All pull requests

Pull requests

• Retry ESP32 emulator installation after transient failures

  CLA Signed
  #23941·rascani opened Oct 16, 2026Contributor·

  ·
  2

• [PyBindings] Preserve compatibility after removing ATen dependency

  CLA Signed
  #23939·JacobSzwejbka opened Oct 16, 2026Contributor·

  ·
  2

• Run WebGPU models in browser Wasm (#23938)

  CLA Signedmeta-exported
  #23938·raedle opened Oct 16, 2026·

  ·
  5

• Add OSS CI for the Native CPU engine

  CLA Signedmeta-exported
  #23937·JakeStevens opened Oct 16, 2026Contributor·

  ·
  3

• Share the Native vision runner across CPU and Vulkan engines

  CLA Signedmeta-exportedmodule: vulkan
  #23936·JakeStevens opened Oct 16, 2026Contributor·

  ·
  3

• Add a Native CPU engine backed by CPUPlan

  CLA Signedmeta-exported
  #23935·JakeStevens opened Oct 16, 2026Contributor·

  ·
  3

• Add torchvision export workflow to CpuBackend

  CLA Signedmeta-exported
  #23934·JakeStevens opened Oct 16, 2026Contributor·

  ·
  3

• Give Triton SDPA's padded keys no softmax weight without a mask

  CLA Signedcommunity: contributionmodule: cudarelease notes: backends
  #23933·edmundshieh opened Oct 16, 2026·

  ·
  3

  

• Fix Python 3.15 CUDA and Windows wheel builds

  CLA Signed
  #23932·JacobSzwejbka opened Oct 16, 2026Contributor·

  ·
  2

• Flip executorch android extension to source-only ABI (#23931)

  CLA Signedmeta-exported
  #23931·rybalkinsd opened Oct 16, 2026·

  ·
  8

• Add CMake CI for native runtime, GGML, and Vulkan (#23930)

  CLA Signedmeta-exportedmodule: vulkan
  #23930·JakeStevens opened Oct 16, 2026Contributor·

  ·
  3

• Pin llama.cpp for the native GGML engine (#23929)

  CLA Signedmeta-exported
  #23929·JakeStevens opened Oct 16, 2026Contributor·

  ·
  3

• BundleIO: Match bundled tensors to program I/O layouts

  ciflow/trunkCLA Signedpartner: armrelease notes: devtools
  #23927·zingo opened Oct 16, 2026Collaborator·

  ·
  3

• Add ETDump tracing to CpuBackend (#23926)

  CLA Signedmeta-exported
  #23926·JakeStevens opened Oct 16, 2026Contributor·

  ·
  3

• NXP backend: Shorten test names

  CLA Signedmodule: nxprelease notes: nxp
  #23925·roman-janik-nxp opened Oct 16, 2026Collaborator·

  ·
  2

• Cortex-M: Fix pass ordering for quantized attention

  ciflow/trunkCLA Signedmodule: microcontrollerspartner: armrelease notes: arm
  #23923·zingo opened Oct 16, 2026Collaborator·

  ·
  1

• FACTO: Add Arm U85 generated-test profile

  CLA Signedpartner: armrelease notes: arm
  #23920·bdemirb opened Oct 16, 2026Collaborator·

  ·
  4

• ci: use selected Python for pinned PyTorch installs

  CLA Signedpartner: armrelease notes: none
  #23919·zingo opened Oct 16, 2026Collaborator·

  ·
  1

• Quacomm AI Engine Direct - Bump QNN SDK version from 2.50 to 2.51

  CLA Signed
  #23918·winskuo-quic opened Oct 16, 2026Contributor·

  ·
  2

• Qualcomm AI Engine Direct - Keep each node's own LPBQ block size

  CLA Signedcommunity: contributionmodule: qnnpartner: qualcomm
  #23916·msluszniak opened Oct 16, 2026Contributor·

  ·
  2

• Qualcomm AI Engine Direct - Keep a non-delegated op's weight dequantize on CPU

  CLA Signedcommunity: contributionmodule: qnnpartner: qualcomm
  #23915·msluszniak opened Oct 16, 2026Contributor·

  ·
  1

  2

• Check quantized embedding qparam dtypes against how the kernels read them

  CLA Signedcommunity: contributionmodule: kernelsrelease notes: ops & kernels
  #23912·Shakhtar-Sankur opened Oct 16, 2026Contributor·

  ·
  2

  

• Skip ggml targets on platforms their backends do not support

  CLA Signedmeta-exported
  #23911·digantdesai opened Oct 16, 2026Contributor·

  ·
  6

• Packed HiFi FC: fused 4-bit kernel on HiFi4

  CLA Signedmeta-exported
  #23910·mcremon-meta opened Oct 16, 2026Contributor·

  ·
  3

• Update MG README with offgraph cache and batching

  CLA Signed
  #23909·metascroy opened Oct 16, 2026Contributor·

  ·
  4