All pull requests

Pull requests

• [ET-VK][reduction] Work around Adreno 8x4 dispatch failure

  CLA Signed
  #23350·pytorchbot opened Oct 2, 2026Collaborator·

  ·
  2

• Leave SDPA calls the Triton kernel cannot take to the regular CUDA lowering

  CLA Signedrelease notes: backends
  #23349·shoumikhin opened Oct 2, 2026Contributor·

  ·
  3

• Arm backend: Reject mismatched Ethos-U output buffers

  ciflow/trunkCLA Signedmodule: armpartner: armrelease notes: none
  #23348·Sebastian-Larsson opened Oct 2, 2026Collaborator·

  ·
  1

• Qualcomm AI Engine Direct - Support QAT for concat observer

  CLA Signed
  #23347·shewu-quic opened Oct 2, 2026Contributor·

  ·
  5

• [LLM serving 10/x] Add a standalone batching server launcher

  CLA Signed
  #23345·metascroy opened Oct 2, 2026Contributor·

  ·
  1

• [LLM serving 9/x] Validate batching-backed serving end to end

  CLA Signed
  #23344·metascroy opened Oct 2, 2026Contributor·

  ·
  1

• [LLM serving 8/x] Run concurrent chat requests with per-session ordering

  CLA Signed
  #23343·metascroy opened Oct 2, 2026Contributor·

  ·
  1

• [LLM serving 7/x] Negotiate a multiplexed Python worker client

  CLA Signed
  #23342·metascroy opened Oct 2, 2026Contributor·

  ·
  1

• Refactor examples/kev into examples/systemone/kev

  CLA Signed
  #23341·mergennachin opened Oct 2, 2026Contributor·3 of 4 tasks·

  ·
  7

• [CoreML] Cache the quantizer's pattern graphs instead of rebuilding them per model

  #23323·vijay-kapse opened Oct 1, 2026·

  ·
  4

• Arm backend: Enable host memory capabilities with fixed buck deps.

  ciflow/trunkCLA Signedmodule: armrelease notes: arm
  #23322·wwwind opened Oct 1, 2026Collaborator·

  ·
  1

• Add generic checks to the PR review skill

  CLA Signed
  #23315·Gasoonjia opened Oct 1, 2026Contributor·

  ·
  2

• Delegate beyond DQLinear by default

  CLA Signedmeta-exported
  #23314·pssrawat opened Oct 1, 2026Contributor·

  ·
  4

• Arm backend: Add optional Model Converter debug-info generation.

  ciflow/trunkCLA Signedmodule: armrelease notes: arm
  #23313·wwwind opened Oct 1, 2026Collaborator·

  ·
  2

• Arm backend: Add generated-case reporting artifacts (FACTO)

  ciflow/trunkCLA Signedhelp wantedmodule: armpartner: armrelease notes: arm
  #23312·bdemirb opened Oct 1, 2026Collaborator·

  ·
  4

• Declare the torch version in release wheels

  CLA Signedrelease notes: build
  #23311·shoumikhin opened Oct 1, 2026Contributor·

  ·
  6

• Cortex-M backend: Wire up explicit layout into testing

  ciflow/trunkCLA Signedpartner: armrelease notes: none
  #23310·AdrianLundell opened Oct 1, 2026Collaborator·

  ·
  1

• Arm backend: Add baseline functionality for zero copy IO project.

  ciflow/trunkCLA Signedmodule: armpartner: armrelease notes: arm
  #23309·wwwind opened Oct 1, 2026Collaborator·

  ·
  4

• NXP backend: use channels-first for reference testing model

  CLA Signedmodule: nxprelease notes: nxp
  #23308·novak-vaclav opened Oct 1, 2026Collaborator·

  ·
  4

• NXP backend: Enable model suite CI testing.

  CLA Signedmodule: nxprelease notes: nxp
  #23305·MartinPavella opened Oct 1, 2026Collaborator·

  ·
  2

  

• Share one FakeQuantize across cat under QAT (#23299)

  CLA Signedmeta-exportedmodule: qnnrelease notes: none
  #23299·billmguo opened Oct 1, 2026Contributor·

  ·
  7

• Make ConcatObserver deepcopy-safe (#23298)

  CLA Signedmeta-exportedmodule: qnnrelease notes: none
  #23298·billmguo opened Oct 1, 2026Contributor·

  ·
  4

• [LLM serving 6/x] Add a bounded multiplexed native serving worker

  CLA Signed
  #23291·metascroy opened Sep 30, 2026Contributor·

  ·
  4

• Fix FlatTensor serializer linkage and header installation

  module: build/installrelease notes: none
  #23290·quic-xiyushi opened Sep 30, 2026Contributor·

  ·
  5

• Automate stable release dependency finalization

  CLA Signed
  #23289·JacobSzwejbka opened Sep 30, 2026Contributor·

  ·
  1