All pull requests

Pull requests

• gpu: skip optimize_module in eager compile mode

  #5301·vmilanov-amd opened Sep 23, 2026Collaborator·

  ·

• chore: add security scanning workflows

  #5299·haribabug opened Sep 22, 2026·

  ·
  3

• Fix hipExtModuleLaunchKernel C linkage on Linux

  #5298·kentqian opened Sep 22, 2026Contributor·

  ·
  1

  

• Suggest license changes as well

  #5295·pfultz2 opened Sep 21, 2026Collaborator·

  ·
  5

• Only to test rocmlirTriton PR#510: Reject unsupported FP8 split-K atomics

  #5294·ahsan-ca opened Sep 21, 2026Contributor·

  ·
  1

  

• Propagate broadcast in binary ops

  #5291·pfultz2 opened Sep 21, 2026Collaborator·

  ·
  4

• Bump rocm-docs-core from 1.40.2 to 1.41.0 in /docs/sphinx

  documentation
  #5290·dependabot[bot] opened Sep 21, 2026Bot·

  ·

• Add missing tests

  #5289·pfultz2 opened Sep 20, 2026Collaborator·

  ·
  4

• [AIMIGRAPHX-1266] GQA fails when the operator has 12 inputs

  #5287·pfultz2 opened Sep 20, 2026Collaborator·

  ·
  5

• [AIMIGRAPHX-1264] QMoE Op not supported

  #5286·pfultz2 opened Sep 20, 2026Collaborator·

  ·
  2

• Update mlir problem key with a md5 sum

  #5284·pfultz2 opened Sep 18, 2026Collaborator·

  ·
  5

• Limit reshape output fusion in mlir

  #5283·pfultz2 opened Sep 18, 2026Collaborator·

  ·
  6

• Onnxruntime Weekly Sync 2026-09-18

  automateddependenciesonnxruntimeskip bot checks
  #5282·github-actions[bot] opened Sep 18, 2026Bot·

  ·
  

• Bump soupsieve from 2.8.4 to 2.9 in /docs/sphinx

  documentation
  #5281·dependabot[bot] opened Sep 18, 2026Bot·

  ·

• Fix horizontal fusion of dependent operations by partitioning key groups into independent subgroups

  #5280·pfultz2 opened Sep 18, 2026Collaborator·

  ·
  4

• Refactor adaptive benchmarking

  #5279·pfultz2 opened Sep 18, 2026Collaborator·

  ·
  15

• Share identical weight literals across programs on the same device

  #5278·aditya-dl opened Sep 17, 2026Collaborator·

  ·

• Fix GRU/LSTM accuracy regression in horizontal fusion

  high priority
  #5276·umangyadav opened Sep 17, 2026Member·

  ·
  13

• Winograd fp16 gfx11

  #5274·pfultz2 opened Sep 17, 2026Collaborator·

  ·

• Add fused M=1 INT4 GEMV kernel for LLM decode (opt-in)

  #5272·aditya-dl opened Sep 16, 2026Collaborator·

  ·

• Fuse reduce slices and squeezed epilogues, tile short reductions, and match attention output through view ops

  llm
  #5271·pfultz2 opened Sep 16, 2026Collaborator·

  ·
  7

• Cache HIP code object compilation

  #5270·dhernandez0 opened Sep 16, 2026Contributor·

  ·
  15

  

• Symbolic and dynamic shape Cpp & Python printing

  #5269·CharlieL7 opened Sep 15, 2026Collaborator·

  ·
  18

• Accept 2D MatMulNBits zero points

  #5266·ghedo opened Sep 15, 2026·

  ·
  5

• Opt int4 2

  #5264·pfultz2 opened Sep 14, 2026Collaborator·

  ·