All pull requests

Pull requests

• Bump compression from 1.8.1 to 1.8.2 in /js/react_native/e2e

  dependenciesjavascript
  #33127·dependabot[bot] opened Oct 6, 2026Bot·

  ·
  2

• Bump @vue/server-renderer and vue in /js/web/test/e2e/exports/testcases/vite-default

  dependenciesjavascript
  #33126·dependabot[bot] opened Oct 6, 2026Bot·

  ·
  2

• Bump source-map-js from 1.2.1 to 1.2.2 in /js/web/test/e2e/exports/testcases/vite-default

  dependenciesjavascript
  #33125·dependabot[bot] opened Oct 6, 2026Bot·

  ·
  2

• [CUDA] Cache cuDNN frontend Conv execution plans

  #33123·Tianlei Wu (tianleiwu) opened Oct 6, 2026Contributor·3 of 4 tasks·

  ·

• Fix WebGPU DequantizeLinear scalar/vector shader cache collisions

  #33122·Yang Gu (gyagp) opened Oct 6, 2026Contributor·

  ·
  1

  5

• [MLAS] Fix null B in SymmQgemm benchmark packing

  #33121·Yair Ben Avraham (yairbenavraham) opened Oct 6, 2026Contributor·

  ·
  2

• Allow onnxruntime_DISABLE_EXCEPTIONS without onnxruntime_MINIMAL_BUILD

  #33120·Ryan VanderMeulen (rvandermeulen) opened Oct 6, 2026Contributor·

  ·
  4

• [CUDA] Single-source remaining GQA preparation scratch sizing

  #33118·Chi Lo (chilo-ms) opened Oct 5, 2026Contributor·

  ·

• Zero initialize OrtEpFactory in provider bridge EP factories

  #33117·maxim-davgalev opened Oct 5, 2026Contributor·

  ·
  7

• Remove windows x86 workflow

  #33116·eserscor opened Oct 5, 2026Contributor·

  ·
  1

• Fix duplicate node name when QDQ fusion absorbs a redundant Clip/Relu

  #33115·Yuduo Wu (1duo) opened Oct 5, 2026Contributor·

  ·
  11

• Pin type query selector

  #33114·eserscor opened Oct 5, 2026Contributor·

  ·

• Optimize CPU BatchNormalization for NC inputs

  #33113·Master Artyom Rabosh (artyomrabosh) opened Oct 5, 2026·

  ·
  5

• Fix optional OpenVINO provider probe without exceptions

  #33112·Mustapha Jaber (mustjab) opened Oct 5, 2026Contributor·

  ·
  7

• Fix custom WebGPU context stream validation and test host-owned caches

  #33110·Mustapha Jaber (mustjab) opened Oct 5, 2026Contributor·

  ·
  8

• [CI] Use Temurin for Java API documentation

  #33108·Xavier Dupré (xadupre) opened Oct 5, 2026Member·

  ·
  1

• [WebNN EP] Restore float32 computation for decomposed SimplifiedLayerNormalization

  #33107·Hrishith Thadicherla (hthadicherla) opened Oct 5, 2026·

  ·
  4

• fix(quantization): avoid intermediate overflow when calculating quantization ranges

  #33104·Ge Li (galleonli) opened Oct 5, 2026·

  ·
  4

• [Java] Check tensor capacity before creating JNI arrays

  #33103·Xavier Dupré (xadupre) opened Oct 5, 2026Member·

  ·
  9

• Add trt_profiling_verbosity option to TensorRT EP

  #33102·Leonid Nikitin (kibenos) opened Oct 5, 2026·

  ·
  1

  6

• [CUDA] Support Offline-Prepacked NVFP4 Weights in QMoE

  #33094·kunal-vaishnavi opened Oct 5, 2026Contributor·

  ·
  3

• Fix ReshapeFusion crash when fused Reshape's shape input is node-produced

  #33091·Anubhav Agrawal (anubhavagr) opened Oct 4, 2026·

  ·
  1

  5

• [float16] Reuse Cast nodes for shared always-float inputs

  #33090·tianleiwu with CopilotAI opened Oct 4, 2026·

  ·
  1

  3

  

• Handle missing ConstantOfShape value during float16 conversion

  #33089·tianleiwu with CopilotAI opened Oct 4, 2026·

  ·
  1

  3

  

• [WebGPU] Restore ORT 1.24.4 compatibility with serial legacy execution

  #33082·Xiaofei Han (xiaofeihan1) opened Oct 3, 2026Contributor·

  ·
  3