Skip to content
Closed
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
20 commits
Select commit Hold shift + click to select a range
1988019
Make zero-input-channel NCHWc regression graph-only (#33398)
chilo-ms Oct 15, 2026
327d11a
[CUDA] Bound GQA non-windowed workspace via the KV-length envelope (r…
chilo-ms Oct 15, 2026
ca64594
Revert "webgpu: Fix Python factory selection for plugin EP hosts (#33…
qjia7 Oct 16, 2026
5e42959
fix: name the activation the WebGPU Conv kernel cannot fuse (#32716) …
ykhrustalev Oct 16, 2026
2fd0afa
[CUDA] Support offline NVFP4 QMoE weights and fused block-FP8 executi…
kunal-vaishnavi Oct 16, 2026
f33c174
[CUDA][WebGPU] Add PackedSparseAttentionIndexerMerge for MTP index sh…
kunal-vaishnavi Oct 16, 2026
52131d9
[WebGPU] Optimize subgroup matrix MatMulNBits for AMD wave64 (#32796)
gyagp Oct 16, 2026
a8fda2f
Fix d3d12_file_buffer_loader so it can build with no exception and cl…
sushraja-msft Oct 16, 2026
b016c16
[CUDA] Optimize packed QSA scoring and multi-token verification (#33328)
kunal-vaishnavi Oct 16, 2026
ec70cae
[Framework] Honor Level-1 estimates for nodes missing from memory pro…
chilo-ms Oct 16, 2026
529efb6
[docs] Experimental C API usage (#29701)
edgchen1 Oct 16, 2026
0fb935a
Bump github/issue-labeler from 3.4 to 3.5 (#33388)
dependabot[bot] Oct 16, 2026
454b3dd
Fix QLinearSoftmax output for empty inputs (#33401)
DavidLuong98 Oct 16, 2026
0767b73
[WebGPU] Upgrade Dawn to the latest version (#33390)
Jiawei-Shao Oct 16, 2026
e88538c
Zero initialize OrtEpFactory in provider bridge EP factories (#33117)
maxim-davgalev Oct 16, 2026
d8a3445
[WebGPU] Add GatedDeltaNetStateReplay contrib operator (#33213)
aciddelgado Oct 17, 2026
fed1f4b
[CUDA] Load fpA_intB tactic cache on session create and save on close…
tianleiwu Oct 17, 2026
b5375da
[WebGPU] Extend FP16 MatMul GEMV fast path to M=2–8 (#33351)
aciddelgado Oct 17, 2026
c64c9fd
Fix 1DS telemetry deadlock during Windows process termination (#33291)
qjia7 Oct 17, 2026
5c8bb04
Merge remote-tracking branch 'origin/master' into sync_msft_18102026
AIFrameworksIntegration Oct 17, 2026
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion .github/workflows/labeler.yml
Original file line number Diff line number Diff line change
Expand Up @@ -15,7 +15,7 @@ jobs:
]
timeout-minutes: 120
steps:
- uses: github/issue-labeler@c1b0f9f52a63158c4adc09425e858e87b32e9685 # v3.4
- uses: github/issue-labeler@ccca1e16fcb66762200e616fc29b57b940c1cb67 # v3.5
with:
repo-token: "${{ secrets.GITHUB_TOKEN }}"
configuration-path: .github/labeler.yml
Expand Down
2 changes: 1 addition & 1 deletion .github/workflows/linux_cuda_ci.yml
Original file line number Diff line number Diff line change
Expand Up @@ -138,4 +138,4 @@ jobs:
--cmake_extra_defines onnxruntime_QUICK_BUILD=ON
--cmake_extra_defines onnxruntime_ENABLE_CUDA_EP_INTERNAL_TESTS=ON
--cmake_extra_defines onnxruntime_BUILD_CUDA_QUANT_PREPROCESS=ON
python_path_prefix: 'ORT_RUN_CUDA_TACTIC_CACHE_TESTS=1 PATH=/opt/python/cp310-cp310/bin:$PATH'
python_path_prefix: 'export ORT_RUN_CUDA_TACTIC_CACHE_TESTS=1 PATH=/opt/python/cp310-cp310/bin:$PATH'
2 changes: 1 addition & 1 deletion .github/workflows/title-only-labeler.yml
Original file line number Diff line number Diff line change
Expand Up @@ -15,7 +15,7 @@ jobs:
]
timeout-minutes: 120
steps:
- uses: github/issue-labeler@c1b0f9f52a63158c4adc09425e858e87b32e9685 # v3.4
- uses: github/issue-labeler@ccca1e16fcb66762200e616fc29b57b940c1cb67 # v3.5
with:
repo-token: "${{ secrets.GITHUB_TOKEN }}"
configuration-path: .github/title-only-labeler.yml
Expand Down
26 changes: 2 additions & 24 deletions .github/workflows/windows_webgpu.yml
Original file line number Diff line number Diff line change
Expand Up @@ -343,9 +343,9 @@ jobs:
--cmake_generator "Visual Studio 17 2022" `
--enable_onnx_tests `
--use_webgpu shared_lib `
--build_wheel `
--use_vcpkg --use_vcpkg_ms_internal_asset_cache `
--cmake_extra_defines onnxruntime_BUILD_UNIT_TESTS=ON onnxruntime_ENABLE_DAWN_BACKEND_D3D12=1 onnxruntime_ENABLE_DAWN_BACKEND_VULKAN=1 `
--disable_rtti `
--enable_lto

if ($lastExitCode -ne 0) {
Expand Down Expand Up @@ -410,7 +410,7 @@ jobs:
architecture: x64

- name: Install python modules
run: python -m pip install -r tools\ci_build\github\windows\python\requirements.txt -r plugin-ep-webgpu\python\requirements-build-wheel.txt
run: python -m pip install -r tools\ci_build\github\windows\python\requirements.txt
shell: cmd
working-directory: ${{ github.workspace }}

Expand All @@ -426,7 +426,6 @@ jobs:

- name: Run tests
shell: pwsh
working-directory: ${{ runner.temp }}
run: |
python.exe ${{ github.workspace }}\tools\python\update_ctest_path.py `
"${{ github.workspace }}\RelWithDebInfo\CTestTestfile.cmake" `
Expand All @@ -444,27 +443,6 @@ jobs:
exit $lastExitCode
}

$binaryDir = "${{ github.workspace }}\RelWithDebInfo\RelWithDebInfo"
$pluginWheelDir = "${{ runner.temp }}\webgpu-plugin-wheels"
python "${{ github.workspace }}\plugin-ep-webgpu\python\build_wheel.py" `
--binary_dir $binaryDir --version 0.0.0 --output_dir $pluginWheelDir
if ($LASTEXITCODE -ne 0) { exit $LASTEXITCODE }

# Install the host from this PR, not the minimum-runtime wheel used by package compatibility CI.
$hostWheels = @(Get-ChildItem "$binaryDir\dist\onnxruntime_webgpu-*.whl")
$pluginWheels = @(Get-ChildItem "$pluginWheelDir\onnxruntime_ep_webgpu-*.whl")
if ($hostWheels.Count -ne 1 -or $pluginWheels.Count -ne 1) {
throw "Expected exactly one source-built host wheel and one plugin wheel"
}
# Dependencies were installed once for this job in Install python modules.
python -m pip install --no-deps --force-reinstall $hostWheels[0].FullName $pluginWheels[0].FullName
if ($LASTEXITCODE -ne 0) { exit $LASTEXITCODE }

$env:ORT_WEBGPU_TEST_REQUIRE_EP_DEVICE = '1'
$env:ORT_WEBGPU_TEST_PROVIDER_NAME = '1'
python -u "${{ github.workspace }}\plugin-ep-webgpu\python\test\test_webgpu_plugin_ep.py"
if ($LASTEXITCODE -ne 0) { exit $LASTEXITCODE }

# Same plugin EP sources as webgpu_plugin_build_x64_RelWithDebInfo, but linked into the ORT binary
# instead of built as a separate shared library.
webgpu_static_plugin_build_x64_RelWithDebInfo:
Expand Down
Loading
Loading