.github/workflows/windows-ort.yml: upgrade to ort 1.23.0 and cuda 12.9.1

Author: WolframRhodium

.github/workflows/windows-ort.yml

@@ -20,7 +20,7 @@ on:
 
 jobs:
   build-windows:
-    runs-on: windows-2022
+    runs-on: windows-2025
 
     defaults:
       run:
@@ -44,7 +44,7 @@ jobs:
       uses: actions/cache@v4
       with:
         path: vsort/protobuf/install
-        key: ${{ runner.os }}-vsort-protobuf-v4
+        key: ${{ runner.os }}-vsort-protobuf-v5
 
     - name: Checkout protobuf
       uses: actions/checkout@v4
@@ -53,9 +53,10 @@ jobs:
         repository: protocolbuffers/protobuf
         # follows protobuf in https://github.com/AmusementClub/onnxruntime/blob/master/cmake/external/onnxruntime_external_deps.cmake#L203
         # if you change this, remember to bump the version of the cache key.
-        ref: v3.21.12
+        ref: v5.29.2
         fetch-depth: 1
         path: vsort/protobuf
+        submodules: 'recursive'
 
     - name: Configure protobuf
       if: steps.cache-protobuf.outputs.cache-hit != 'true'
@@ -76,7 +77,7 @@ jobs:
       uses: actions/cache@v4
       with:
         path: vsort/onnx/install
-        key: ${{ runner.os }}-vsort-onnx-v5
+        key: ${{ runner.os }}-vsort-onnx-v6
 
     - name: Checkout onnx
       if: steps.cache-onnx.outputs.cache-hit != 'true'
@@ -85,7 +86,7 @@ jobs:
         repository: onnx/onnx
         # follows onnx in https://github.com/AmusementClub/onnxruntime/tree/master/cmake/external
         # if you change this, remember to bump the version of the cache key.
-        ref: 990217f043af7222348ca8f0301e17fa7b841781
+        ref: v1.19.0
         fetch-depth: 1
         path: vsort/onnx
 
@@ -116,21 +117,21 @@ jobs:
 
     - name: Download ONNX Runtime Precompilation
       run: |
-        curl -s -o ortgpu.zip -LJO https://github.com/AmusementClub/onnxruntime/releases/download/orttraining_rc2-8036-geb41d57f21-240425-0428/onnxruntime-gpu-win64.zip
+        curl -s -o ortgpu.zip -LJO https://github.com/AmusementClub/onnxruntime/releases/download/orttraining_rc2-10488-g4754a1d64e-250830-1356/onnxruntime-gpu-win64.zip
         unzip -q ortgpu.zip
 
     - name: Cache CUDA
       id: cache-cuda
       uses: actions/cache@v4
       with:
         path: C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA
-        key: ${{ runner.os }}-cuda-12.4.1
+        key: ${{ runner.os }}-cuda-12.9.1
 
     - name: Setup CUDA
       if: steps.cache-cuda.outputs.cache-hit != 'true'
       run: |
-        curl -s -o cuda_installer.exe -L https://developer.download.nvidia.com/compute/cuda/12.4.1/network_installers/cuda_12.4.1_windows_network.exe
-        cuda_installer.exe -s nvcc_12.4 cudart_12.4
+        curl -s -o cuda_installer.exe -L https://developer.download.nvidia.com/compute/cuda/12.9.1/network_installers/cuda_12.9.1_windows_network.exe
+        cuda_installer.exe -s nvcc_12.9 cudart_12.9
 
     - name: Configure
       run: cmake -S . -B build -G Ninja -LA
@@ -142,7 +143,7 @@ jobs:
         -D ONNX_RUNTIME_API_DIRECTORY=onnxruntime-gpu\include\onnxruntime
         -D ONNX_RUNTIME_LIB_DIRECTORY=onnxruntime-gpu\lib
         -D ENABLE_CUDA=1
-        -D CUDAToolkit_ROOT="C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.4"
+        -D CUDAToolkit_ROOT="C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.9"
         -D ENABLE_DML=1
         -D CMAKE_CXX_STANDARD=20
 
@@ -161,7 +162,7 @@ jobs:
     - name: Download DirectML Library
       # follows DirectML in https://github.com/AmusementClub/onnxruntime/blob/master/cmake/external/dml.cmake#L44
       run: |
-        curl -s -o directml.nupkg -LJO https://www.nuget.org/api/v2/package/Microsoft.AI.DirectML/1.14.1
+        curl -s -o directml.nupkg -LJO https://www.nuget.org/api/v2/package/Microsoft.AI.DirectML/1.15.4
         unzip -q directml.nupkg -d dml
         copy dml\bin\x64-win\DirectML.dll artifact\vsort\