All pull requests

Pull requests

• Fix MultiOptimizer crash when an optimizer's group is empty

  #4570·ishtihoss opened Sep 26, 2026Contributor·

  ·

• Perf: Metal kernel source hash cache

  #4568·STRML opened Sep 26, 2026·

  ·

• Adding VJP for gated delta net.

  #4565·tpegolotti opened Sep 25, 2026Collaborator·

  ·

• Make scheduler and CPU allocator fork-safe

  #4564·MarcosAsh opened Sep 25, 2026Contributor·

  ·

• Adding VJP for the scaled dot product attention.

  #4563·tpegolotti opened Sep 25, 2026Collaborator·

  ·

• metal: runtime setter for command-buffer limits

  #4562·aidiffuser opened Sep 25, 2026·

  ·

• python: Fix bare ellipsis __setitem__ by squeezing leading singleton dimensions

  #4561·shaeantoine opened Sep 25, 2026·

  ·

• register_group: Strict init fallback

  #4558·ronaldmannak opened Sep 24, 2026·

  ·

• ReduceScatter::eval_cpu: Assert inputs.size = 1

  #4557·ronaldmannak opened Sep 24, 2026·

  ·
  4

• Fix eval cleanup deadlock by committing GPU streams before waiting

  #4556·ronaldmannak opened Sep 24, 2026·

  ·

• Add Python 3.15 and 3.15t support

  #4555·XXXXRT666 opened Sep 24, 2026Contributor·

  ·

• [CUDA] port Metal rhs gather_qmm optimizations

  #4554·dhiltgen opened Sep 23, 2026Contributor·

  ·

• Contraction docs addition

  low priority
  #4550·simeetnayan81 opened Sep 23, 2026·

  ·

• fix(reduce): support 64-bit indexing for reductions on views past 2^31 and large arrays

  #4549·nicoloangileri opened Sep 23, 2026·

  ·
  1

• Fix empty unsigned sum/prod failing on Metal

  #4546·CodeWithMoin opened Sep 22, 2026·

  ·

• Fixes grouped Conv1d/Conv2d init to use in_channels // groups

  #4544·CodeWithMoin opened Sep 22, 2026·

  ·
  1

• Fix/split data size span

  await verification
  #4540·michalk8 opened Sep 21, 2026Collaborator·

  ·

• Add C++ linear SVM training example with automatic differentiation

  #4537·StarxSky opened Sep 20, 2026·

  ·

• Fix CustomKernel Metal dispatch silently truncating the requested threadgrid

  #4535·pseudobacon opened Sep 19, 2026·

  ·
  1

• Advances file offset during partial reads and adds EOF/signal handling in ParallelFileReader::read #4526

  #4532·Ritabanm opened Sep 18, 2026Contributor·

  ·
  1

• fix(CUDA): Do not throw from ~CudaHandle while the driver is unloading

  #4531·inureyes opened Sep 18, 2026·

  ·

• Report a dead peer in JACCL instead of spinning forever; name the lin…

  low priority
  #4530·Sofille65 opened Sep 18, 2026·

  ·

• Fix col_reduce_longcolumn for negative-stride views

  low priority
  #4529·mateuuszzzzz opened Sep 17, 2026·

  ·
  1

• Fix default.metallib lookup under swift test with Swift Build

  low priority
  #4527·heckj opened Sep 17, 2026Contributor·

  ·
  1

  1

• Clear CUDA errors once reported and raise eval errors in buffer protocol

  low priority
  #4523·aleroot opened Sep 16, 2026Contributor·

  ·