All pull requests

Pull requests

• Fix invalid padding in shared pooling tests

  CLA Signedhelp wantedpartner: arm
  #25034·zingo opened Nov 13, 2026Collaborator·

  ·
  2

• Add quantized KV attention primitives

  ciflow/trunkCLA Signedhelp wantedmodule: armrelease notes: arm
  #25033·xingguo01 opened Nov 13, 2026Collaborator·

  ·
  1

• Harden custom SDPA export path

  CLA Signedhelp wantedpartner: armrelease notes: arm
  #25032·bdemirb opened Nov 13, 2026Collaborator·

  ·
  3

• Declare minimal network access for executorch C++ RE tests

  CLA Signedmeta-exported
  #25030·IanChilds opened Nov 13, 2026·

  ·
  4

• LFM3-VL-2B and LFM3-Omni-2B image, speech in and speech out (#25028)

  CLA Signedmeta-exported
  #25028·psiddh opened Nov 13, 2026Contributor·

  ·
  5

• Compare integer tensors with a scalar as integers

  CLA Signedmeta-exported
  #25027·psiddh opened Nov 13, 2026Contributor·

  ·
  4

• Add an aten.mul.Scalar kernel (#25026)

  CLA Signedmeta-exported
  #25026·psiddh opened Nov 13, 2026Contributor·

  ·
  5

• Phonon: keep the Mimi decoder state on the GPU for CUDA

  CLA Signedmeta-exported
  #25021·psiddh opened Nov 13, 2026Contributor·

  ·
  4

• Phonon: run the Mimi transposed convolutions as convolutions on Vulkan, MLX and CUDA (#25020)

  CLA Signedmeta-exportedmodule: vulkan
  #25020·psiddh opened Nov 13, 2026Contributor·

  ·
  5

• Lower grouped GEMM on sm_80 and above (#25017)

  CLA Signedmeta-exported
  #25017·Gasoonjia opened Nov 13, 2026Contributor·

  ·
  3

• Pin planned host buffers of device programs (#25016)

  CLA Signedmeta-exported
  #25016·Gasoonjia opened Nov 13, 2026Contributor·

  ·
  3

• Add a Diffusers image generation CLI

  CLA Signed
  #25015·kiymetakdemir opened Nov 13, 2026Contributor·

  ·
  1

• Add a C++ runner for Diffusers programs

  CLA Signed
  #25014·kiymetakdemir opened Nov 13, 2026Contributor·

  ·
  1

• Port the Diffusers schedulers and torch's CPU RNG to C++

  CLA Signed
  #25013·kiymetakdemir opened Nov 13, 2026Contributor·

  ·
  2

• Qualcomm AI Engine Direct - Add Authors

  CLA Signed
  #25012·winskuo-quic opened Nov 13, 2026Contributor·

  ·
  2

• Run SDK-independent MediaTek preprocessing tests on Python 3.14

  CLA Signedmeta-exported
  #25011·KasraGhodsi opened Nov 13, 2026·

  ·
  3

• Separate MediaTek preprocessing unit coverage from the vendor SDK

  CLA Signedmeta-exported
  #25010·KasraGhodsi opened Nov 13, 2026·

  ·
  3

• [Native] Select the logits rows an off-graph export returns

  CLA Signed
  #25009·metascroy opened Nov 13, 2026Contributor·

  ·
  1

• [Native] Gather the rows of x[:, rows] in the ggml engine

  CLA Signed
  #25008·metascroy opened Nov 13, 2026Contributor·

  ·
  1

• [Native] Serve many sequences from one ggml engine KV cache

  CLA Signed
  #25007·metascroy opened Nov 13, 2026Contributor·

  ·
  1

• [Native] Separate the ggml engine's K/V storage from its sequence cache

  CLA Signed
  #25006·metascroy opened Nov 13, 2026Contributor·

  ·
  1

• [XNNPACK] Delegate GEMMs that take parts of one packed weight or bias

  CLA Signed
  #25005·john-rocky opened Nov 13, 2026Contributor·

  ·
  3

• Phonon: run the Mimi decoder's transposed convolutions as convolutions on XNNPACK

  CLA Signedmeta-exported
  #25004·psiddh opened Nov 12, 2026Contributor·

  ·
  4

• feat: shortening test names POC

  CLA Signedmodule: nxprelease notes: nxp
  #25002·novak-vaclav opened Nov 12, 2026Collaborator·

  ·
  2

  

• Recompose decomposed hardswish in the XNNPACK provider (#25001)

  CLA Signedmeta-exported
  #25001·JakeStevens opened Nov 12, 2026Contributor·

  ·
  3