Commits

Branch selector

 main

User selector

ProExpertProg

All time

Commit History

Commits on Feb 11, 2026

• [torch.compile] Enable AR+rms fusion by default available for `-O2` (#34299)

  Show description for addac0e
  ProExpertProg

  authored
  addac0e

  Copy full SHA for addac0e

Commits on Feb 9, 2026

• [CI][torch.compile] Fix incorrect filtering for E2E fusion tests on B200 (#34031)

  Show description for 781ddf7
  ProExpertProg

  authored
  781ddf7

  Copy full SHA for 781ddf7

Commits on Feb 6, 2026

• [torch.compile] Reorganize vllm/compilation and tests/compile (0/N for vLLM IR) (#33731)

  Show description for ac32e66
  ProExpertProg

  authored
  ac32e66

  Copy full SHA for ac32e66

Commits on Feb 5, 2026

• Revert "[Attention][FA3] Update FA3 to include new swizzle optimization" (#33841)

  ProExpertProg

  authored
  e3bf79f

  Copy full SHA for e3bf79f

• [CI][torch.compile] Reduce e2e fusion test time (#33293)

  Show description for 4d95135
  ProExpertProg

  authored
  4d95135

  Copy full SHA for 4d95135

Commits on Feb 1, 2026

• Change defaults for vllm bench startup (#33489)

  ProExpertProg

  authored
  6720238

  Copy full SHA for 6720238

Commits on Jan 31, 2026

• [fix][torch.compile] Fix cold-start compilation time increase by adding kv cache update to splitting ops (#33441)

  Show description for 15f40b2
  

  ProExpertProg

  and

  zou3519

  authored
  15f40b2

  Copy full SHA for 15f40b2

Commits on Jan 23, 2026

• [torch.compile][CI] Add back attn fusion on hopper/ada (#32940)

  Show description for bbbd696
  ProExpertProg

  authored
  bbbd696

  Copy full SHA for bbbd696

• [torch.compile] Compile `CustomOp.forward_native` for `SiluAndMul` and `QuantFP8` to avoid raw torch ops inside opaque custom ops (#32806)

  Show description for 5e4e0e5
  

  ProExpertProg

  and

  mgoin

  authored
  5e4e0e5

  Copy full SHA for 5e4e0e5

Commits on Oct 17, 2025

• [torch.compile] Enable attention and allreduce fusion without custom ops enabled (#24604)

  Show description for bd7157a
  ProExpertProg

  authored
  bd7157a

  Copy full SHA for bd7157a

Commits on Oct 14, 2025

• [torch.compile] Fix tests for torch==2.9 inductor partition (#26116)

  Show description for 2dcd12d
  ProExpertProg

  authored
  2dcd12d

  Copy full SHA for 2dcd12d

Commits on Sep 23, 2025

• [Bugfix] fix custom op test (#25429)

  Show description for d588cd2
  ProExpertProg

  authored
  d588cd2

  Copy full SHA for d588cd2

Commits on Sep 22, 2025

• [torch.compile] Cleanup compilation tests and custom passes, add debug utils, fix DCE bug (#23091), fix test (#24376), and prep for custom op matching (#24604) (#24542)

  Show description for d5e0fca
  ProExpertProg

  authored
  d5e0fca

  Copy full SHA for d5e0fca

Commits on Aug 27, 2025

• [BugFix] Fix topk_softmax assert (#19764)

  Show description for 4f35be1
  ProExpertProg

  authored
  4f35be1

  Copy full SHA for 4f35be1

Commits on Aug 13, 2025

• Move checklist in PR template (#22852)

  Show description for 4e8614e
  ProExpertProg

  authored
  4e8614e

  Copy full SHA for 4e8614e

Commits on Jul 11, 2025

• [Bugfix] Upgrade depyf to 0.19 and streamline custom pass logging (#20777)

  Show description for 762be26
  ProExpertProg

  authored
  762be26

  Copy full SHA for 762be26

• [Perf][fp8] Use CustomOp abstraction for fp8 quant for better perf (#19830)

  Show description for 31d5c17
  

  ProExpertProg

  and

  mgoin

  authored
  31d5c17

  Copy full SHA for 31d5c17

Commits on Jul 1, 2025

• [CLI] Improve CLI arg parsing for `-O`/`--compilation-config` (#20156)

  Show description for 6d42ce8
  ProExpertProg

  authored
  6d42ce8

  Copy full SHA for 6d42ce8

Commits on Jun 27, 2025

• [Fix][torch.compile] Enable custom ops by default when Inductor off (#20102)

  Show description for aafabaa
  ProExpertProg

  authored
  aafabaa

  Copy full SHA for aafabaa

Commits on Jun 13, 2025

• [CUDA] Enable full cudagraph for FlashMLA (#18581)

  Show description for 3597b06
  ProExpertProg

  authored
  3597b06

  Copy full SHA for 3597b06

Commits on Jun 12, 2025

• [torch.compile][ROCm] Fuse quantization onto attention using a torch.compile pass (#16756)

  Show description for f98548b
  

  ProExpertProg

  and

  SageMoore

  authored
  f98548b

  Copy full SHA for f98548b

Commits on Jun 7, 2025

• [BugFix][V1] Fix memory profiling bug (#18974)

  Show description for 2d8476e
  ProExpertProg

  authored
  2d8476e

  Copy full SHA for 2d8476e

Commits on May 31, 2025

• [BugFix] Pydantic part 2 (#18911)

  Show description for c55d804
  ProExpertProg

  authored
  c55d804

  Copy full SHA for c55d804

Commits on May 29, 2025

• [BugFix] Update pydantic to fix error on python 3.10 (#18852)

  Show description for f274581
  ProExpertProg

  authored
  f274581

  Copy full SHA for f274581

Commits on May 28, 2025

• [Build] Fixes for CMake install (#18570)

  ProExpertProg

  authored
  a3896c7

  Copy full SHA for a3896c7

Commits on May 15, 2025

• [Fix][ROCm] Enforce eager for all encoder-decoder models on ROCm (#18154)

  Show description for 83f74c6
  ProExpertProg

  authored
  83f74c6

  Copy full SHA for 83f74c6

Commits on May 14, 2025

• [Fix] Support CUDAGraph capture for encoder-decoder on ROCm (#18104)

  Show description for 176a95c
  ProExpertProg

  authored
  176a95c

  Copy full SHA for 176a95c

Commits on Apr 18, 2025

• [ROCm] [Attention] Cleanup ROCm output passing (#16431)

  Show description for aaec845
  ProExpertProg

  authored
  aaec845

  Copy full SHA for aaec845

Commits on Apr 9, 2025

• [BugFix] Fix fusion test and add them to CI (#16287)

  Show description for 9cdde47
  ProExpertProg

  authored
  9cdde47

  Copy full SHA for 9cdde47

Commits on Mar 28, 2025

• [Bugfix] [torch.compile] Add Dynamo metrics context during compilation (#15639)

  Show description for 04437e3
  ProExpertProg

  authored
  04437e3

  Copy full SHA for 04437e3

Commits on Mar 24, 2025

• [Fix] [torch.compile] Improve UUID system for custom passes (#15249)

  Show description for f622dbc
  ProExpertProg

  authored
  f622dbc

  Copy full SHA for f622dbc

Commits on Mar 7, 2025

• [FP8] Refactor apply_fp8_linear and apply_fp8_linear_generic into an object (#14390)

  Show description for e174450
  ProExpertProg

  authored
  e174450

  Copy full SHA for e174450

Commits on Feb 28, 2025

• [torch.compile] Fix RMSNorm + quant fusion in the non-cutlass-fp8 case, rename RedundantReshapesPass to NoopEliminationPass (#10902)

  Show description for bd56c98
  ProExpertProg

  authored
  bd56c98

  Copy full SHA for bd56c98

Commits on Dec 13, 2024

• [torch.compile] Dynamic fp8 + rms_norm fusion (#10906)

  Show description for 30870b4
  

  ProExpertProg

  and

  Varun Sundar Rabindranath

  authored
  30870b4

  Copy full SHA for 30870b4

Commits on Nov 21, 2024

• [torch.compile] Inductor code caching fix (#10273)

  Show description for 8b0fe06
  ProExpertProg

  authored
  8b0fe06

  Copy full SHA for 8b0fe06

Pagination

Previous

Next