All pull requests

Pull requests

• [CUDA] Add QMoEOffload operator with doorbell CUDA-graph decode, cross-layer promotion, and DMA prefill

  #33452·Tianlei Wu (tianleiwu) opened Oct 19, 2026Contributor·3 tasks done·

  ·

• [CUDA] Add session-scoped QMoE offload lifecycle, device memory ledger, and arena shrink hooks

  #33451·Tianlei Wu (tianleiwu) opened Oct 19, 2026Contributor·3 tasks done·

  ·

• [CPU/MLAS] Add QMoE shared 4-bit expert layouts, AVX-512 kernels, and FileExpertStore

  #33450·Tianlei Wu (tianleiwu) opened Oct 19, 2026Contributor·3 tasks done·

  ·
  6

• [WebGPU] Fix FP16 MatMulNBits wide-tile results on Qualcomm D3D12

  #33449·Yang Gu (gyagp) opened Oct 19, 2026Contributor·

  ·
  1

• [WebGPU] Generate component inventories from fetched Dawn dependencies

  #33448·Yang Gu (gyagp) opened Oct 19, 2026Contributor·

  ·
  1

• Add EP device memory capacity query API with CUDA support

  #33447·Selena Yang (selenayang888) opened Oct 19, 2026Contributor·

  ·

• Speed up CPU 1D depthwise ConvTranspose

  #33446· Stern (Hosi121) opened Oct 18, 2026Contributor·

  ·
  7

• Fix: [Bug][CPU] TransposeOptimizer inserts an unsupported INT32 QuantizeLinear after valid dequantization

  #33445·SIDDARTHA REDDY (SIDDARTHAREDDY8) opened Oct 18, 2026Contributor·

  ·
  1

  4

• fix(optimizer): preserve DequantizeLinear before ArgMax when scale is non-positive (#33440)

  #33444·Prashanth (dpr2106) opened Oct 18, 2026·

  ·
  1

  4

• FreeBSD support in onnxruntime/core/platform/posix/env.cc

  #33433·Yuri Victorovich (yurivict) opened Oct 18, 2026Contributor·

  ·
  2

• Fix BatchNormalization CPU return NaN for infinte scale

  #33428·Ishan Undre (ishanundre2004) opened Oct 17, 2026·

  ·
  1

  6

• Fix incorrect bias scaling in MatMulIntegerToFloat fusion

  #33425·Shoryamishra61 opened Oct 17, 2026·

  ·
  1

  4

• GQA: fall back from auto-enabled cuDNN SDPA when the graph can't be built

  #33424·Dante (KafKafrnZ) opened Oct 17, 2026·

  ·
  1

  10

• Preserve score and layout semantics in Python attention fusion

  #33423·Sherman Gaucia Laing (0-Actual) opened Oct 17, 2026·

  ·
  2

• Remove Rust subtree

  #33422·Carson M. (decahedron1) opened Oct 17, 2026Contributor·

  ·
  3

• Preserve reduction and constant semantics in Python LayerNormalization fusion

  #33421·Sherman Gaucia Laing (0-Actual) opened Oct 17, 2026·

  ·
  3

• [CUDA] Tie GQA phase admission to runtime classification

  #33420·Chi Lo (chilo-ms) opened Oct 17, 2026Contributor·

  ·

• Avoid module-load dynamic initializer for OrtEnv::m_

  #33419·Mustapha Jaber (mustjab) opened Oct 16, 2026Contributor·

  ·
  4

  

• [CUDA] QMoE: honour router_weights and allow top-k 9 single-token decode GEMV

  model:nemotron
  #33418·Tianlei Wu (tianleiwu) opened Oct 16, 2026Contributor·2 of 3 tasks·

  ·
  11

• [WebGPU] Fix weight load acceleration so that weight load acceleration off session followed by on works

  #33417·Sushanth Rajasankar (sushraja-msft) opened Oct 16, 2026Contributor·

  ·
  1

• Wasm without Emscripten

  #33416·Mendy Berger (MendyBerger) opened Oct 16, 2026·

  ·
  9

• Fix dynamic Gemm quantization with annotated or shared weights

  #33415·Ge Li (galleonli) opened Oct 16, 2026Contributor·

  ·
  1

  6

• Parallelize gemm operations in xnnpack provider

  #33413·Josh Minor (jishminor) opened Oct 16, 2026·

  ·
  3

• Specify MIT license expression in NuGet packages

  #33411·Federico Fissore (ffissore) opened Oct 16, 2026·

  ·
  3

• [WebGPU] Move copy implementation into data transfer

  #33410·Xiaofei Han (xiaofeihan1) opened Oct 16, 2026Contributor·

  ·
  2