Commits

Branch selector

 main

User selector

All users

Datepicker

All time

Commit History

Commits on Dec 12, 2024

• don't save inputs buffer of FlashAttenFunc to reduce memory usage for inference mode (#1383)

  XiaobingSuper

  authored
  0dfb281

  Copy full SHA for 0dfb281

Commits on Dec 7, 2024

• [CI] Use MAX_JOBS=1 with nvcc 12.3, don't need OLD_GENERATOR_PATH

  tridao

  committed
  f86e3dd

  Copy full SHA for f86e3dd

• Bump to v2.7.2

  tridao

  committed
  b7d29fb

  Copy full SHA for b7d29fb

• [CI] Use MAX_JOBS=2 even with nvcc 12.3

  tridao

  committed
  d3b1cd1

  Copy full SHA for d3b1cd1

• Update to Cutlass 3.6.0

  tridao

  committed
  61a23ea

  Copy full SHA for 61a23ea

• [CI] Don't include <ATen/cuda/CUDAGraphsUtils.cuh>

  tridao

  committed
  9375ac9

  Copy full SHA for 9375ac9

• [CI] Change torch #include to make it work with torch 2.1 Philox

  tridao

  committed
  e782d28

  Copy full SHA for e782d28

Commits on Dec 6, 2024

• [CI] Use torch 2.6.0.dev20241001, reduce torch #include

  tridao

  committed
  073afd5

  Copy full SHA for 073afd5

• [CI] Fix CUDA version for torch 2.6

  tridao

  committed
  cf0f4c3

  Copy full SHA for cf0f4c3

• Bump to v2.7.1

  tridao

  committed
  cc408f9

  Copy full SHA for cc408f9

• [CI] Drop Python 3.8, add Python 3.13, add pytorch 2.6.0.dev20241010

  tridao

  committed
  8878692

  Copy full SHA for 8878692

• Add how to import FA3 to documentation (#1112)

  AdamLouly

  authored
  df1a744

  Copy full SHA for df1a744

• [AMD] Triton Backend for ROCm (#1203)

  Show description for b518517
  micmelesse

  authored
  b518517

  Copy full SHA for b518517

Commits on Dec 4, 2024

• Fix compilation with clang on ARM64 (#1285)

  sclarkson

  authored
  1feb711

  Copy full SHA for 1feb711

Commits on Dec 2, 2024

• Fix FA3 Varlen Performance regression (#1361)

  kadeng

  authored
  0823cf7

  Copy full SHA for 0823cf7

• flash_bwd_kernel.h: add maybe_unused annotation to suppress lengthy warnings about dtanh not being used in some compile permutations. (#1363)

  acgessler

  authored
  ca71144

  Copy full SHA for ca71144

Commits on Nov 16, 2024

• Fix swiglu backwards return type (#1337)

  ntenenz

  authored
  7153673

  Copy full SHA for 7153673

Commits on Nov 13, 2024

• [CI] Pytorch 2.5.1 does not support python 3.8

  tridao

  committed
  641db75

  Copy full SHA for 641db75

• Update README for FA3

  tridao

  committed
  7435839

  Copy full SHA for 7435839

Commits on Nov 12, 2024

• [CI] Switch back to CUDA 12.4

  tridao

  committed
  241c682

  Copy full SHA for 241c682

• Bump to v2.7.0

  tridao

  committed
  c555642

  Copy full SHA for c555642

• [CI] Still use CUDA 12.3 but pull the right pytorch version

  tridao

  committed
  6ffeb57

  Copy full SHA for 6ffeb57

• Use CUDA 12.4 in the build system (#1326)

  Show description for 42f2b8b
  EthanSteinberg

  authored
  42f2b8b

  Copy full SHA for 42f2b8b

• Drop support for Pytorch 2.0

  tridao

  committed
  2f6c633

  Copy full SHA for 2f6c633

• [AMD ROCm] Fix KVcache bug and improve performance (#1328)

  Show description for 88d1657
  

  rocking5566

  and

  poyenc

  authored
  88d1657

  Copy full SHA for 88d1657

• Make FA3 paged attention ready for upgrade to Cutlass 3.6 (#1331)

  kadeng

  authored
  284e2c6

  Copy full SHA for 284e2c6

Commits on Nov 10, 2024

• Paged Attention support for FA3 (#1268)

  kadeng

  authored
  b443207

  Copy full SHA for b443207

Commits on Nov 7, 2024

• Feat: Add support for PyTorch 2.5 in workflows (#1284)

  Show description for f0bf3ed
  NanoCode012

  authored
  f0bf3ed

  Copy full SHA for f0bf3ed

Commits on Oct 31, 2024

• Make namespace comment consistent (#1305)

  Show description for 478ee66
  

  ngocson2vn

  and

  Sony Nguyen

  authored
  478ee66

  Copy full SHA for 478ee66

Commits on Oct 15, 2024

• Fix copy-paste error in hopper tests (#1279)

  milesvant

  authored
  c1d146c

  Copy full SHA for c1d146c

• FA3 kvcache + split kv + gqa parallelization (#1236)

  jayhshah

  authored
  a5a7527

  Copy full SHA for a5a7527

Commits on Oct 5, 2024

• [CrossEntropy] Fix where labels address not aligned to 16 bytes

  tridao

  committed
  bedf877

  Copy full SHA for bedf877

Commits on Sep 20, 2024

• Hotfix due to change of upstream api (#1239)

  rocking5566

  authored
  53a4f34

  Copy full SHA for 53a4f34

• Fix FAv3 compilation with MSVC (#1240)

  hlky

  authored
  8476986

  Copy full SHA for 8476986

• Merge pull request #1233 from Dao-AILab/ipiszy/local_attn

  Show description for 9cafd4a
  ipiszy

  authored
  9cafd4a

  Copy full SHA for 9cafd4a

Pagination

Previous

Next