All pull requests

Pull requests

• feat(openarm): control Damiao motors via motorbridge + macOS PCAN support

  #4833·CarolinePascal opened Oct 1, 2026Collaborator·

  ·
  1

• feat(teleoperators): mirror option for the SO leaders

  documentationrobotstests
  #4832·stepankandrushin opened Oct 1, 2026·4 of 5 tasks·

  ·
  1

• fix(datasets): accept channel-first frames for (H, W, C) image features

  datasettests
  #4831·Zhuoxi2000 opened Oct 1, 2026·2 of 5 tasks·

  ·

• feat(robots): add bimanual YAM with motorbridge and gravity feed-forward

  documentationexamplesrobotstests
  #4829·pkooij opened Oct 1, 2026Member·

  ·
  1

• feat(rollout): direct VLM control with inspect-robots native tools

  #4828·pkooij opened Oct 1, 2026Member·

  ·

• fix(xvla): compute select_action chunks through predict_action_chunk

  policiestests
  #4827·shoumikhin opened Oct 1, 2026Contributor·

  ·
  2

• feat(calibration): lerobot-calibrate --auto for the SO-101 leader and follower

  documentationrobotstests
  #4826·stepankandrushin opened Oct 1, 2026·4 of 5 tasks·

  ·
  1

• feat(rollout): show loop rate, inference time and memory while the robot runs

  tests
  #4824·shoumikhin opened Oct 1, 2026Contributor·

  ·
  3

• fix(rewards): apply SARM stage_loss_weight to the total loss

  tests
  #4822·Jxy-yxJ opened Sep 30, 2026·

  ·

• Let the rollout control loop take NumPy actions from the inference engine

  configurationdocumentationevaluationpoliciesprocessortests
  #4820·shoumikhin opened Sep 30, 2026Contributor·

  ·
  2

• fix(annotations): store VQA bbox and keypoint coordinates as [0, 1] image fractions

  documentationtests
  #4819·Maximellerbach opened Sep 30, 2026Member·3 of 5 tasks·

  ·
  2

• feat(policies): accept the starting noise in predict_action_chunk for GR00T, pi0 and pi0.5

  policiestests
  #4817·shoumikhin opened Sep 30, 2026Contributor·3 of 5 tasks·

  ·
  2

• feat(policies): accept the starting noise in predict_action_chunk for Multi-task DiT and FastWAM

  policiestests
  #4816·shoumikhin opened Sep 30, 2026Contributor·3 of 5 tasks·

  ·
  2

• feat(policies): accept the starting noise of the action and video streams in FLUX3 and LingBot-VA

  policiestests
  #4815·shoumikhin opened Sep 30, 2026Contributor·3 of 5 tasks·

  ·
  2

• fix(policies): use the noise passed to predict_action_chunk in X-VLA, LaWAM and VLA-JEPA

  policiestests
  #4814·shoumikhin opened Sep 30, 2026Contributor·3 of 5 tasks·

  ·
  2

• feat(molmoact2): accept the starting noise in predict_action_chunk

  policiestests
  #4813·shoumikhin opened Sep 30, 2026Contributor·3 of 5 tasks·

  ·
  2

• (docs) improvements and fixes

  documentation
  #4812·NikodemBartnik opened Sep 30, 2026Member·

  ·
  1

  

• feat(policies): accept the starting noise in predict_action_chunk for EO-1, EVO-1 and WALL-OSS

  policiestests
  #4811·shoumikhin opened Sep 30, 2026Contributor·3 of 5 tasks·

  ·
  2

• fix(datasets): report a 29.97 fps stream rate

  datasettests
  #4810·SashaMIT opened Sep 30, 2026·5 tasks·

  ·

• feat(annotations): optional robot-to-human demonstration videos via Inference Providers

  #4809·pkooij opened Sep 30, 2026Member·

  ·
  1

• perf(datasets): batch Lance depth video decoding

  datasettests
  #4808·yanghua opened Sep 30, 2026Contributor·5 tasks·

  ·
  4

• fix(datasets): keep per-element stats for multi-dimensional features

  datasettests
  #4807·breken-ai opened Sep 30, 2026·5 tasks·

  ·
  1

• fix(async): round-trip policy server configuration

  tests
  #4806·ethanvillalovoz opened Sep 29, 2026Contributor·3 of 5 tasks·

  ·
  1

• feat(rollout): add episodic recording mode to DAgger strategy

  #4805·grahamstelzer opened Sep 29, 2026·1 of 5 tasks·

  ·
  1

• perf(datasets): optimize Lance RGB batch decoding

  datasettests
  #4804·yanghua opened Sep 29, 2026Contributor·5 tasks·

  ·
  7