Commits

Branch selector

 v2.9.0

User selector

All users

Datepicker

All time

Commit History

Commits on Oct 9, 2025

• CUDA aarch64 12.6 and 12.8 builds fix triton constraints (#165022)

  Show description for 0fabc3b
  

  pytorchbot

  and

  atalman

  authored
  0fabc3b

  Copy full SHA for 0fabc3b

Commits on Oct 8, 2025

• [MPS] Update OS version in error message (#164949)

  Show description for 26e023a
  

  pytorchbot

  and

  malfet

  authored
  26e023a

  Copy full SHA for 26e023a

• CUDA 13.0 builds fix on Amazon Linux 2023 (#164893)

  Show description for 6f12be2
  

  3 peopleauthored
  6f12be2

  Copy full SHA for 6f12be2

Commits on Oct 7, 2025

• update the baseline data for the operator benchmark (#164789)

  Show description for 42f0c2c
  

  pytorchbot

  and

  LifengWang

  authored
  42f0c2c

  Copy full SHA for 42f0c2c

Commits on Oct 6, 2025

• fix cpp extension distributed warning spew (#164785)

  Show description for b015422
  

  pytorchbot

  and

  msaroufim

  authored
  b015422

  Copy full SHA for b015422

• Fix docker build issue after 164575 (#164779)

  Show description for d4c4307
  

  pytorchbot

  and

  atalman

  authored
  d4c4307

  Copy full SHA for d4c4307

• [ROCm] Increase binary build timeout to 5 hours (300 minutes) (#164770)

  Show description for 3b57315
  

  3 peopleauthored
  3b57315

  Copy full SHA for 3b57315

Commits on Oct 3, 2025

• Pin conda version for Docker builds (#164579)

  Show description for c74f057
  

  pytorchbot

  and

  atalman

  authored
  c74f057

  Copy full SHA for c74f057

Commits on Oct 2, 2025

• [Cherry-Pick] Work Around exposing statically linked libstdc++ CXX11 ABI strong symbols (#163980) (#164508)

  Show description for fd36458
  

  atalman

  and

  malfet

  authored
  fd36458

  Copy full SHA for fd36458

• [CherrryPick][2.9] Cherry pick request for `Reapply "Make functionalization ViewMeta serializable with pickle #163769` (#163873)

  Show description for 2f6387e
  

  Lucaskabela

  and

  bdhirsh

  authored
  2f6387e

  Copy full SHA for 2f6387e

• fix pickling for BitwiseFn (#163861)

  Show description for 017d857
  

  3 peopleauthored
  017d857

  Copy full SHA for 017d857

• Make sure Windows CUDA 12.8 build follow same arches as Linux builds (#164477)

  Show description for d6e8411
  

  pytorchbot

  and

  atalman

  authored
  d6e8411

  Copy full SHA for d6e8411

Commits on Oct 1, 2025

• [Flex] Fix silent correctness w/ backpropping grads (#164366)

  Show description for 10b501f
  

  pytorchbot

  and

  drisspg

  authored
  10b501f

  Copy full SHA for 10b501f

• [a2av] Separate in/out splits into two tensors (#164028)

  Show description for 31c72b8
  

  pytorchbot

  and

  kwen2501

  authored
  31c72b8

  Copy full SHA for 31c72b8

• [Flex attention] Fix flex attention head broadcast (#164368)

  Show description for 1cd83de
  

  pytorchbot

  and

  Isalia20

  authored
  1cd83de

  Copy full SHA for 1cd83de

• Update Gloo submodule (#164371)

  Show description for 881c2cc
  

  pytorchbot

  and

  malfet

  authored
  881c2cc

  Copy full SHA for 881c2cc

• [MPS] Chunk fillBuffer into 4Gb slices (#164370)

  Show description for 764f655
  

  pytorchbot

  and

  malfet

  authored
  764f655

  Copy full SHA for 764f655

• Update Microsoft C++ Redistributable to the latest version (#164369)

  Show description for 3e8a062
  

  pytorchbot

  and

  saman-amd

  authored
  3e8a062

  Copy full SHA for 3e8a062

• Fix warn message (#164367)

  Show description for 3abee62
  

  pytorchbot

  and

  drisspg

  authored
  3abee62

  Copy full SHA for 3abee62

• [MPSHooks] Release pending command encoder (#164365)

  Show description for f227c88
  

  pytorchbot

  and

  malfet

  authored
  f227c88

  Copy full SHA for f227c88

• [SDPA] [MPS] Fixes regression in 2.8.0 for scaled_dot_product_attention using mps (#164364)

  Show description for a5feacb
  

  pytorchbot

  and

  Vismai-Khanderao

  authored
  a5feacb

  Copy full SHA for a5feacb

• Update Sphinx theme (#164147) (#164254)

  Show description for 71282c8
  svekars

  authored
  71282c8

  Copy full SHA for 71282c8

• [vllm hash update] update the pinned vllm hash (#164190) (#164312)

  Show description for e70d9f5
  

  huydhn

  and

  pytorchupdatebot

  authored
  e70d9f5

  Copy full SHA for e70d9f5

Commits on Sep 30, 2025

• Clean up obsoleted vLLM tests (#164282)

  Show description for 005e3e8
  

  pytorchbot

  and

  huydhn

  authored
  005e3e8

  Copy full SHA for 005e3e8

• [AARCH64][CD][CUDA13][Triton][PTXAS] Turn on BUILD_BUNDLE_PTXAS=1 (#164236)

  Show description for 72cf48e
  

  pytorchbot

  and

  nWEIdia

  authored
  72cf48e

  Copy full SHA for 72cf48e

Commits on Sep 29, 2025

• [CI] Move libtorch-cpu-shared-with-deps-release-build to python 3.10 (#164182)

  Show description for a21a4bf
  

  pytorchbot

  and

  atalman

  authored
  a21a4bf

  Copy full SHA for a21a4bf

• Use linux.g4dn.4xlarge.nvidia.gpu for cuda 12.4 legacy driver tests (#164172)

  Show description for 21fec65
  

  3 peopleauthored
  21fec65

  Copy full SHA for 21fec65

• [CUDA] revert PR 130472 (#163379)

  Show description for 22d46b5
  

  pytorchbot

  and

  thenumberouscode

  authored
  22d46b5

  Copy full SHA for 22d46b5

• Skip test_conv3d_cudnn_broken on ROCM (#164163)

  Show description for d1b63e2
  

  pytorchbot

  and

  malfet

  authored
  d1b63e2

  Copy full SHA for d1b63e2

• [c10d] P2P tensors must be dense (#163981)

  Show description for 20100b7
  

  pytorchbot

  and

  kwen2501

  authored
  20100b7

  Copy full SHA for 20100b7

• Add operator benchmarking run to CI nightly (#164151)

  Show description for a2c7704
  

  3 peopleauthored
  a2c7704

  Copy full SHA for a2c7704

• Fix operator benchmark issue#162708 (#164140)

  Show description for b64fc8e
  

  pytorchbot

  and

  jainapurva

  authored
  b64fc8e

  Copy full SHA for b64fc8e

• [cuDNN][Convolution] Disable cuDNN for 3D convolutions with kernel size != 1 for cuDNN 9.8+ (#164027)

  Show description for 709f4f6
  

  3 peopleauthored
  709f4f6

  Copy full SHA for 709f4f6

• [cuDNN][SDPA] Disable dropout for cuDNN SDPA on 9.11 - 9.13 (#164026)

  Show description for 11f776c
  

  pytorchbot

  and

  eqy

  authored
  11f776c

  Copy full SHA for 11f776c

• [cuDNN][conv][64-bit] Disable cuDNN for 64-bit depthwise convs again (#164023)

  Show description for 45e257f
  

  pytorchbot

  and

  eqy

  authored
  45e257f

  Copy full SHA for 45e257f

Pagination

Previous

Next