All pull requests

Pull requests

• Add a TensorRT backend to the ExecuTorch export

  #275·shoumikhin opened Nov 12, 2026·

  ·

• SAC: distributed training and CNN support on top of the CUDA graph optimizations

  #274·lgulich opened Nov 11, 2026Collaborator·

  ·

• SAC: capture updates and environment interaction in CUDA graphs

  #273·lgulich opened Nov 11, 2026Collaborator·

  ·

• feat: Add torch XPU support

  #272·nathantsao opened Nov 11, 2026·

  ·

• Add ExecuTorch export for policies

  #271·shoumikhin opened Nov 11, 2026·

  ·

• Return only the ONNX outputs from the GRU ONNX model

  #270·shoumikhin opened Nov 11, 2026·

  ·

• [main] Add distributed SAC and CNN model support

  #268·jnskkmhr opened Nov 9, 2026Contributor·

  ·

• Aggregate episode metrics without mutating environment extras

  #267·Afloat16 opened Nov 7, 2026Contributor·

  ·

• Preserve import errors and validate resolved callables

  #266·Afloat16 opened Nov 7, 2026Contributor·

  ·

• Preserve recurrent hidden-state dtypes in rollout storage

  #264·Afloat16 opened Nov 3, 2026Contributor·

  ·

• Fix bucketed reduction of noncontiguous gradients

  #263·Afloat16 opened Nov 3, 2026Contributor·

  ·

• fix: preserve recurrent state dtype in TorchScript exports

  #261·Yushu-Li opened Nov 1, 2026·

  ·

• Keep computed CNN padding nonnegative for small kernels

  #259·Afloat16 opened Oct 30, 2026Contributor·

  ·

• Normalize vector rewards without ambiguous tensor conditions

  #258·Afloat16 opened Oct 30, 2026Contributor·

  ·

• Include action-range scaling in Beta entropy

  #257·Afloat16 opened Oct 30, 2026Contributor·

  ·

• [main] Add SuperMimic model-based training

  #255·rafaelcathomen opened Oct 28, 2026Contributor·

  ·
  4

• Add PPO critic lifecycle hooks and DistributionalPPO

  #236·samindaa opened Sep 27, 2026·

  ·
  1

• [Feature] Multi critic PPO

  #224·ManuelActisCassa opened Jul 31, 2026·

  ·
  1