infra: win11 nvidia github runner image builder

.github/workflows/vulkaninfo.yml

@@ -4,108 +4,6 @@ on:
 permissions: {}
 
 jobs:
-  # build-win11-nvidia-image:
-  #   runs-on:  generator-windows11
-  #   snapshot: win11-nvidia-grid-image
-  #   steps:
-  #     - name: Check OS version
-  #       shell: powershell
-  #       run: |
-  #         Write-Host "=== OS Version ==="
-  #         $os = Get-CimInstance Win32_OperatingSystem
-  #         Write-Host "Caption:      $($os.Caption)"
-  #         Write-Host "Version:      $($os.Version)"
-  #         Write-Host "Build Number: $($os.BuildNumber)"
-  #         Write-Host "Architecture: $($os.OSArchitecture)"
-
-  #     - name: Install Chocolatey
-  #       shell: powershell
-  #       run: |
-  #         Set-ExecutionPolicy Bypass -Scope Process -Force
-  #         [System.Net.ServicePointManager]::SecurityProtocol = [System.Net.ServicePointManager]::SecurityProtocol -bor 3072
-  #         iex ((New-Object System.Net.WebClient).DownloadString('https://community.chocolatey.org/install.ps1'))
-
-  #     - name: Install WinGet
-  #       uses: Cyberboss/install-winget@fc922bc7e0adb16f39ceab057cb2af88bd3799e1 # 1.2.0
-
-  #     - name: Install git for windows with WinGet
-  #       shell: powershell
-  #       run: winget install --id Git.Git -e --source winget --disable-interactivity --accept-package-agreements --custom '/o:PathOption=CmdTools'
-
-  #     - name: Enable Windows 11 Developer Mode via Registry
-  #       shell: powershell
-  #       run: |
-  #         $registryPath = "HKLM:\SOFTWARE\Microsoft\Windows\CurrentVersion\AppModelUnlock"
-  #         $propertyName = "AllowDevelopmentWithoutDevLicense"
-  #         $propertyValue = 1
-
-  #         # Create the registry key path if it doesn't exist
-  #         if (-not (Test-Path -Path $registryPath)) {
-  #             New-Item -Path $registryPath -ItemType Directory -Force | Out-Null
-  #         }
-
-  #         # Set the registry value to enable Developer Mode
-  #         Set-ItemProperty -Path $registryPath -Name $propertyName -Value $propertyValue -Type DWORD -Force
-
-  #         Write-Host "Developer Mode enabled successfully."
-
-  #     - name: Install latest Visual C++ Redistributable
-  #       shell: powershell
-  #       run: |
-  #         # Download the latest x64 redistributable
-  #         $vcredistUrl  = "https://aka.ms/vc14/vc_redist.x64.exe"
-  #         $vcInstaller = "$env:TEMP\vc_redist.x64.exe"
-  #         Invoke-WebRequest -Uri $vcredistUrl -OutFile $vcInstaller
-  #         Write-Host "Running silent install ..."
-  #         $proc = Start-Process -FilePath $vcInstaller `
-  #                               -ArgumentList "/install", "/quiet", "/norestart" `
-  #                               -Wait -PassThru
-  #         Write-Host "Installer exit code: $($proc.ExitCode)"
-
-  #         if ($proc.ExitCode -notin @(0, 1)) {
-  #           Write-Host "=== Windows Event Log (System, last 20 errors) ==="
-  #           Get-EventLog -LogName System -EntryType Error -Newest 20 | Format-List TimeGenerated, Source, EventID, Message
-  #           Write-Error "Visual C++ Redistributable installation failed with exit code $($proc.ExitCode)"
-  #           exit 1
-  #         }
-
-  #     - name: Install latest NVIDIA GRID Driver (Windows 11)
-  #       shell: powershell
-  #       run: |
-  #         Write-Host "=== Installing NVIDIA driver ==="
-
-  #         Write-Host "Downloading driver ..."
-  #         # $driverUrl   = "https://us.download.nvidia.com/tesla/591.59/591.59-data-center-tesla-desktop-winserver-2022-2025-dch-international.exe"
-  #         $driverUrl    = "https://storage.googleapis.com/nvidia-drivers-us-public/GRID/vGPU19.4/582.16_grid_win10_win11_server2022_server2025_dch_64bit_international.exe"
-  #         $installer   = "$env:TEMP\nvidia_driver.exe"
-  #         wget $driverUrl -o $installer
-
-  #         Write-Host "Running silent install ..."
-  #         $proc = Start-Process -FilePath $installer `
-  #                               -ArgumentList "-s", "-noreboot", "-noeula" `
-  #                               -Wait -PassThru
-  #         Write-Host "Installer exit code: $($proc.ExitCode)"
-
-  #         if ($proc.ExitCode -notin @(0, 1)) {
-  #           Write-Host "=== Windows Event Log (System, last 20 errors) ==="
-  #           Get-EventLog -LogName System -EntryType Error -Newest 20 | Format-List TimeGenerated, Source, EventID, Message
-  #           Write-Error "NVIDIA driver installation failed with exit code $($proc.ExitCode)"
-  #           exit 1
-  #         }
-
-  #         Write-Host "Waiting 15s for driver to register ..."
-  #         Start-Sleep -Seconds 15
-
-  #         $env:PATH = [System.Environment]::GetEnvironmentVariable("PATH", "Machine") + ";" + [System.Environment]::GetEnvironmentVariable("PATH", "User")
-
-  #         Write-Host "Driver installed -- verifying ..."
-  #         nvidia-smi
-  #         if ($LASTEXITCODE -ne 0) {
-  #           Write-Error "nvidia-smi still fails after installation"
-  #           exit 1
-  #         }
-  #         Write-Host "NVIDIA driver installation verified successfully."
-
   run-vulkaninfo:
     runs-on: ai-run-windows11-gpu
     steps:

.github/workflows/win11-nvidia-image-builder.yml

@@ -0,0 +1,132 @@
+on:
+  workflow_dispatch:
+
+jobs:
+  build-win11-nvidia-image:
+    runs-on:  generator-windows11
+    snapshot: win11-nvidia-grid-image
+    steps:
+      - name: Check OS version
+        shell: powershell
+        run: |
+          Write-Host "=== OS Version ==="
+          $os = Get-CimInstance Win32_OperatingSystem
+          Write-Host "Caption:      $($os.Caption)"
+          Write-Host "Version:      $($os.Version)"
+          Write-Host "Build Number: $($os.BuildNumber)"
+          Write-Host "Architecture: $($os.OSArchitecture)"
+
+      - name: Install Chocolatey
+        shell: powershell
+        run: |
+          Set-ExecutionPolicy Bypass -Scope Process -Force
+          [System.Net.ServicePointManager]::SecurityProtocol = [System.Net.ServicePointManager]::SecurityProtocol -bor 3072
+          iex ((New-Object System.Net.WebClient).DownloadString('https://community.chocolatey.org/install.ps1'))
+
+      - name: Install WinGet
+        uses: Cyberboss/install-winget@v1
+
+      - name: Install git for windows with WinGet
+        shell: powershell
+        run: winget install --id Git.Git -e --source winget --disable-interactivity --accept-package-agreements --custom '/o:PathOption=CmdTools'
+
+      - name: Enable Windows 11 Developer Mode via Registry
+        shell: powershell
+        run: |
+          $registryPath = "HKLM:\SOFTWARE\Microsoft\Windows\CurrentVersion\AppModelUnlock"
+          $propertyName = "AllowDevelopmentWithoutDevLicense"
+          $propertyValue = 1
+
+          # Create the registry key path if it doesn't exist
+          if (-not (Test-Path -Path $registryPath)) {
+              New-Item -Path $registryPath -ItemType Directory -Force | Out-Null
+          }
+
+          # Set the registry value to enable Developer Mode
+          Set-ItemProperty -Path $registryPath -Name $propertyName -Value $propertyValue -Type DWORD -Force
+
+          Write-Host "Developer Mode enabled successfully."
+
+      - name: Install latest Visual C++ Redistributable
+        shell: powershell
+        run: |
+          # Download the latest x64 redistributable
+          $vcredistUrl  = "https://aka.ms/vc14/vc_redist.x64.exe"
+          $vcInstaller = "$env:TEMP\vc_redist.x64.exe"
+          Invoke-WebRequest -Uri $vcredistUrl -OutFile $vcInstaller
+          Write-Host "Running silent install ..."
+          $proc = Start-Process -FilePath $vcInstaller `
+                                -ArgumentList "/install", "/quiet", "/norestart" `
+                                -Wait -PassThru
+          Write-Host "Installer exit code: $($proc.ExitCode)"
+
+          if ($proc.ExitCode -notin @(0, 1)) {
+            Write-Host "=== Windows Event Log (System, last 20 errors) ==="
+            Get-EventLog -LogName System -EntryType Error -Newest 20 | Format-List TimeGenerated, Source, EventID, Message
+            Write-Error "Visual C++ Redistributable installation failed with exit code $($proc.ExitCode)"
+            exit 1
+          }
+
+      - name: Install Visual Studio Build Tools
+        run: |
+          # This command ensures the "Desktop development with C++" workload is installed if needed
+          winget install Microsoft.VisualStudio.BuildTools -e --source winget --disable-interactivity --accept-package-agreements --override "--wait --quiet --add Microsoft.VisualStudio.Workload.VCTools --includeRecommended"
+        shell: powershell
+
+      - name: Install LLVM with WinGet
+        shell: powershell
+        run: winget install --id=LLVM.LLVM -e --silent --source winget --disable-interactivity --accept-source-agreements --accept-package-agreements
+
+      - name: Install latest NVIDIA GRID Driver (Windows 11)
+        shell: powershell
+        run: |
+          Write-Host "=== Installing NVIDIA driver ==="
+
+          Write-Host "Downloading driver ..."
+          # $driverUrl   = "https://us.download.nvidia.com/tesla/591.59/591.59-data-center-tesla-desktop-winserver-2022-2025-dch-international.exe"
+          $driverUrl    = "https://storage.googleapis.com/nvidia-drivers-us-public/GRID/vGPU19.4/582.16_grid_win10_win11_server2022_server2025_dch_64bit_international.exe"
+          $installer   = "$env:TEMP\nvidia_driver.exe"
+          wget $driverUrl -o $installer
+
+          Write-Host "Running silent install ..."
+          $proc = Start-Process -FilePath $installer `
+                                -ArgumentList "-s", "-noreboot", "-noeula" `
+                                -Wait -PassThru
+          Write-Host "Installer exit code: $($proc.ExitCode)"
+
+          if ($proc.ExitCode -notin @(0, 1)) {
+            Write-Host "=== Windows Event Log (System, last 20 errors) ==="
+            Get-EventLog -LogName System -EntryType Error -Newest 20 | Format-List TimeGenerated, Source, EventID, Message
+            Write-Error "NVIDIA driver installation failed with exit code $($proc.ExitCode)"
+            exit 1
+          }
+
+          Write-Host "Waiting 15s for driver to register ..."
+          Start-Sleep -Seconds 15
+
+          $env:PATH = [System.Environment]::GetEnvironmentVariable("PATH", "Machine") + ";" + [System.Environment]::GetEnvironmentVariable("PATH", "User")
+
+          Write-Host "Driver installed -- verifying ..."
+          nvidia-smi
+          if ($LASTEXITCODE -ne 0) {
+            Write-Error "nvidia-smi still fails after installation"
+            exit 1
+          }
+          Write-Host "NVIDIA driver installation verified successfully."
+
+
+  # run-vulkaninfo:
+  #   runs-on: ai-run-windows11-gpu
+  #   steps:
+      - name: GPU detection via nvidia-smi and vulkaninfo
+        continue-on-error: true
+        shell: powershell
+        run: |
+          Write-Host "=== GPU Device List ==="
+          nvidia-smi -L
+          Write-Host ""
+          Write-Host "=== Running nvidia-smi ==="
+          nvidia-smi
+          Write-Host ""
+          Write-Host "=== Running Vulkaninfo ==="
+          vulkaninfo.exe --summary