refactor: move GPU drivers to config and refactor pipelines for CUDA …

…driver (#46) * Move driver config to justfile * Add set config step in main.yaml, remove cuda var in justfile * add load config step to ci.yaml * change relative path for GH actions * update main builds to use var from load config * testing * echo in load config * gs * change matrix.driver_version to steps.load_config.outputs.cuda_version * update to use steps.load_config.outputs.cuda_versionin main.yaml * GH action directory debug for main.yaml * move the fetch up * remove debug statements
Azure · Aug 7, 2024 · 9094a1a · 9094a1a
1 parent b40b851
commit 9094a1a
Show file tree

Hide file tree

Showing 4 changed files with 30 additions and 14 deletions.
diff --git a/.github/workflows/ci.yaml b/.github/workflows/ci.yaml
@@ -5,17 +5,22 @@ on:
       - main
   workflow_dispatch: {}
 
-jobs: 
+jobs:
   cuda:
     runs-on: ubuntu-latest
     strategy:
       matrix:
-        driver_version: ["550.90.07"]
         driver_kind: ["cuda"]
     steps:
       - uses: actions/checkout@v2
         with:
           fetch-depth: 0
+      - name: Load CUDA config
+        id: load_config
+        run: |
+          cuda_version=$(yq e '.cuda.version' driver_config.yml)
+          echo "CUDA_VERSION=$cuda_version"
+          echo "cuda_version=$cuda_version" >> $GITHUB_OUTPUT
       - name: Set up Docker Buildx
         uses: docker/setup-buildx-action@v1
       - name: Cache Docker layers
@@ -28,7 +33,7 @@ jobs:
       - uses: paulhatch/[email protected]
         with:
           bump_each_commit: false
-          version_format: "${{ matrix.driver_kind}}-${{ matrix.driver_version }}-sha-${GITHUB_SHA:0:6}"
+          version_format: "cuda-${{ steps.load_config.outputs.cuda_version }}-sha-${GITHUB_SHA:0:6}"
         id: semver
       - name: 'Check version'
         run: |
@@ -39,7 +44,7 @@ jobs:
           set -x
           echo "tag is: "
           echo ${{ steps.semver.outputs.version }}
-          docker buildx build --build-arg DRIVER_KIND=${{ matrix.driver_kind }} --build-arg DRIVER_VERSION=${{ matrix.driver_version }} --cache-from=type=local,src=/tmp/.buildx-cache --cache-to=type=local,dest=/tmp/.buildx-cache-new --output=type=docker -t ${{ secrets.AZURE_REGISTRY_SERVER }}/public/aks/aks-gpu:${{ steps.semver.outputs.version }} .
+          docker buildx build --build-arg DRIVER_KIND=${{ matrix.driver_kind }} --build-arg DRIVER_VERSION=${{ steps.load_config.outputs.cuda_version }} --cache-from=type=local,src=/tmp/.buildx-cache --cache-to=type=local,dest=/tmp/.buildx-cache-new --output=type=docker -t ${{ secrets.AZURE_REGISTRY_SERVER }}/public/aks/aks-gpu:${{ steps.semver.outputs.version }} .
           docker images
       - name: Move cache
         run: |
@@ -62,9 +67,9 @@ jobs:
           uses: actions/cache@v2
           with:
             path: /tmp/.buildx-cache
-            key: ${{ runner.os }}-buildx-${{ matrix.driver_kind}}-${{ matrix.driver_version }}-${{ github.sha }}
+            key: ${{ runner.os }}-buildx-${{ matrix.driver_kind}}-${{ steps.load_config.outputs.cuda_version }}-${{ github.sha }}
             restore-keys: |
-              ${{ runner.os }}-buildx-${{ matrix.driver_kind}}-${{ matrix.driver_version }}
+              ${{ runner.os }}-buildx-${{ matrix.driver_kind}}-${{ steps.load_config.outputs.cuda_version }}
         - uses: paulhatch/[email protected]
           with:
             bump_each_commit: false

diff --git a/.github/workflows/main.yaml b/.github/workflows/main.yaml
@@ -8,17 +8,23 @@ permissions:
   id-token: write
   contents: read
 
-jobs: 
+jobs:
   cuda:
     runs-on: ubuntu-latest
     strategy:
       matrix:
-        driver_version: ["550.90.07"]
+        driver_version: ["${{ needs.load_config.outputs.cuda_version }}"]
         driver_kind: ["cuda"]
     steps:
       - uses: actions/checkout@v2
         with:
           fetch-depth: 0
+      - name: Load CUDA config
+        id: load_config
+        run: |
+          cuda_version=$(yq e '.cuda.version' driver_config.yml)
+          echo "CUDA_VERSION=$cuda_version"
+          echo "cuda_version=$cuda_version" >> $GITHUB_OUTPUT      
       - name: Set up Docker Buildx
         uses: docker/setup-buildx-action@v1
       - name: Cache Docker layers
@@ -31,7 +37,7 @@ jobs:
       - uses: paulhatch/[email protected]
         with:
           bump_each_commit: false
-          version_format: "${{ matrix.driver_kind}}-${{ matrix.driver_version }}-sha-${GITHUB_SHA:0:6}"
+          version_format: "cuda-${{ steps.load_config.outputs.cuda_version }}-sha-${GITHUB_SHA:0:6}"
         id: semver
       - name: 'Check version'
         run: |
@@ -48,7 +54,7 @@ jobs:
           set -x
           echo "tag is: "
           echo ${{ steps.semver.outputs.version }}
-          docker buildx build --build-arg DRIVER_KIND=${{ matrix.driver_kind }} --build-arg DRIVER_VERSION=${{ matrix.driver_version }} --cache-from=type=local,src=/tmp/.buildx-cache --cache-to=type=local,dest=/tmp/.buildx-cache-new --output=type=docker -t ${{ secrets.AZURE_REGISTRY_SERVER }}/public/aks/aks-gpu:${{ steps.semver.outputs.version }} .
+          docker buildx build --build-arg DRIVER_KIND=${{ matrix.driver_kind }} --build-arg DRIVER_VERSION=${{ steps.load_config.outputs.cuda_version }} --cache-from=type=local,src=/tmp/.buildx-cache --cache-to=type=local,dest=/tmp/.buildx-cache-new --output=type=docker -t ${{ secrets.AZURE_REGISTRY_SERVER }}/public/aks/aks-gpu:${{ steps.semver.outputs.version }} .
           docker images
           az acr login -n ${{ secrets.AZURE_REGISTRY_SERVER }}
           docker push ${{ secrets.AZURE_REGISTRY_SERVER }}/public/aks/aks-gpu:${{ steps.semver.outputs.version }}

diff --git a/driver_config.yml b/driver_config.yml
@@ -0,0 +1,6 @@
+cuda:
+  version: "550.90.07"
+
+grid:
+  version: "535.161.08"
+  url: "https://download.microsoft.com/download/8/d/a/8da4fb8e-3a9b-4e6a-bc9a-72ff64d7a13c/NVIDIA-Linux-x86_64-535.161.08-grid-azure.run"
diff --git a/justfile b/justfile
@@ -2,17 +2,16 @@ grid_535_url    := "https://download.microsoft.com/download/8/d/a/8da4fb8e-3a9b-
 
 grid_535_driver := "535.161.08"
 
-cuda_550_driver := "550.90.07"
 registry := "docker.io/alexeldeib"
 
 default:
 
-pushallcuda: (pushcuda cuda_550_driver)
+pushallcuda: (pushcuda)
 
 pushallgrid: (pushgrid grid_535_driver)
 
-pushcuda VERSION: (buildcuda VERSION)
-	docker push {{ registry }}/aks-gpu:{{VERSION}}-cuda
+pushcuda: (buildcuda)
+	docker push {{ registry }}/aks-gpu:$(yq e '.cuda.version' driver_config.yml)-cuda
 
 pushgrid VERSION URL: (buildgrid VERSION URL)
 	docker push {{ registry }}/aks-gpu:{{VERSION}}-grid