2 days ago
-
Cap trailing hold padding at 50 frames per episode; update README
-
Re-merge data-fixed (1517 eps, 646079 frames): MP tasks re-targeted for soft Dex3 gains with failed replays removed; trailing hold padding on episodes whose plain replay failed; stats recomputed; README updated
3 days ago
-
MP tasks: waist state follows the AMO convention (last commanded value); README: teleop vs MP conventions
-
Recompute 433 stale per-episode stats rows (episodes with appended tail/hold frames: BendHandover 90, BendPickAndPlace 48, BendPick 97, LocomotionPickBetweenTables 99, XMovePick 99)
-
Rebuild with updated G1WholebodyBendHandoverTeleop-v0 (full teleop tails: 40775 frames); recomputed stats
-
Document 36-d action / 32-d state layout; add joint names to info.json
4 days ago
-
Merge 17 SIMPLE data-fixed tasks into one LeRobot v2.1 dataset (recomputed stats)
