ONNX Runtime Windows GPU Doc Gen CI

Workflow runs for ONNX Runtime Windows GPU Doc Gen CI

• [CUDA] Add fast INT3 QMoE kernels and offline SM80 INT2/INT4 prepack …

  currently running:
  ONNX Runtime Windows GPU Doc Gen CI#11433:Commitbd6010cpushedbyjiafatom
  main

  Oct 25, 2026In progress
  More actions for workflow run 11433

• [CUDA] Add SwiGLU contrib operator and SwiGLU fusion

  currently running:
  ONNX Runtime Windows GPU Doc Gen CI#11432:Pull request#32126synchronizebytianleiwu
  tlwu/20260817/moe_router_swiglu

  Oct 25, 2026In progress
  More actions for workflow run 11432

• [CUDA] Add BFloat16 ArgMax/ArgMin support, extend GatedAdd, and add SigmoidGateMul

  currently running:
  ONNX Runtime Windows GPU Doc Gen CI#11431:Pull request#33470synchronizebytianleiwu
  asonawane/main-ops-hyperconnection-tensor

  Oct 25, 2026In progress
  More actions for workflow run 11431

• [CUDA] Add SpeculativePrefixCommit and SpeculativeStepPublish contrib operators

  currently running:
  ONNX Runtime Windows GPU Doc Gen CI#11430:Pull request#33474synchronizebytianleiwu
  asonawane/main-ops-mtp-fold-state

  Oct 25, 2026In progress
  More actions for workflow run 11430

• [CUDA] Optimize HyperConnectionLowRankMix 4-row decode, rank-320 specialization, and deterministic prefill

  completed successfully:
  ONNX Runtime Windows GPU Doc Gen CI#11429:Pull request#33548synchronizebytianleiwu
  asonawane/dflash2-s4-hyperconnection

  Oct 24, 20262h 18m 4s
  More actions for workflow run 11429

• Optimize quantized MoE and block-scaled FP8 CUDA projections

  failed:
  ONNX Runtime Windows GPU Doc Gen CI#11428:Pull request#33473synchronizebytianleiwu
  asonawane/main-ops-quantized-moe-fp8

  Oct 24, 20262h 27m 7s
  More actions for workflow run 11428

• Optimize quantized MoE and block-scaled FP8 CUDA projections

  cancelled:
  ONNX Runtime Windows GPU Doc Gen CI#11427:Pull request#33473synchronizebytianleiwu
  asonawane/main-ops-quantized-moe-fp8

  Oct 24, 20266m 10s
  More actions for workflow run 11427

• Add exact normalization and rotary CUDA operator modes (#33469)

  completed successfully:
  ONNX Runtime Windows GPU Doc Gen CI#11426:Commitc787328pushedbyapsonawane
  main

  Oct 24, 20262h 23m 41s
  More actions for workflow run 11426

• Optimize quantized MoE and block-scaled FP8 CUDA projections

  cancelled:
  ONNX Runtime Windows GPU Doc Gen CI#11425:Pull request#33473synchronizebytianleiwu
  asonawane/main-ops-quantized-moe-fp8

  Oct 24, 202641m 32s
  More actions for workflow run 11425

• [CUDA] Add fast INT3 QMoE kernels and offline SM80 INT2/INT4 prepack support

  completed successfully:
  ONNX Runtime Windows GPU Doc Gen CI#11424:Pull request#33569synchronizebytianleiwu
  tlwu/20261023/qmoe-int3-int2-prepack-cuda

  Oct 24, 20262h 31m 39s
  More actions for workflow run 11424

• [CUDA/CPU] Add INT4 and INT3 QMoE CPU/GPU offload support

  completed successfully:
  ONNX Runtime Windows GPU Doc Gen CI#11423:Pull request#33575stackedbytianleiwu
  tlwu/20261024/qmoe-offload-int4-int3

  Oct 24, 20262h 27m 58s
  More actions for workflow run 11423

• [CUDA] Extend DynamicSparseAttention INT8 KV cache and PackedSparseAttentionIndexer QSA

  completed successfully:
  ONNX Runtime Windows GPU Doc Gen CI#11422:Pull request#33472synchronizebytianleiwu
  asonawane/main-ops-sparse-attention-indexer

  Oct 24, 20262h 26m 57s
  More actions for workflow run 11422

• [CUDA] Optimize HyperConnectionLowRankMix 4-row decode, rank-320 specialization, and deterministic prefill

  completed successfully:
  ONNX Runtime Windows GPU Doc Gen CI#11421:Pull request#33548synchronizebytianleiwu
  asonawane/dflash2-s4-hyperconnection

  Oct 24, 20262h 23m 31s
  More actions for workflow run 11421

• [CUDA] Add fast INT3 QMoE kernels and offline SM80 INT2/INT4 prepack support

  cancelled:
  ONNX Runtime Windows GPU Doc Gen CI#11420:Pull request#33569synchronizebytianleiwu
  tlwu/20261023/qmoe-int3-int2-prepack-cuda

  Oct 24, 20261h 25m 37s
  More actions for workflow run 11420

• Fix MatMulNBits benchmark model IR compatibility

  requires action with the application:
  ONNX Runtime Windows GPU Doc Gen CI#11419:Pull request#33577openedbyarcusbuilds
  arcusbuilds:fix/matmulnbits-profile-model-ir

  Oct 24, 2026Action required
  More actions for workflow run 11419

• Add INT3 QMoE CUDA execution and versioned private prepack

  completed successfully:
  ONNX Runtime Windows GPU Doc Gen CI#11418:Pull request#33555synchronizebyjiafatom
  jiafa/int3-qmoe-pr2-pr3

  Oct 24, 20262h 24m 35s
  More actions for workflow run 11418

• [CUDA] Extend DynamicSparseAttention INT8 KV cache and PackedSparseAttentionIndexer QSA

  completed successfully:
  ONNX Runtime Windows GPU Doc Gen CI#11417:Pull request#33472synchronizebytianleiwu
  asonawane/main-ops-sparse-attention-indexer

  Oct 24, 20262h 19m 8s
  More actions for workflow run 11417

• [CUDA] Extend GatedDeltaNet and CausalConvWithState for S<=8 decode and prefix states

  completed successfully:
  ONNX Runtime Windows GPU Doc Gen CI#11416:Pull request#33471synchronizebytianleiwu
  asonawane/main-ops-gdn-convolution

  Oct 24, 20262h 22m 3s
  More actions for workflow run 11416

• [CUDA] Add SpeculativePrefixCommit and SpeculativeStepPublish contrib operators

  completed successfully:
  ONNX Runtime Windows GPU Doc Gen CI#11415:Pull request#33474synchronizebytianleiwu
  asonawane/main-ops-mtp-fold-state

  Oct 24, 20262h 15m 42s
  More actions for workflow run 11415

• [CUDA] Optimize HyperConnectionLowRankMix 4-row decode, rank-320 specialization, and deterministic prefill

  completed successfully:
  ONNX Runtime Windows GPU Doc Gen CI#11414:Pull request#33548synchronizebytianleiwu
  asonawane/dflash2-s4-hyperconnection

  Oct 24, 20262h 18m 35s
  More actions for workflow run 11414

• [CUDA] Add fast INT3 QMoE kernels and offline SM80 INT2/INT4 prepack support

  completed successfully:
  ONNX Runtime Windows GPU Doc Gen CI#11413:Pull request#33569synchronizebytianleiwu
  tlwu/20261023/qmoe-int3-int2-prepack-cuda

  Oct 24, 20262h 23m 35s
  More actions for workflow run 11413

• Add 2-bit SQNBIT_CompFp32 support to MlasQNBitGemm (MatMulNBits bits=2, x64)

  failed:
  ONNX Runtime Windows GPU Doc Gen CI#11412:Pull request#33573openedbyRipefog
  Ripefog:q2bit-compfp32

  Oct 24, 202610m 34s
  More actions for workflow run 11412

• [CUDA] Add portable QMoE CPU/GPU offload backend with doorbell decode, adaptive promotion, and DMA prefill

  completed successfully:
  ONNX Runtime Windows GPU Doc Gen CI#11411:Pull request#33452synchronizebytianleiwu
  tlwu/20261019/qmoe-offload-p3-cuda-operator

  Oct 24, 20262h 16m 38s
  More actions for workflow run 11411

• [CUDA] Add portable QMoE CPU/GPU offload backend with doorbell decode, adaptive promotion, and DMA prefill

  cancelled:
  ONNX Runtime Windows GPU Doc Gen CI#11410:Pull request#33452synchronizebytianleiwu
  tlwu/20261019/qmoe-offload-p3-cuda-operator

  Oct 24, 202625m 42s
  More actions for workflow run 11410

• [WebGPU] Rename same-named sources to avoid MSVC SARIF log collisions

  completed successfully:
  ONNX Runtime Windows GPU Doc Gen CI#11409:Pull request#33572openedbyadrastogi
  adrastogi/webgpu-unique-source-basenames

  Oct 24, 20262h 33m 59s
  More actions for workflow run 11409