All pull requests

Pull requests

• Remove the max_episode_steps assignment after Timeout

  #104·YannBerthelot opened Oct 29, 2026·

  ·
  1

• Clear the task embedding's policy gradient after update_pi

  #103·YannBerthelot opened Oct 29, 2026·

  ·
  1

• Broadcast scalar task embeddings across time and batch dimensions

  #97·Afloat16 opened Oct 6, 2026Contributor·

  ·

• Isolate observation encoders between world model instances

  #96·Afloat16 opened Oct 6, 2026Contributor·

  ·

• Preserve the tanh Jacobian in the dimension-scaled policy entropy

  #95·Afloat16 opened Oct 6, 2026Contributor·

  ·

• Restore learning in raw and symlog scalar regression modes

  #94·Afloat16 opened Oct 2, 2026Contributor·

  ·

• Preserve signed value transform precision and gradients at zero

  #93·Afloat16 opened Sep 29, 2026Contributor·

  ·

• Respect the requested dimension in gumbel_softmax_sample

  #92·Afloat16 opened Sep 29, 2026Contributor·

  ·

• Testing/td mpc2/isaac

  #66·ameyj17 opened May 8, 2025·

  ·