Pepijn 8455efc474 feat(umi): simplify to derive_state_from_action and cam0-only ...

- Remove fix_dataset.py (user fixes dataset at source)
- evaluate.py: replace observation.pose/joints with observation.state
  (8D, derived from action during training, from FK at inference)
- evaluate.py: remove cam1 — training uses only cam0
- docs: rewrite workflow around derive_state_from_action=true,
  updated recompute-stats and training commands with
  relative_exclude_joints for gripper dims

Made-with: Cursor

2026-04-02 15:02:20 +02:00

..

backward_compatibility

refactor(dataset): split LeRobotDataset into DatasetReader & DatasetWriter (+ API cleanup) (#3180)

2026-03-26 19:09:25 +01:00

dataset

refactor(dataset): split LeRobotDataset into DatasetReader & DatasetWriter (+ API cleanup) (#3180)

2026-03-26 19:09:25 +01:00

lekiwi

refactor(dataset): split LeRobotDataset into DatasetReader & DatasetWriter (+ API cleanup) (#3180)

2026-03-26 19:09:25 +01:00

phone_to_so100

refactor(dataset): split LeRobotDataset into DatasetReader & DatasetWriter (+ API cleanup) (#3180)

2026-03-26 19:09:25 +01:00

port_datasets

refactor(dataset): modular files (#3171)

2026-03-15 23:58:09 -07:00

rtc

feat(policies): add relative action support for pi0, pi0.5, and pi0_fast (#2970)

2026-04-01 12:59:12 +02:00

so100_to_so100_EE

refactor(dataset): split LeRobotDataset into DatasetReader & DatasetWriter (+ API cleanup) (#3180)

2026-03-26 19:09:25 +01:00

training

refactor(dataset): modular files (#3171)

2026-03-15 23:58:09 -07:00

tutorial

refactor(dataset): modular files (#3171)

2026-03-15 23:58:09 -07:00

umi_pi0_relative_ee

feat(umi): simplify to derive_state_from_action and cam0-only

2026-04-02 15:02:20 +02:00