffs.argmax())
773
- loguru.logger.info(
774
- "Smoothness: max frame-to-frame qpos diff={:.6f} at frame {}/{}",
775
- max_jump,
776
- max_jump_frame,
777
- qpos_traj.shape[0],
778
- )
779
- for tick in range(config.ctrl_steps, qpos_traj.shape[0], config.ctrl_steps):
780
- boundary_jump = np.linalg.norm(qpos_traj[tick] - qpos_traj[tick - 1])
781
- if boundary_jump > 0.5:
782
- loguru.logger.warning(
783
- "Large jump at tick boundary frame {}: diff={:.4f}",
784
- tick,
785
- boundary_jump,
786
- )
787
-
788
- if "viser" in config.viewer and config.save_viser:
789
- viser_path = f"{config.output_dir}/visualization_mjwp_fast.viser"
790
- try:
791
- from spider.viewers.viser_viewer import (
792
- _STATE as _viser_state,
793
- end_recording_and_save,
794
- )
795
-
796
- # If a recording was started, prefer the timeline recording.
797
- # Falls back to a single-frame snapshot when no recording exists.
798
- ok = end_recording_and_save(viser_path)
799
- if ok:
800
- loguru.logger.info(
801
- f"Saved viser optimization recording to {viser_path}"
802
- )
803
- else:
804
- server = _viser_state.server
805
- if server is not None:
806
- Path(viser_path).write_bytes(
807
- server.get_scene_serializer().serialize()
808
- )
809
- loguru.logger.info(f"Saved viser scene to {viser_path}")
810
- except Exception as e:
811
- loguru.logger.warning(f"Failed to save viser scene: {e}")
812
-
813
- if "viser" in config.viewer and config.wait_on_finish:
814
- loguru.logger.info(
815
- "Optimization complete! Keeping Viser server alive. Press Ctrl+C to exit."
816
- )
817
- try:
818
- while True:
819
- time.sleep(1)
820
- except KeyboardInterrupt:
821
- pass
822
-
823
- return errors
824
-
825
-
826
- @hydra.main(version_base=None, config_path="config", config_name="default")
827
- def run_main(cfg: DictConfig) -> None:
828
- """Entry point for Hydra configuration runner."""
829
- config_dict = dict(cfg)
830
-
831
- load_config_path = config_dict.get("load_config_path", "")
832
- if load_config_path:
833
- loaded_config = load_config_yaml(load_config_path)
834
- cli_overrides = _extract_cli_overrides(cfg)
835
- config_dict = {**loaded_config, **cli_overrides}
836
- else:
837
- config_dict = filter_config_fields(config_dict)
838
-
839
- if "noise_scale" in config_dict and config_dict["noise_scale"] is None:
840
- config_dict.pop("noise_scale")
841
-
842
- if "pair_margin_range" in config_dict:
843
- config_dict["pair_margin_range"] = tuple(config_dict["pair_margin_range"])
844
- if "xy_offset_range" in config_dict:
845
- config_dict["xy_offset_range"] = tuple(config_dict["xy_offset_range"])
846
-
847
- config = Config(**config_dict)
848
- main(config)
849
-
850
-
851
- if __name__ == "__main__":
852
- run_main()
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/runtime_source/sha256.json DELETED
@@ -1,9 +0,0 @@
1
- {
2
- "spider/config.py": "dc555763920efb095625791cfe7b85867ff61477a77ac5fb2b0f8334bdaae09b",
3
- "spider/simulators/mjwp.py": "4ea554adbf2ece1bc5cb4a27d5f81e37182a5f3f449db5ed23dbcde17a71e330",
4
- "spider/optimizers/sampling.py": "7dbe311531dde912c586d8626838f78850e75510835bdabcd5e35279f974d1cb",
5
- "spider/optimizers/sampling_fast.py": "6c9249fa9024287fdf115d33a0c8ccec4ecd74cfbb4e978770113937645c07b2",
6
- "examples/run_mjwp_fast.py": "e016907910699d80997c69f9f72eeedd9c420e540bd6945afd5a9a486f09960a",
7
- "spider/io.py": "94bd7204c0310825ba12609860f87988b2981bc360ef69a917b6412f8d2c39c1",
8
- "spider/pipeline.py": "78e5a0afc537929d3172a676f0c54d40fe1eca9b68af290561e666ed001f99f3"
9
- }
 
 
 
 
 
 
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/runtime_source/spider/config.py DELETED
@@ -1,578 +0,0 @@
1
- # Copyright (c) Meta Platforms, Inc. and affiliates.
2
- # All rights reserved.
3
-
4
- # This source code is licensed under the license found in the
5
- # LICENSE file in the root directory of this source tree.
6
-
7
- """Define the configuration for the optimizer.
8
-
9
- Author: Chaoyi Pan
10
- Date: 2025-08-10
11
- """
12
-
13
- from __future__ import annotations
14
-
15
- import json
16
- import os
17
- import random
18
- from dataclasses import dataclass, field, fields
19
-
20
- import loguru
21
- import mujoco
22
- import numpy as np
23
- import torch
24
- from omegaconf import OmegaConf
25
-
26
- import spider
27
- from spider.io import get_processed_data_dir
28
-
29
-
30
- @dataclass
31
- class Config:
32
- # === TASK CONFIGURATION ===
33
- robot_type: str = "sharpa" # "xhand", "inspire", "allegro", "g1"
34
- embodiment_type: str = "bimanual" # "left", "right", "bimanual", "CMU"
35
- task: str = "pick_spoon_bowl"
36
- seed: int = 0
37
- # Derived from explicit task_info.json metadata, never inferred from a
38
- # stationary/identity object pose. Placeholder DOFs remain in the schema.
39
- hand_only: bool = False
40
-
41
- # === SANITY CHECK CONFIGURATION ===
42
- # Holds controls at ctrl_ref[0] for sanity_check_seconds and verifies the
43
- # object does not drift more than the thresholds. Hand-only tasks check
44
- # palms/fingertips with the base thresholds. Set seconds to 0.0 to disable.
45
- sanity_check_seconds: float = 3.0
46
- sanity_check_pos_thresh: float = 0.01
47
- sanity_check_rot_thresh: float = 0.3
48
-
49
- # === DATASET CONFIGURATION ===
50
- dataset_dir: str = f"{spider.ROOT}/../example_datasets"
51
- dataset_name: str = "oakink"
52
- data_id: int = 0
53
- model_path: str = ""
54
- data_path: str = ""
55
- # Optional config loader (used by CLI runner)
56
- load_config_path: str = ""
57
-
58
- # === SIMULATOR CONFIGURATION ===
59
- simulator: str = "mjwp" # "isaac" | "mujoco" | "mjwp" | "mjwp_cons" | "mjwp_eq" | "mjwp_cons_eq" | "kinematic"
60
- device: str = "cuda:0"
61
- # Simulation timing
62
- sim_dt: float = 0.01 # simulation timestep
63
- ctrl_dt: float = 0.4 # control timestep
64
- ref_dt: float = 0.02 # reference data timestep
65
- render_dt: float = 0.02 # rendering timestep
66
- horizon: float = 1.6 # planning horizon
67
- knot_dt: float = 0.4 # knot point spacing
68
- max_sim_steps: int = -1 # maximum simulation steps (-1 for unlimited)
69
- # Simulation constraints
70
- nconmax_per_env: int = 100 # max contacts per environment
71
- njmax_per_env: int = 350 # max joints per environment
72
- # Simulation annealing
73
- num_dyn: int = (
74
- 1 # number of environments for annealing, used for virtual contact constraint
75
- )
76
- # Domain randomization
77
- num_dr: int = (
78
- 1 # number of domain randomization groups, used for domain randomization
79
- )
80
- pair_margin_range: tuple[float, float] = (-0.005, 0.005)
81
- xy_offset_range: tuple[float, float] = (-0.005, 0.005)
82
- perturb_force: float = 0.0
83
- perturb_torque: float = 0.0
84
- contact_guidance: bool = False
85
- # Explicit control initialization/tuning; reference data and joint/control
86
- # limits are unchanged. Gains scale stiffness, with damping scaled by sqrt.
87
- hand_control_mode: str = "reference" # reference | velocity | inverse_dynamics
88
- hand_gain_scale: float = 1.0
89
- object_pos_actuator_names: list[str] = field(
90
- default_factory=lambda: [
91
- "right_object_pos_x",
92
- "right_object_pos_y",
93
- "right_object_pos_z",
94
- "left_object_pos_x",
95
- "left_object_pos_y",
96
- "left_object_pos_z",
97
- ]
98
- )
99
- object_rot_actuator_names: list[str] = field(
100
- default_factory=lambda: [
101
- "right_object_rot_x",
102
- "right_object_rot_y",
103
- "right_object_rot_z",
104
- "left_object_rot_x",
105
- "left_object_rot_y",
106
- "left_object_rot_z",
107
- ]
108
- )
109
- object_action_dims: int = 0
110
- object_actuator_ids: list[int] = field(default_factory=list)
111
- object_actuator_names: list[str] = field(default_factory=list)
112
- init_pos_actuator_gain: float = 10.0
113
- init_pos_actuator_bias: float = 10.0
114
- init_rot_actuator_gain: float = 0.1
115
- init_rot_actuator_bias: float = 0.1
116
- guidance_decay_ratio: float = 0.5
117
- gibbs_sampling: bool = False
118
-
119
- # === OPTIMIZER CONFIGURATION ===
120
- # Sampling parameters
121
- num_samples: int = 2048
122
- temperature: float = 0.3
123
- max_num_iterations: int = 16
124
- improvement_threshold: float = 0.01
125
- improvement_check_steps: int = 1
126
- # Termination parameters
127
- terminate_resample: bool = False
128
- object_pos_threshold: float = 0.1
129
- object_rot_threshold: float = 0.3
130
- max_local_retries: int = 3
131
- max_revert_depth: int = 3
132
- # Preserve humanoid defaults. Hand-only runs should explicitly set 0.02 m /
133
- # 0.2 rad; these base thresholds remain independent of object thresholds.
134
- base_pos_threshold: float = 0.5
135
- base_rot_threshold: float = 0.4
136
- # Compilation
137
- use_torch_compile: bool = True # use torch.compile for acceleration
138
- # Noise scheduling
139
- first_ctrl_noise_scale: float = 0.5
140
- last_ctrl_noise_scale: float = 1.0
141
- final_noise_scale: float = 0.1
142
- exploit_ratio: float = 0.01
143
- exploit_noise_scale: float = 0.01
144
- # Noise scaling by component
145
- joint_noise_scale: float = 0.15
146
- pos_noise_scale: float = 0.03
147
- rot_noise_scale: float = 0.03
148
- # Reward mode
149
- use_rl_reward: bool = False # use dexmachina RL training reward formulation
150
- # Reward scaling
151
- base_pos_rew_scale: float = 1.0
152
- base_rot_rew_scale: float = 0.3
153
- joint_rew_scale: float = 0.003
154
- pos_rew_scale: float = 1.0
155
- rot_rew_scale: float = 0.3
156
- vel_rew_scale: float = 0.0001
157
- terminal_rew_scale: float = 1.0
158
- contact_rew_scale: float = 0.0
159
-
160
- # === VISUALIZATION CONFIGURATION ===
161
- show_viewer: bool = True
162
- viewer: str = "mujoco" # "mujoco" | "rerun" | "viser" | "isaac"
163
- wait_on_finish: bool = True # block after optimization to keep viewer alive
164
- rerun_spawn: bool = False
165
- save_video: bool = True
166
- save_info: bool = True
167
- save_rerun: bool = False
168
- save_viser: bool = False
169
- save_metrics: bool = True
170
- save_config: bool = True
171
-
172
- # === TRACE RECORDING ===
173
- trace_dt: float = 1 / 50.0
174
- num_trace_uniform_samples: int = 4
175
- num_trace_topk_samples: int = 2
176
- trace_site_ids: list = field(default_factory=list)
177
-
178
- # === CONTACT GUIDANCE (DERIVED) ===
179
- contact_order: list = field(default_factory=list)
180
- hand_contact_site_ids: list = field(default_factory=list)
181
- right_contact_indices: list = field(default_factory=list)
182
- left_contact_indices: list = field(default_factory=list)
183
- right_pos_ctrl_ids: list = field(default_factory=list)
184
- left_pos_ctrl_ids: list = field(default_factory=list)
185
- contact_len: int = 0
186
-
187
- # === AUTOMATICALLY SET PROPERTIES ===
188
- # Computed timesteps
189
- horizon_steps: int = -1
190
- knot_steps: int = -1
191
- ref_steps: int = -1
192
- ctrl_steps: int = -1
193
- # Model dimensions
194
- nq_obj: int = -1 # object DOF
195
- nq: int = -1 # total position DOF
196
- nv: int = -1 # total velocity DOF
197
- nu: int = -1 # total control DOF
198
- npair: int = -1 # total pair DOF
199
- # Computed tensors
200
- noise_scale: torch.Tensor = field(default_factory=lambda: torch.ones(1))
201
- beta_traj: float = -1.0
202
- # Runtime state
203
- env_params_list: list = field(default_factory=list)
204
- viewer_body_entity_and_ids: list = field(default_factory=list)
205
- output_dir: str = ""
206
-
207
-
208
- def resolve_object_actuator_ids(
209
- model: mujoco.MjModel,
210
- desired_names: list[str],
211
- object_action_dims: int,
212
- ) -> tuple[list[int], list[str]]:
213
- """Resolve object actuator ids by name, with a fallback to last N actuators."""
214
- resolved_ids: list[int] = []
215
- resolved_names: list[str] = []
216
- for name in desired_names:
217
- aid = mujoco.mj_name2id(model, mujoco.mjtObj.mjOBJ_ACTUATOR, name)
218
- if aid != -1:
219
- resolved_ids.append(int(aid))
220
- resolved_names.append(name)
221
- if resolved_ids:
222
- if len(resolved_ids) != len(desired_names):
223
- loguru.logger.info(
224
- "Resolved {} / {} object actuators by name.",
225
- len(resolved_ids),
226
- len(desired_names),
227
- )
228
- return resolved_ids, resolved_names
229
-
230
- obj_start = max(model.nu - max(object_action_dims, 0), 0)
231
- fallback_ids = list(range(obj_start, model.nu))
232
- fallback_names = [
233
- mujoco.mj_id2name(model, mujoco.mjtObj.mjOBJ_ACTUATOR, int(aid))
234
- for aid in fallback_ids
235
- ]
236
- loguru.logger.warning(
237
- "No named object actuators found; falling back to last {} actuators.",
238
- len(fallback_ids),
239
- )
240
- return fallback_ids, fallback_names
241
-
242
-
243
- def load_config_yaml(path: str) -> dict:
244
- """Load a config YAML into a plain dict, resolving OmegaConf types."""
245
- if not path:
246
- return {}
247
- abs_path = os.path.abspath(os.path.expanduser(path))
248
- if not os.path.isfile(abs_path):
249
- raise FileNotFoundError(f"Config file not found: {abs_path}")
250
- cfg = OmegaConf.load(abs_path)
251
- cfg_dict = OmegaConf.to_container(cfg, resolve=True)
252
- if not isinstance(cfg_dict, dict):
253
- raise ValueError(f"Config at {abs_path} did not resolve to a mapping.")
254
- cfg_dict.pop("hydra", None)
255
- return cfg_dict
256
-
257
-
258
- def filter_config_fields(config_dict: dict) -> dict:
259
- """Filter config dict keys to those defined in Config."""
260
- allowed = {field.name for field in fields(Config)}
261
- return {key: value for key, value in config_dict.items() if key in allowed}
262
-
263
-
264
- def build_hand_contact_site_ids(
265
- mj_model: mujoco.MjModel, embodiment_type: str
266
- ) -> tuple[list[tuple[str, str]], list[int | None]]:
267
- contact_order = []
268
- if embodiment_type in ["bimanual", "right"]:
269
- contact_order.extend(
270
- [
271
- ("right", "thumb"),
272
- ("right", "index"),
273
- ("right", "middle"),
274
- ("right", "ring"),
275
- ("right", "pinky"),
276
- ]
277
- )
278
- if embodiment_type in ["bimanual", "left"]:
279
- contact_order.extend(
280
- [
281
- ("left", "thumb"),
282
- ("left", "index"),
283
- ("left", "middle"),
284
- ("left", "ring"),
285
- ("left", "pinky"),
286
- ]
287
- )
288
-
289
- site_ids: list[int | None] = [None] * len(contact_order)
290
- for sid in range(mj_model.nsite):
291
- name = mujoco.mj_id2name(mj_model, mujoco.mjtObj.mjOBJ_SITE, sid)
292
- if name is None:
293
- continue
294
- name_l = name.lower()
295
- if "track" not in name_l or "hand" not in name_l:
296
- continue
297
- for idx, (side, finger) in enumerate(contact_order):
298
- if side in name_l and finger in name_l:
299
- if site_ids[idx] is None:
300
- site_ids[idx] = sid
301
- break
302
-
303
- missing = [i for i, sid in enumerate(site_ids) if sid is None]
304
- if missing:
305
- loguru.logger.warning(
306
- "Missing {} hand contact sites for guidance; indices: {}",
307
- len(missing),
308
- missing,
309
- )
310
- return contact_order, site_ids
311
-
312
-
313
- def get_object_pos_ctrl_indices(config: Config) -> tuple[list[int], list[int]]:
314
- right_ids: list[int] = []
315
- left_ids: list[int] = []
316
- if config.object_actuator_ids and config.object_actuator_names:
317
- for aid, name in zip(
318
- config.object_actuator_ids, config.object_actuator_names, strict=False
319
- ):
320
- name_l = (name or "").lower()
321
- if "_pos_" not in name_l:
322
- continue
323
- if "right" in name_l:
324
- right_ids.append(int(aid))
325
- elif "left" in name_l:
326
- left_ids.append(int(aid))
327
-
328
- if right_ids or left_ids:
329
- return right_ids, left_ids
330
-
331
- obj_dims = int(config.object_action_dims) if config.object_action_dims > 0 else 0
332
- if obj_dims == 0:
333
- obj_dims = 12 if config.embodiment_type == "bimanual" else 6
334
- start = max(int(config.nu) - obj_dims, 0)
335
- if config.embodiment_type == "bimanual" and obj_dims >= 12:
336
- right_ids = list(range(start, start + 3))
337
- left_ids = list(range(start + 3, start + 6))
338
- else:
339
- if config.embodiment_type == "right":
340
- right_ids = list(range(start, start + 3))
341
- elif config.embodiment_type == "left":
342
- left_ids = list(range(start, start + 3))
343
- else:
344
- right_ids = list(range(start, start + 3))
345
- return right_ids, left_ids
346
-
347
-
348
- def get_noise_scale(config: Config) -> torch.Tensor:
349
- """Get the noise scale for sampling.
350
-
351
- Args:
352
- config: Config
353
-
354
- Returns:
355
- Noise scale, shape (num_samples, knot_steps, nu)
356
- """
357
- noise_scale = torch.logspace(
358
- start=torch.log10(torch.tensor(config.first_ctrl_noise_scale)),
359
- end=torch.log10(torch.tensor(config.last_ctrl_noise_scale)),
360
- steps=int(round(config.horizon / config.knot_dt)),
361
- device=config.device,
362
- base=10,
363
- )[None, :, None] # Shape: (1, num_knot_steps, 1)
364
- noise_scale = noise_scale.repeat(1, 1, config.nu)
365
- if config.embodiment_type in ["bimanual", "right", "left"]:
366
- object_action_dims = max(int(config.object_action_dims), 0)
367
- robot_nu = max(int(config.nu - object_action_dims), 0)
368
- noise_scale[:, :, :3] *= config.pos_noise_scale
369
- noise_scale[:, :, 3:6] *= config.rot_noise_scale
370
- if config.embodiment_type == "bimanual":
371
- half_dof = robot_nu // 2
372
- noise_scale[:, :, 6:half_dof] *= config.joint_noise_scale
373
- noise_scale[:, :, half_dof : half_dof + 3] *= config.pos_noise_scale
374
- noise_scale[:, :, half_dof + 3 : half_dof + 6] *= config.rot_noise_scale
375
- noise_scale[:, :, half_dof + 6 : robot_nu] *= config.joint_noise_scale
376
- elif config.embodiment_type in ["right", "left"]:
377
- noise_scale[:, :, 6:robot_nu] *= config.joint_noise_scale
378
- else:
379
- noise_scale *= config.joint_noise_scale
380
- if config.contact_guidance and config.object_actuator_ids:
381
- object_ids = torch.as_tensor(
382
- config.object_actuator_ids, device=config.device, dtype=torch.long
383
- )
384
- noise_scale[:, :, object_ids] *= 0.0
385
- # repeat to match num_samples; same samples used across DR groups
386
- noise_scale = noise_scale.repeat(config.num_samples, 1, 1)
387
- # set first sample to 0
388
- noise_scale[0] *= 0.0
389
- # set last few samples to exploit_noise_scale
390
- num_exploit_samples = int(config.num_samples * config.exploit_ratio)
391
- if num_exploit_samples > 0:
392
- noise_scale[-num_exploit_samples:] *= config.exploit_noise_scale
393
- return noise_scale
394
-
395
-
396
- def compute_steps(config: Config):
397
- # make sure every dt can be divided by sim_dt
398
- config.horizon_steps = int(np.round(config.horizon / config.sim_dt))
399
- config.knot_steps = int(np.round(config.knot_dt / config.sim_dt))
400
- config.ref_steps = int(np.round(config.ref_dt / config.sim_dt))
401
- config.ctrl_steps = int(np.round(config.ctrl_dt / config.sim_dt))
402
- assert np.isclose(
403
- config.horizon - config.horizon_steps * config.sim_dt, 0, atol=1e-5
404
- ), "horizon must be divisible by sim_dt"
405
- assert np.isclose(
406
- config.ctrl_dt - config.ctrl_steps * config.sim_dt, 0, atol=1e-5
407
- ), "ctrl_dt must be divisible by sim_dt"
408
- assert np.isclose(
409
- config.knot_dt - config.knot_steps * config.sim_dt, 0, atol=1e-5
410
- ), "knot_dt must be divisible by sim_dt"
411
- return config
412
-
413
-
414
- def compute_noise_schedule(config: Config) -> Config:
415
- config.noise_scale = get_noise_scale(config)
416
- if config.max_num_iterations > 0:
417
- config.beta_traj = config.final_noise_scale ** (1 / config.max_num_iterations)
418
- else:
419
- config.beta_traj = 1.0
420
- return config
421
-
422
-
423
- def process_config(config: Config):
424
- """Process the configuration to fill in the missing fields."""
425
- if config.hand_control_mode not in ("reference", "velocity", "inverse_dynamics"):
426
- raise ValueError(f"Unknown hand_control_mode: {config.hand_control_mode}")
427
- if not np.isfinite(config.hand_gain_scale) or config.hand_gain_scale <= 0:
428
- raise ValueError("hand_gain_scale must be positive and finite")
429
- config = compute_steps(config)
430
- trace_steps_tmp = int(np.round(config.trace_dt / config.sim_dt))
431
- assert np.isclose(
432
- config.trace_dt - trace_steps_tmp * config.sim_dt, 0, atol=1e-3
433
- ), "trace_dt must be divisible by sim_dt"
434
-
435
- # Set object DOF based on hand type
436
- if config.contact_guidance:
437
- config.nq_obj = {
438
- "bimanual": 12,
439
- "right": 6,
440
- "left": 6,
441
- }.get(config.embodiment_type, 0)
442
- else:
443
- config.nq_obj = {
444
- "bimanual": 14,
445
- "right": 7,
446
- "left": 7,
447
- }.get(config.embodiment_type, 0)
448
-
449
- # resolve processed directories for this trial
450
- dataset_dir_abs = os.path.abspath(config.dataset_dir)
451
- processed_dir_robot = get_processed_data_dir(
452
- dataset_dir=dataset_dir_abs,
453
- dataset_name=config.dataset_name,
454
- robot_type=config.robot_type,
455
- embodiment_type=config.embodiment_type,
456
- task=config.task,
457
- data_id=config.data_id,
458
- )
459
- task_info_path = f"{processed_dir_robot}/../task_info.json"
460
- try:
461
- with open(task_info_path, encoding="utf-8") as f:
462
- task_info = json.load(f)
463
- except FileNotFoundError:
464
- loguru.logger.warning(
465
- f"task_info.json not found at {task_info_path}, using default values"
466
- )
467
- task_info = {}
468
- config.hand_only = task_info.get("hand_only", False)
469
- if not isinstance(config.hand_only, bool):
470
- raise ValueError("task_info.json hand_only must be a boolean")
471
- if config.hand_only:
472
- if config.embodiment_type not in ("left", "right", "bimanual"):
473
- raise ValueError("hand_only requires a hand embodiment")
474
- if config.contact_guidance or config.contact_rew_scale > 0:
475
- raise ValueError("hand_only tasks cannot use object contact guidance/rewards")
476
- for name in ("base_pos_threshold", "base_rot_threshold"):
477
- value = getattr(config, name)
478
- if not np.isfinite(value) or value <= 0:
479
- raise ValueError(f"hand_only requires a positive finite {name}")
480
- # model and data within processed directory (scene_eq.xml support for annealing over equality constraints)
481
- if config.contact_guidance:
482
- scene_xml = "scene_act.xml"
483
- else:
484
- scene_xml = "scene.xml" if config.num_dyn == 1 else "scene_eq.xml"
485
- config.model_path = f"{processed_dir_robot}/../{scene_xml}"
486
- # default to MJWP retargeted trajectory if available
487
- if config.contact_guidance:
488
- config.data_path = f"{processed_dir_robot}/trajectory_kinematic_act.npz"
489
- else:
490
- config.data_path = f"{processed_dir_robot}/trajectory_kinematic.npz"
491
-
492
- # get model data
493
- if config.simulator == "mjwp":
494
- model = mujoco.MjModel.from_xml_path(config.model_path)
495
- if config.hand_only:
496
- for side in ("left", "right"):
497
- body = mujoco.mj_name2id(
498
- model, mujoco.mjtObj.mjOBJ_BODY, f"{side}_object"
499
- )
500
- if body < 0:
501
- continue
502
- # Geometry on a descendant also makes this a physical object.
503
- for geom_body in model.geom_bodyid:
504
- ancestor = int(geom_body)
505
- while ancestor and ancestor != body:
506
- ancestor = int(model.body_parentid[ancestor])
507
- if ancestor == body:
508
- raise ValueError(
509
- "hand_only metadata contradicts physical object geometry"
510
- )
511
- config.nq = model.nq
512
- config.nv = model.nv
513
- config.nu = model.nu
514
- config.npair = model.npair
515
- if config.contact_guidance:
516
- if config.object_action_dims <= 0:
517
- config.object_action_dims = (
518
- 12 if config.embodiment_type == "bimanual" else 6
519
- )
520
- desired_names = (
521
- config.object_pos_actuator_names + config.object_rot_actuator_names
522
- )
523
- object_ids, object_names = resolve_object_actuator_ids(
524
- model,
525
- desired_names,
526
- config.object_action_dims,
527
- )
528
- config.object_actuator_ids = object_ids
529
- config.object_actuator_names = object_names
530
- config.right_pos_ctrl_ids, config.left_pos_ctrl_ids = (
531
- get_object_pos_ctrl_indices(config)
532
- )
533
- config.contact_order, config.hand_contact_site_ids = (
534
- build_hand_contact_site_ids(model, config.embodiment_type)
535
- )
536
- config.right_contact_indices = [
537
- idx
538
- for idx, (side, finger) in enumerate(config.contact_order)
539
- if (side == "right") and (finger in ["thumb"])
540
- ]
541
- config.left_contact_indices = [
542
- idx
543
- for idx, (side, finger) in enumerate(config.contact_order)
544
- if side == "left" and (finger in ["thumb"])
545
- ]
546
-
547
- # get noise scale
548
- config = compute_noise_schedule(config)
549
-
550
- # output dir: write artifacts alongside the trial
551
- config.output_dir = processed_dir_robot
552
- os.makedirs(config.output_dir, exist_ok=True)
553
-
554
- if "ref_dt" in task_info:
555
- config.ref_dt = task_info["ref_dt"]
556
- config.ref_steps = int(np.round(config.ref_dt / config.sim_dt))
557
- loguru.logger.info(f"overriding ref_dt: {config.ref_dt} from task_info.json")
558
-
559
- # override contact site ids
560
- if config.contact_rew_scale > 0.0:
561
- if "contact_site_ids" in task_info:
562
- config.contact_site_ids = task_info["contact_site_ids"]
563
- loguru.logger.info(
564
- f"overriding contact_site_ids: {config.contact_site_ids} from task_info.json"
565
- )
566
- else:
567
- raise ValueError(
568
- "contact_site_ids not found in task_info.json while contact_rew_scale > 0.0"
569
- )
570
-
571
- # set seed
572
- np.random.seed(config.seed)
573
- torch.manual_seed(config.seed)
574
- torch.cuda.manual_seed(config.seed)
575
- torch.cuda.manual_seed_all(config.seed)
576
- random.seed(config.seed)
577
-
578
- return config
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
agentic_pipeline/2026-10-06-full-retarget/attempts/egoverse_trial2/runtime_source/spider/io.py DELETED
@@ -1,202 +0,0 @@
1
- # Copyright (c) Meta Platforms, Inc. and affiliates.
2
- # All rights reserved.
3
-
4
- # This source code is licensed under the license found in the
5
- # LICENSE file in the root directory of this source tree.
6
-
7
- """Define functions to load and save the data.
8
-
9
- Author: Chaoyi Pan
10
- Date: 2025-08-10
11
- """
12
-
13
- from __future__ import annotations
14
-
15
- import os
16
- from typing import TYPE_CHECKING
17
-
18
- import loguru
19
- import numpy as np
20
- import torch
21
-
22
-
23
- if TYPE_CHECKING:
24
- from spider.config import Config
25
-
26
-
27
- def load_data(
28
- config: Config,
29
- data_path: str = "../datasets/inspire/bimanual_wipe.npz",
30
- ) -> tuple[torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor]:
31
- """Load trajectory data from NPZ file."""
32
- with np.load(data_path) as archive:
33
- raw_data = dict(archive)
34
- qpos_ref = raw_data["qpos"]
35
- qvel_ref = raw_data["qvel"]
36
- if "frequency" in raw_data:
37
- frequency = float(raw_data["frequency"])
38
- if not np.isfinite(frequency) or frequency <= 0:
39
- raise ValueError("Trajectory frequency must be positive and finite")
40
- config.ref_dt = 1.0 / frequency
41
- if not len(qpos_ref):
42
- raise ValueError("Cannot load an empty reference trajectory")
43
- if config.ref_dt <= 0 or config.sim_dt <= 0:
44
- raise ValueError("Reference and simulation timesteps must be positive")
45
- # if contact_rew_scale > 0.0, we need to make sure both contact and contact_pos are provided.
46
- if config.contact_rew_scale > 0.0:
47
- if "contact" not in raw_data:
48
- raise ValueError("contact data not found while contact_rew_scale > 0.0")
49
- if "contact_pos" not in raw_data:
50
- raise ValueError("contact_pos data not found while contact_rew_scale > 0.0")
51
- try:
52
- contact = raw_data["contact"]
53
- except:
54
- contact = np.zeros((qpos_ref.shape[0], 10))
55
- loguru.logger.warning("contact data not found")
56
- try:
57
- contact_pos = raw_data["contact_pos"]
58
- except:
59
- contact_pos = np.zeros((qpos_ref.shape[0], 10, 3))
60
- loguru.logger.warning("contact_pos data not found")
61
- if "ctrl" in raw_data:
62
- ctrl_ref = raw_data["ctrl"]
63
- else:
64
- # TODO: disable automatic reference control generation, instead, move it to preprocess
65
- # Fallback if 'ctrl' is not in the data file
66
- loguru.logger.warning(
67
- "ctrl data not found, using 'qpos' as a initial guess for control."
68
- )
69
- if config.embodiment_type in ["bimanual", "right", "left"]:
70
- ctrl_ref = qpos_ref[:, : -config.nq_obj]
71
- elif config.embodiment_type in ["CMU", "DanceDB", "humanoid"]:
72
- ctrl_ref = qpos_ref[:, 7:]
73
- else:
74
- raise ValueError(f"Invalid embodiment_type: {config.embodiment_type}")
75
- # move to device
76
- qpos_ref_torch = torch.from_numpy(qpos_ref).to(config.device).to(torch.float32)
77
- qvel_ref_torch = torch.from_numpy(qvel_ref).to(config.device).to(torch.float32)
78
- ctrl_ref_torch = torch.from_numpy(ctrl_ref).to(config.device).to(torch.float32)
79
- contact_ref_torch = torch.from_numpy(contact).to(config.device).to(torch.float32)
80
- contact_pos_ref_torch = (
81
- torch.from_numpy(contact_pos).to(config.device).to(torch.float32)
82
- )
83
- # Sample at actual simulation timestamps. Integer repeat/stride factors
84
- # distort 30 Hz and 120 Hz recordings when the simulator runs at 100 Hz.
85
- duration = (len(qpos_ref) - 1) * config.ref_dt
86
- sample_count = int(np.floor(duration / config.sim_dt + 1e-9)) + 1
87
- positions = torch.arange(sample_count, device=config.device, dtype=torch.float64)
88
- positions *= config.sim_dt / config.ref_dt
89
- positions.clamp_(max=len(qpos_ref) - 1)
90
- lower = positions.floor().long()
91
- upper = (lower + 1).clamp(max=len(qpos_ref) - 1)
92
- fraction = (positions - lower).to(torch.float32)
93
-
94
- def resample(values):
95
- if len(values) != len(qpos_ref):
96
- raise ValueError(
97
- "Reference pose, velocity, control and contact frame counts must match"
98
- )
99
- weight = fraction.reshape((-1,) + (1,) * (values.ndim - 1))
100
- return values[lower] * (1 - weight) + values[upper] * weight
101
-
102
- qpos_ref_interp = resample(qpos_ref_torch)
103
- qvel_ref_interp = resample(qvel_ref_torch)
104
- ctrl_ref_interp = resample(ctrl_ref_torch)
105
- contact_ref_interp = resample(contact_ref_torch)
106
- contact_pos_ref_interp = resample(contact_pos_ref_torch)
107
-
108
- # Quaternion coordinates need shortest-path interpolation rather than
109
- # component interpolation (q and -q denote the same rotation).
110
- if config.embodiment_type in ("right", "left", "bimanual"):
111
- quaternion_starts = (
112
- [
113
- qpos_ref.shape[1] - config.nq_obj + offset + 3
114
- for offset in range(0, config.nq_obj, 7)
115
- ]
116
- if config.nq_obj in (7, 14)
117
- else []
118
- )
119
- else:
120
- quaternion_starts = [3]
121
- for start in quaternion_starts:
122
- left = qpos_ref_torch[lower, start : start + 4]
123
- right = qpos_ref_torch[upper, start : start + 4]
124
- identity = left.new_tensor([1, 0, 0, 0])
125
- left = torch.where(left.norm(dim=-1, keepdim=True) < 1e-12, identity, left)
126
- right = torch.where(right.norm(dim=-1, keepdim=True) < 1e-12, identity, right)
127
- left = torch.nn.functional.normalize(left, dim=-1)
128
- right = torch.nn.functional.normalize(right, dim=-1)
129
- dot = (left * right).sum(-1, keepdim=True)
130
- right = torch.where(dot < 0, -right, right)
131
- dot = dot.abs().clamp(max=1)
132
- angle = torch.acos(dot)
133
- denominator = torch.sin(angle).clamp(min=1e-7)
134
- weight = fraction[:, None]
135
- spherical = (
136
- torch.sin((1 - weight) * angle) * left + torch.sin(weight * angle) * right
137
- ) / denominator
138
- linear = (1 - weight) * left + weight * right
139
- quat = torch.where(dot > 0.9995, linear, spherical)
140
- qpos_ref_interp[:, start : start + 4] = torch.nn.functional.normalize(
141
- quat, dim=-1
142
- )
143
- # repeat the last frame with extra config.horizon_steps
144
- for _ in range(config.horizon_steps + config.ctrl_steps):
145
- qpos_ref_interp = torch.cat([qpos_ref_interp, qpos_ref_interp[-1:]], dim=0)
146
- qvel_ref_interp = torch.cat([qvel_ref_interp, qvel_ref_interp[-1:]], dim=0)
147
- ctrl_ref_interp = torch.cat([ctrl_ref_interp, ctrl_ref_interp[-1:]], dim=0)
148
- contact_ref_interp = torch.cat(
149
- [contact_ref_interp, contact_ref_interp[-1:]], dim=0
150
- )
151
- contact_pos_ref_interp = torch.cat(
152
- [contact_pos_ref_interp, contact_pos_ref_interp[-1:]], dim=0
153
- )
154
- return (
155
- qpos_ref_interp,
156
- qvel_ref_interp,
157
- ctrl_ref_interp,
158
- contact_ref_interp,
159
- contact_pos_ref_interp,
160
- )
161
-
162
-
163
- def get_processed_data_dir(
164
- dataset_dir: str,
165
- dataset_name: str,
166
- robot_type: str,
167
- embodiment_type: str,
168
- task: str,
169
- data_id: int,
170
- ) -> str:
171
- """Get the data directory from the data path."""
172
- return f"{dataset_dir}/processed/{dataset_name}/{robot_type}/{embodiment_type}/{task}/{data_id}"
173
-
174
-
175
- def get_all_tasks(
176
- dataset_dir: str,
177
- dataset_name: str,
178
- robot_type: str,
179
- embodiment_type: str,
180
- ) -> list[str]:
181
- """Get all tasks from the data path."""
182
- return [
183
- d
184
- for d in os.listdir(
185
- f"{dataset_dir}/processed/{dataset_name}/{robot_type}/{embodiment_type}"
186
- )
187
- if os.path.isdir(
188
- f"{dataset_dir}/processed/{dataset_name}/{robot_type}/{embodiment_type}/{d}"
189
- )
190
- ]
191
-
192
-
193
- def get_mesh_dir(dataset_dir: str, dataset_name: str, object_name: str) -> str:
194
- """Get the mesh directory from the data path."""
195
- return f"{dataset_dir}/processed/{dataset_name}/assets/objects/{object_name}"
196
-
197
-
198
- def resolve_dataset_path(dataset_dir: str, value: str | None) -> str | None:
199
- """Resolve metadata paths, including absolute paths from older processors."""
200
- if value is None:
201
- return None
202
- return os.path.join(os.path.abspath(os.path.expanduser(dataset_dir)), value)