All pull requests

Pull requests

• Separate ESP32 CI runs by commit

  CLA Signed
  #24472·rascani opened Nov 2, 2026Contributor·

  ·
  2

• Reuse committed prompt prefixes from active text requests

  CLA Signed
  #24471·mergennachin opened Nov 2, 2026Contributor·6 of 9 tasks·

  ·
  2

• CI: stop the torch-free bindings check from matching load addresses

  CLA Signedtopic: not user facing
  #24470·shoumikhin opened Nov 2, 2026Contributor·

  ·
  5

• Move backend-neutral Cadence passes to backends/transforms

  CLA Signedmeta-exported
  #24469·mcremon-meta opened Nov 2, 2026Contributor·

  ·
  3

• CUDA backend: keep wide attention heads that are not a power of 2 off the Triton SDPA kernel

  CLA Signed
  #24468·pytorchbot opened Nov 2, 2026Collaborator·

  ·
  1

• [MLX] Experimental int8 x int4 MPP GEMM + Muse Glimmer benchmark

  CLA Signed
  #24467·metascroy opened Nov 2, 2026Contributor·

  ·
  2

• OpenVINO: fix wrong outputs when a partition returns a copy of its input

  CLA Signedrelease notes: openvino
  #24466·shoumikhin opened Nov 2, 2026Contributor·

  ·
  10

• Serialize a program without copying its delegate blobs

  CLA Signed
  #24465·pytorchbot opened Nov 2, 2026Collaborator·

  ·
  1

• XNNPACK: fix enable_bf16 load failures and BatchNorm folding into a 3-D linear output

  CLA Signedrelease notes: xnnpack
  #24464·shoumikhin opened Nov 2, 2026Contributor·

  ·
  18

• Validate and lock SharedBuffer allocations (#24463)

  CLA Signedmeta-exported
  #24463·alirezamonfared opened Nov 2, 2026Contributor·

  ·
  4

• Isolate cached delegates by Module lifetime (#24462)

  CLA Signedmeta-exportedmodule: qnn
  #24462·alirezamonfared opened Nov 2, 2026Contributor·

  ·
  4

• Reinstall pip from PyPI in the native wheel pre-build script

  CLA Signed
  #24461·JakeStevens opened Nov 2, 2026Contributor·

  ·
  2

• Support 3-D convolution in the portable convolution kernel

  CLA Signed
  #24460·pytorchbot opened Nov 2, 2026Collaborator·

  ·
  1

• Mark the 3-D convolution cases that fail inside a backend as expected failures

  ciflow/nightlyciflow/trunkCLA Signedmodule: armrelease notes: none
  #24459·shoumikhin opened Nov 2, 2026Contributor·

  ·
  12

• Add boltnn_executorch_llm_runner: run weight-shared QNN LLM exports from the AP

  CLA Signedmeta-exported
  #24456·billmguo opened Nov 2, 2026Contributor·

  ·
  3

• CUDA backend: lower fused attention variants

  CLA Signed
  #24455·pytorchbot opened Nov 2, 2026Collaborator·

  ·
  1

• Lower static FP32 matmul to existing matrix kernels

  CLA Signedmeta-exported
  #24454·JakeStevens opened Nov 2, 2026Contributor·

  ·
  3

• Remove clones that only request a memory format

  CLA Signedmeta-exported
  #24453·zonglinpeng opened Nov 2, 2026Contributor·

  ·
  3

• Lower tanh with the existing unary kernel

  CLA Signedmeta-exported
  #24452·JakeStevens opened Nov 2, 2026Contributor·

  ·
  3

• Lower exp with the existing unary kernel

  CLA Signedmeta-exported
  #24451·JakeStevens opened Nov 2, 2026Contributor·

  ·
  3

• CUDA backend: delete each output handle once when a program returns one tensor twice

  ciflow/trunkCLA Signedmodule: arm
  #24448·pytorchbot opened Nov 2, 2026Collaborator·

  ·
  1

• Add a portable kernel for antialiased bicubic resize

  ciflow/trunkCLA Signedmodule: arm
  #24447·pytorchbot opened Nov 2, 2026Collaborator·

  ·
  1

• OpenVINO: fix bool min/max and masked read lowering, and follow-ups to #24294

  ciflow/trunkCLA Signedmodule: arm
  #24446·pytorchbot opened Nov 2, 2026Collaborator·

  ·
  1

• Keep ops above 6 dims out of XNNPACK and fix quantized ops it leaves out

  CLA Signed
  #24445·pytorchbot opened Nov 2, 2026Collaborator·

  ·
  1

• Bump node-forge from 1.3.1 to 1.4.2 in /examples/demo-apps/react-native/rnllama

  CLA Signeddependenciesjavascript
  #24439·dependabot[bot] opened Nov 2, 2026Bot·

  ·
  2