Review comments

.github/workflows/release.yaml

@@ -183,10 +183,17 @@ jobs:
           name: windows-rocm-deps
           path: dist/deps/*
 
-  # CUDA v11 generation step
-  generate-windows-cuda-v11:
+  # CUDA generation step
+  generate-windows-cuda:
     environment: release
     runs-on: windows
+    strategy:
+      matrix:
+        cuda:
+          - version: "11"
+            url: 'https://developer.download.nvidia.com/compute/cuda/11.3.1/local_installers/cuda_11.3.1_465.89_win10.exe'
+          - version: "12"
+            url: 'https://developer.download.nvidia.com/compute/cuda/12.4.0/local_installers/cuda_12.4.0_551.61_windows.exe'
     env:
       KEY_CONTAINER: ${{ vars.KEY_CONTAINER }}
     steps:
@@ -220,11 +227,11 @@ jobs:
         with:
           go-version-file: go.mod
           cache: true
-      - name: 'Install CUDA'
+      - name: 'Install CUDA ${{ matrix.cuda.version }}'
         run: |
           $ErrorActionPreference = "Stop"
           write-host "downloading CUDA Installer"
-          Invoke-WebRequest -Uri "https://developer.download.nvidia.com/compute/cuda/11.3.1/local_installers/cuda_11.3.1_465.89_win10.exe" -OutFile "${env:RUNNER_TEMP}\cuda-install.exe"
+          Invoke-WebRequest -Uri "${{ matrix.cuda.url }}" -OutFile "${env:RUNNER_TEMP}\cuda-install.exe"
           write-host "Installing CUDA"
           Start-Process "${env:RUNNER_TEMP}\cuda-install.exe" -ArgumentList '-s' -NoNewWindow -Wait
           write-host "Completed CUDA"
@@ -256,7 +263,7 @@ jobs:
           cp "${NVIDIA_DIR}\cublasLt64_*.dll" "dist\deps\"
       - uses: actions/upload-artifact@v4
         with:
-          name: generate-windows-cuda-v11
+          name: generate-windows-cuda-${{ matrix.cuda.version }}
           path: |
             llm/build/**/bin/*
             dist/windows-amd64/**
@@ -265,95 +272,13 @@ jobs:
           name: windows-cuda-deps
           path: dist/deps/*
 
-  # CUDA v12 generation step
-  generate-windows-cuda-v12:
-    environment: release
-    runs-on: windows
-    env:
-      KEY_CONTAINER: ${{ vars.KEY_CONTAINER }}
-    steps:
-      - uses: actions/checkout@v4
-      - name: Set Version
-        shell: bash
-        run: echo "VERSION=${GITHUB_REF_NAME#v}" >> $GITHUB_ENV
-      - uses: 'google-github-actions/auth@v2'
-        with:
-          project_id: 'ollama'
-          credentials_json: '${{ secrets.GOOGLE_SIGNING_CREDENTIALS }}'
-      - run: echo "${{ vars.OLLAMA_CERT }}" > ollama_inc.crt
-      - name: install Windows SDK 8.1 to get signtool
-        run: |
-          $ErrorActionPreference = "Stop"
-          write-host "downloading SDK"
-          Invoke-WebRequest -Uri "https://go.microsoft.com/fwlink/p/?LinkId=323507" -OutFile "${env:RUNNER_TEMP}\sdksetup.exe"
-          Start-Process "${env:RUNNER_TEMP}\sdksetup.exe" -ArgumentList @("/q") -NoNewWindow -Wait
-          write-host "Win SDK 8.1 installed"
-          gci -path 'C:\Program Files (x86)\Windows Kits\' -r -fi 'signtool.exe'
-      - name: install signing plugin
-        run: |
-          $ErrorActionPreference = "Stop"
-          write-host "downloading plugin"
-          Invoke-WebRequest -Uri "https://github.com/GoogleCloudPlatform/kms-integrations/releases/download/cng-v1.0/kmscng-1.0-windows-amd64.zip" -OutFile "${env:RUNNER_TEMP}\plugin.zip"
-          Expand-Archive -Path "${env:RUNNER_TEMP}\plugin.zip" -DestinationPath ${env:RUNNER_TEMP}\plugin\
-          write-host "Installing plugin"
-          & "${env:RUNNER_TEMP}\plugin\*\kmscng.msi" /quiet
-          write-host "plugin installed"
-      - uses: actions/setup-go@v5
-        with:
-          go-version-file: go.mod
-          cache: true
-      - name: 'Install CUDA'
-        run: |
-          $ErrorActionPreference = "Stop"
-          write-host "downloading CUDA Installer"
-          Invoke-WebRequest -Uri "https://developer.download.nvidia.com/compute/cuda/12.4.0/local_installers/cuda_12.4.0_551.61_windows.exe" -OutFile "${env:RUNNER_TEMP}\cuda-install.exe"
-          write-host "Installing CUDA"
-          Start-Process "${env:RUNNER_TEMP}\cuda-install.exe" -ArgumentList '-s' -NoNewWindow -Wait
-          write-host "Completed CUDA"
-          $cudaPath=((resolve-path "c:\Program Files\NVIDIA*\CUDA\v*\bin\nvcc.exe")[0].path | split-path | split-path)
-          $cudaVer=($cudaPath | split-path -leaf ) -replace 'v(\d+).(\d+)', '$1_$2' 
-          echo "$cudaPath\bin" >> $env:GITHUB_PATH
-          echo "CUDA_PATH=$cudaPath" >> $env:GITHUB_ENV
-          echo "CUDA_PATH_V${cudaVer}=$cudaPath" >> $env:GITHUB_ENV
-          echo "CUDA_PATH_VX_Y=CUDA_PATH_V${cudaVer}" >> $env:GITHUB_ENV
-      - name: 'Verify CUDA'
-        run: nvcc -V
-      - run: go get ./...
-      - name: go generate
-        run: |
-          $gopath=(get-command go).source | split-path -parent
-          $cudabin=(get-command nvcc).source | split-path
-          & "C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise\Common7\Tools\Launch-VsDevShell.ps1"
-          cd $env:GITHUB_WORKSPACE
-          $env:CMAKE_SYSTEM_VERSION="10.0.22621.0"
-          $env:PATH="$gopath;$cudabin;$env:PATH"
-          $env:OLLAMA_SKIP_CPU_GENERATE="1"
-          go generate -x ./...
-      - name: 'gather cuda dependencies'
-        run: |
-          $NVIDIA_DIR=(resolve-path 'C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\*\bin\')[0]
-          md "dist\deps"
-          cp "${NVIDIA_DIR}\cudart64_*.dll" "dist\deps\"
-          cp "${NVIDIA_DIR}\cublas64_*.dll" "dist\deps\"
-          cp "${NVIDIA_DIR}\cublasLt64_*.dll" "dist\deps\"
-      - uses: actions/upload-artifact@v4
-        with:
-          name: generate-windows-cuda-v12
-          path: |
-            llm/build/**/bin/*
-            dist/windows-amd64/**
-      - uses: actions/upload-artifact@v4
-        with:
-          name: windows-cuda-deps
-          path: dist/deps/*
 
   # Import the prior generation steps and build the final windows assets
   build-windows:
     environment: release
     runs-on: windows
     needs:
-      - generate-windows-cuda-v11
-      - generate-windows-cuda-v12
+      - generate-windows-cuda
       - generate-windows-rocm
       - generate-windows-cpu
     env:
@@ -397,7 +322,10 @@ jobs:
           name: generate-windows-cpu
       - uses: actions/download-artifact@v4
         with:
-          name: generate-windows-cuda-v11
+          name: generate-windows-cuda-11
+      - uses: actions/download-artifact@v4
+        with:
+          name: generate-windows-cuda-12
       - uses: actions/download-artifact@v4
         with:
           name: windows-cuda-deps