ONNX Runtime CUDA Builds

windows_cuda.yml

2500+ workflow runs

[CUDA] Add portable QMoE CPU/GPU offload backend with doorbell decode, adaptive promotion, and DMA prefill
currently running:
ONNX Runtime CUDA Builds#26232:Pull request#33452synchronizebytianleiwu
tlwu/20261019/qmoe-offload-p3-cuda-operator
Oct 31, 2026In progress
[CUDA] Pass cached SM version to cuDNN frontend Conv graphs (#33770)
completed successfully:
ONNX Runtime CUDA Builds#26231:Commit4edd78apushedbytianleiwu
main
Oct 31, 20261h 30m 36s
[CUDA] Add portable QMoE CPU/GPU offload backend with doorbell decode, adaptive promotion, and DMA prefill
failed:
ONNX Runtime CUDA Builds#26230:Pull request#33452synchronizebytianleiwu
tlwu/20261019/qmoe-offload-p3-cuda-operator
Oct 31, 202655m 29s
[CUDA] Add portable QMoE CPU/GPU offload backend with doorbell decode, adaptive promotion, and DMA prefill
cancelled:
ONNX Runtime CUDA Builds#26229:Pull request#33452synchronizebytianleiwu
tlwu/20261019/qmoe-offload-p3-cuda-operator
Oct 31, 20263m 13s
[CUDA] Extend NVFP4 QMoE decode GEMV envelope and add 256-expert fused routing
failed:
ONNX Runtime CUDA Builds#26228:Pull request#33545synchronizebytianleiwu
tlwu/20261022/qmoe-nvfp4-256e-8row-decode
Oct 31, 202623m 11s
[WebGPU] Fuse rank-32 quantized LoRA with MatMulNBits
completed successfully:
ONNX Runtime CUDA Builds#26227:Pull request#33370synchronizebyqjia7
qjia7:feat/webgpu-matmul-nbits-qlora
Oct 31, 20261h 31m 41s
[CUDA] Add H256 BF16-query INT8-cache group-6/16 paged speculative XQ…
completed successfully:
ONNX Runtime CUDA Builds#26226:Commit5a83ed9pushedbytianleiwu
main
Oct 31, 20261h 29m 57s
Update Kusto dashboard SDK dependencies to 6.0.0
completed successfully:
ONNX Runtime CUDA Builds#26225:Pull request#33783openedbymustjab
users/mustjab/kusto-dashboard-sdk-6
Oct 31, 20261h 35m 16s
[CUDA] Add portable QMoE CPU/GPU offload backend with doorbell decode, adaptive promotion, and DMA prefill
failed:
ONNX Runtime CUDA Builds#26224:Pull request#33452synchronizebytianleiwu
tlwu/20261019/qmoe-offload-p3-cuda-operator
Oct 31, 202627m 25s
[CUDA] Add INT4 and INT3 QMoE CPU/GPU offload support
completed successfully:
ONNX Runtime CUDA Builds#26223:Pull request#33575synchronizebytianleiwu
tlwu/20261024/qmoe-offload-int4-int3
Oct 31, 20261h 30m 35s
[CUDA] Optimize INT4 QMoE offload prefill/decode and S=8 doorbell execution
completed successfully:
ONNX Runtime CUDA Builds#26222:Pull request#33622synchronizebytianleiwu
tlwu/20261026/qmoe-offload-s8-int4-opt
Oct 31, 20261h 29m 32s
[CUDA] Add 8-route grouped INT3 offload prefill, zero-cold doorbell short-circuit, and S=8 streaming optimizations
completed successfully:
ONNX Runtime CUDA Builds#26221:Pull request#33629synchronizebytianleiwu
tlwu/20261026/qmoe-offload-avx2-int3-s8-opt
Oct 31, 20261h 30m 21s
[CUDA] Add session-scoped MoE/QMoE offload planner, lifecycle control…
completed successfully:
ONNX Runtime CUDA Builds#26220:Commit649f516pushedbytianleiwu
main
Oct 31, 20261h 29m 29s
Align Windows WebGPU profiling and visualize CPU/GPU timelines
requires action with the application:
ONNX Runtime CUDA Builds#26219:Pull request#33781synchronizebygyagp
gyagp:pr-improve-profile-a82348b4-e08a-4244-9eb6-5ffa9998ddec
Oct 31, 2026Action required
fix(web): handle opaque origins for bundled file URLs
requires action with the application:
ONNX Runtime CUDA Builds#26218:Pull request#33782openedbyleninworld
leninworld:fix-web-file-origin-33735
Oct 31, 2026Action required
Align Windows WebGPU profiling and visualize CPU/GPU timelines
requires action with the application:
ONNX Runtime CUDA Builds#26217:Pull request#33781openedbygyagp
gyagp:pr-improve-profile-a82348b4-e08a-4244-9eb6-5ffa9998ddec
Oct 31, 2026Action required
Fix severe CPU STFT performance for batched short transforms
requires action with the application:
ONNX Runtime CUDA Builds#26216:Pull request#32457synchronizebythe0cp
the0cp:stft-cpu-decomposition
Oct 31, 2026Action required
[CUDA] Add session-scoped MoE/QMoE offload planner, lifecycle controller, and memory budget ledger
completed successfully:
ONNX Runtime CUDA Builds#26215:Pull request#33451synchronizebytianleiwu
tlwu/20261019/qmoe-offload-p2-cuda-memory-lifecycle
Oct 31, 20261h 33m 34s
Fix spelling typos flagged by misspell linter
completed successfully:
ONNX Runtime CUDA Builds#26214:Pull request#33780openedbyadrastogi
adityar-microsoft-fix-misspell-typos
Oct 31, 20261h 47m 51s
Make misspell lint actionable: PR-only, blocking, no false positives
completed successfully:
ONNX Runtime CUDA Builds#26213:Pull request#33779openedbyadrastogi
adityar-microsoft-fix-lint-misspell-on-push
Oct 31, 20261h 33m 34s
[WebGPU] Use segment accessors in SkipLayerNormalization (#33760)
completed successfully:
ONNX Runtime CUDA Builds#26212:Commitf72d03dpushedbyqjia7
main
Oct 31, 20261h 33m 20s
Migrate GitHub Actions away from deprecated macos-14 image
completed successfully:
ONNX Runtime CUDA Builds#26211:Pull request#33778synchronizebyedgchen1
edgchen1/update_macos_image_for_objc_static_analysis_job
Oct 31, 20261h 31m 27s
[CUDA] Add session-scoped MoE/QMoE offload planner, lifecycle controller, and memory budget ledger
cancelled:
ONNX Runtime CUDA Builds#26210:Pull request#33451synchronizebytianleiwu
tlwu/20261019/qmoe-offload-p2-cuda-memory-lifecycle
Oct 31, 202646m 45s
Migrate GitHub Actions away from deprecated macos-14 image
cancelled:
ONNX Runtime CUDA Builds#26209:Pull request#33778openedbyedgchen1
edgchen1/update_macos_image_for_objc_static_analysis_job
Oct 31, 20269m 22s
[CUDA] Add session-scoped MoE/QMoE offload planner, lifecycle controller, and memory budget ledger
failed:
ONNX Runtime CUDA Builds#26208:Pull request#33451synchronizebytianleiwu
tlwu/20261019/qmoe-offload-p2-cuda-memory-lifecycle
Oct 31, 202632m 24s

Navigation Menu

ONNX Runtime CUDA Builds

2500+ workflow runs

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

[CUDA] Pass cached SM version to cuDNN frontend Conv graphs (#33770)

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

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

[CUDA] Extend NVFP4 QMoE decode GEMV envelope and add 256-expert fused routing

[WebGPU] Fuse rank-32 quantized LoRA with MatMulNBits

[CUDA] Add H256 BF16-query INT8-cache group-6/16 paged speculative XQ…

Update Kusto dashboard SDK dependencies to 6.0.0

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

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

[CUDA] Optimize INT4 QMoE offload prefill/decode and S=8 doorbell execution

[CUDA] Add 8-route grouped INT3 offload prefill, zero-cold doorbell short-circuit, and S=8 streaming optimizations

[CUDA] Add session-scoped MoE/QMoE offload planner, lifecycle control…

Align Windows WebGPU profiling and visualize CPU/GPU timelines

fix(web): handle opaque origins for bundled file URLs

Align Windows WebGPU profiling and visualize CPU/GPU timelines

Fix severe CPU STFT performance for batched short transforms

[CUDA] Add session-scoped MoE/QMoE offload planner, lifecycle controller, and memory budget ledger

Fix spelling typos flagged by misspell linter

Make misspell lint actionable: PR-only, blocking, no false positives

[WebGPU] Use segment accessors in SkipLayerNormalization (#33760)

Migrate GitHub Actions away from deprecated `macos-14` image

[CUDA] Add session-scoped MoE/QMoE offload planner, lifecycle controller, and memory budget ledger

Migrate GitHub Actions away from deprecated `macos-14` image

[CUDA] Add session-scoped MoE/QMoE offload planner, lifecycle controller, and memory budget ledger

Workflow runs for ONNX Runtime CUDA Builds

2500+ workflow runs