Commits

Breadcrumbs

History for

onmain

User selector

All users

Datepicker

All time

Commit History

Commits on Jan 23, 2025

• Revert "Reverting the PR adding Kleidiai-based int4 kernels (#145392)" (#145505)

  Show description for 41b38f7
  

  nikhil-arm

  authored and

  pytorchmergebot

  committed
  41b38f7

  Copy full SHA for 41b38f7

Commits on Jan 22, 2025

• Reverting the PR adding Kleidiai-based int4 kernels (#145392)

  Show description for 0940eb6
  

  albanD

  authored and

  pytorchmergebot

  committed
  0940eb6

  Copy full SHA for 0940eb6

Commits on Jan 3, 2025

• [ROCm] CK Flash Attention Backend (#143695)

  Show description for 0a94bb4
  

  3 peopleauthored and

  pytorchmergebot

  committed
  0a94bb4

  Copy full SHA for 0a94bb4

Commits on Dec 20, 2024

• [ARM][feat]: Add 4 bit dynamic quantization matmuls & KleidiAI Backend (#134124)

  Show description for 94737e8
  

  nikhil-arm

  authored and

  pytorchmergebot

  committed
  94737e8

  Copy full SHA for 94737e8

Commits on Dec 19, 2024

• Revert "[ARM][feat]: Add 4 bit dynamic quantization matmuls & KleidiAI Backend (#134124)"

  Show description for 8136daf
  pytorchmergebot

  committed
  8136daf

  Copy full SHA for 8136daf

• [ARM][feat]: Add 4 bit dynamic quantization matmuls & KleidiAI Backend (#134124)

  Show description for 4b82251
  

  nikhil-arm

  authored and

  pytorchmergebot

  committed
  4b82251

  Copy full SHA for 4b82251

• Revert "[ARM][feat]: Add 4 bit dynamic quantization matmuls & KleidiAI Backend (#134124)"

  Show description for 14fe1f7
  pytorchmergebot

  committed
  14fe1f7

  Copy full SHA for 14fe1f7

Commits on Dec 18, 2024

• [ARM][feat]: Add 4 bit dynamic quantization matmuls & KleidiAI Backend (#134124)

  Show description for d3ff2d4
  

  nikhil-arm

  authored and

  pytorchmergebot

  committed
  d3ff2d4

  Copy full SHA for d3ff2d4

Commits on Dec 17, 2024

• Revert "[ROCm] CK Flash Attention Backend (#138947)"

  Show description for 969b07b
  pytorchmergebot

  committed
  969b07b

  Copy full SHA for 969b07b

• [ROCm] CK Flash Attention Backend (#138947)

  Show description for 500d029
  

  2 peopleauthored and

  pytorchmergebot

  committed
  500d029

  Copy full SHA for 500d029

Commits on Oct 9, 2024

• Clarify opt-einsum usage, fix #127109 (#137596)

  Show description for cfe9702
  

  janeyx99

  authored and

  pytorchmergebot

  committed
  cfe9702

  Copy full SHA for cfe9702

Commits on Sep 24, 2024

• Add option to configure reduced precision math backend for SDPA (#135964)

  Show description for 0a35986
  

  jianyuh

  authored and

  pytorchmergebot

  committed
  0a35986

  Copy full SHA for 0a35986

Commits on Aug 21, 2024

• [sparse] Add cuSPARSELt as a backend (#128534)

  Show description for 255cd75
  

  jcaip

  authored and

  pytorchmergebot

  committed
  255cd75

  Copy full SHA for 255cd75

Commits on Jul 31, 2024

• Expose function to probe whether PyTorch was built with FlashAttention (#131894)

  Show description for f4f7aba
  

  lw

  authored and

  pytorchmergebot

  committed
  f4f7aba

  Copy full SHA for f4f7aba

Commits on Jun 30, 2024

• [cuDNN][SDPA] Remove `TORCH_CUDNN_SDPA_ENABLED=1`, enable cuDNN SDPA by default on H100 and 2nd on other archs >= sm80 (#125343)

  Show description for f845a7a
  

  eqy

  authored and

  pytorchmergebot

  committed
  f845a7a

  Copy full SHA for f845a7a

Commits on Jun 28, 2024

• Revert "[cuDNN][SDPA] Remove `TORCH_CUDNN_SDPA_ENABLED=1`, enable cuDNN SDPA by default on H100 and 2nd on other archs >= sm80 (#125343)"

  Show description for 999eec8
  pytorchmergebot

  committed
  999eec8

  Copy full SHA for 999eec8

Commits on Jun 26, 2024

• [cuDNN][SDPA] Remove `TORCH_CUDNN_SDPA_ENABLED=1`, enable cuDNN SDPA by default on H100 and 2nd on other archs >= sm80 (#125343)

  Show description for b7e7a4c
  

  eqy

  authored and

  pytorchmergebot

  committed
  b7e7a4c

  Copy full SHA for b7e7a4c

Commits on Jun 12, 2024

• Revert "[cuDNN][SDPA] Remove `TORCH_CUDNN_SDPA_ENABLED=1`, enable cuDNN SDPA by default on H100 and 2nd on other archs >= sm80 (#125343)"

  Show description for 817ce68
  pytorchmergebot

  committed
  817ce68

  Copy full SHA for 817ce68

Commits on Jun 9, 2024

• [cuDNN][SDPA] Remove `TORCH_CUDNN_SDPA_ENABLED=1`, enable cuDNN SDPA by default on H100 and 2nd on other archs >= sm80 (#125343)

  Show description for 4c97193
  

  eqy

  authored and

  pytorchmergebot

  committed
  4c97193

  Copy full SHA for 4c97193

Commits on Apr 22, 2024

• preferred blas library; cublaslt gemm implementation (#122106)

  Show description for 6ede882
  

  jeffdaily

  authored and

  pytorchmergebot

  committed
  6ede882

  Copy full SHA for 6ede882

Commits on Feb 14, 2024

• [CUDNN][SDPA] Experimental cuDNN Flash Attention v2 Inference (#115663)

  Show description for cd380c7
  

  eqy

  authored and

  pytorchmergebot

  committed
  cd380c7

  Copy full SHA for cd380c7

Commits on Jan 24, 2024

• Remove sdp_kernel and replace with sdpa_kernel in attention namespace (#114689)

  Show description for 4e29f01
  

  drisspg

  authored and

  pytorchmergebot

  committed
  4e29f01

  Copy full SHA for 4e29f01

Commits on Jan 18, 2024

• Revert "[CUDNN][SDPA] Experimental cuDNN Flash Attention v2 Inference (#115663)"

  Show description for 2f84a9d
  pytorchmergebot

  committed
  2f84a9d

  Copy full SHA for 2f84a9d

• [CUDNN][SDPA] Experimental cuDNN Flash Attention v2 Inference (#115663)

  Show description for 5aa92b5
  

  eqy

  authored and

  pytorchmergebot

  committed
  5aa92b5

  Copy full SHA for 5aa92b5

Commits on Jan 2, 2024

• Add config to disable TransformerEncoder/MHA fastpath (#112212)

  Show description for 0f6f582
  

  mikaylagawarecki

  authored and

  pytorchmergebot

  committed
  0f6f582

  Copy full SHA for 0f6f582

Commits on Dec 27, 2023

• Add setUserEnabledNNPACK config (#116152)

  Show description for 6b91e69
  

  angelayi

  authored and

  pytorchmergebot

  committed
  6b91e69

  Copy full SHA for 6b91e69

Commits on Nov 15, 2023

• expose sdpa helpers to python (#110496)

  Show description for 9b0f2f8
  

  drisspg

  authored and

  pytorchmergebot

  committed
  9b0f2f8

  Copy full SHA for 9b0f2f8

Commits on Oct 6, 2023

• Doc test non packages (#110568)

  Show description for c4db607
  

  albanD

  authored and

  pytorchmergebot

  committed
  c4db607

  Copy full SHA for c4db607

Commits on May 22, 2023

• Fix Backend docs search items (#101214)

  Show description for a630328
  

  ain-soph

  authored and

  pytorchmergebot

  committed
  a630328

  Copy full SHA for a630328

Commits on May 3, 2023

• Publicly exposing `torch.backends.cpu.get_cpu_capability()` (#100164)

  Show description for 6a12f10
  

  vfdev-5

  authored and

  pytorchmergebot

  committed
  6a12f10

  Copy full SHA for 6a12f10

Commits on Mar 14, 2023

• [CUDA][cuFFT] Minor fix for cuFFT plan cache docs (#96373)

  Show description for 6e3e22d
  

  eqy

  authored and

  pytorchmergebot

  committed
  6e3e22d

  Copy full SHA for 6e3e22d

Commits on Dec 21, 2022

• [cuBLAS] Add an option to disable reduced precision reductions for BF16 GEMM (#89172)

  Show description for 8b617f8
  

  eqy

  authored and

  pytorchmergebot

  committed
  8b617f8

  Copy full SHA for 8b617f8

Commits on Nov 16, 2022

• Create native function for determining which implementation of SDP to call (#89029)

  Show description for b291c12
  

  drisspg

  authored and

  pytorchmergebot

  committed
  b291c12

  Copy full SHA for b291c12

Commits on Oct 28, 2022

• Add mem efficient backend flag (#87946)

  Show description for 35c611d
  

  drisspg

  authored and

  pytorchmergebot

  committed
  35c611d

  Copy full SHA for 35c611d

Commits on Oct 5, 2022

• Add opteinsum backend to give users control (#86219)

  Show description for a348975
  

  janeyx99

  authored and

  pytorchmergebot

  committed
  a348975

  Copy full SHA for a348975

Pagination

Previous

Next