E2E Preset tuning Test #9
Workflow file for this run
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| name: E2E Preset tuning Test | |
| on: | |
| workflow_run: | |
| workflows: ["Build and Push Preset Models"] | |
| types: | |
| - completed | |
| workflow_dispatch: {} | |
| env: | |
| GO_VERSION: "1.22" | |
| permissions: | |
| id-token: write | |
| contents: read | |
| jobs: | |
| e2e-preset-tuning-tests: | |
| if: github.event_name == 'workflow_dispatch' || github.event.workflow_run.conclusion == 'success' | |
| runs-on: ubuntu-latest | |
| environment: preset-env | |
| steps: | |
| - name: Checkout | |
| uses: actions/[email protected] | |
| with: | |
| submodules: true | |
| fetch-depth: 0 | |
| - name: 'Az CLI login' | |
| uses: azure/[email protected] | |
| with: | |
| client-id: ${{ secrets.AZURE_CLIENT_ID }} | |
| tenant-id: ${{ secrets.AZURE_TENANT_ID }} | |
| allow-no-subscriptions: true | |
| - name: 'Set ACR Subscription' | |
| run: az account set --subscription ${{secrets.AZURE_SUBSCRIPTION_ID}} | |
| - name: Set up kubectl context | |
| run: | | |
| az aks get-credentials --resource-group llm-test --name GitRunner | |
| - name: Get test meta | |
| id: get_test_meta | |
| run: | | |
| jq --version | |
| sudo apt install jq -y | |
| CONFIG=$(jq '.matrix.image[] | select(.name == "tuning-example")' .github/e2e-preset-configs.json) | |
| echo "TAG=0.0.7" >> $GITHUB_OUTPUT | |
| echo "model=$CONFIG" >> $GITHUB_OUTPUT | |
| - name: Create Nodepool | |
| run: | | |
| NODEPOOL_EXIST=$(az aks nodepool show \ | |
| --name ${{ steps.get_test_meta.outputs.model.name }} \ | |
| --cluster-name GitRunner \ | |
| --resource-group llm-test \ | |
| --query 'name' -o tsv || echo "") | |
| echo "NODEPOOL_EXIST: $NODEPOOL_EXIST" | |
| if [ -z "$NODEPOOL_EXIST" ]; then | |
| az aks nodepool add \ | |
| --name ${{ steps.get_test_meta.outputs.model.name }} \ | |
| --cluster-name GitRunner \ | |
| --resource-group llm-test \ | |
| --node-count ${{ steps.get_test_meta.outputs.model.node-count }} \ | |
| --node-vm-size ${{ steps.get_test_meta.outputs.model.node-vm-size }} \ | |
| --node-osdisk-size ${{ steps.get_test_meta.outputs.model.node-osdisk-size }} \ | |
| --labels pool=${{ steps.get_test_meta.outputs.model.name }} \ | |
| --node-taints sku=gpu:NoSchedule \ | |
| --aks-custom-headers UseGPUDedicatedVHD=true | |
| else | |
| NODEPOOL_STATE=$(az aks nodepool show \ | |
| --name ${{ steps.get_test_meta.outputs.model.name }} \ | |
| --cluster-name GitRunner \ | |
| --resource-group llm-test \ | |
| --query 'provisioningState' -o tsv) | |
| echo "NODEPOOL_STATE: $NODEPOOL_STATE" | |
| if [ "$NODEPOOL_STATE" != "Succeeded" ]; then | |
| echo "Nodepool exists but is not in a Succeeded state. Please check manually." | |
| exit 1 | |
| else | |
| echo "Nodepool already exists and is in a running state." | |
| fi | |
| fi | |
| - name: Replace repo and Deploy Resource to K8s | |
| run: | | |
| sed -i "s/REPO_HERE/${{ secrets.ACR_AMRT_USERNAME }}/g" presets/workspace/test/tuning/tuning-job.yaml | |
| sed -i "s/TAG_HERE/${{ steps.get_test_meta.outputs.TAG }}/g" presets/workspace/test/tuning/tuning-job.yaml | |
| kubectl apply -f presets/workspace/test/tuning/tuning-job.yaml | |
| - name: Wait for tuning job to be ready | |
| shell: bash {0} | |
| run: | | |
| retval_complete=1 | |
| retval_failed=1 | |
| count=0 | |
| max_retries=60 | |
| while [[ $retval_complete -ne 0 ]] && [[ $retval_failed -ne 0 ]] && [[ $count -lt $max_retries ]]; do | |
| sleep 10 | |
| output=$(kubectl wait --for=condition=failed job/tuning-example --timeout=0 2>&1) | |
| retval_failed=$? | |
| output=$(kubectl wait --for=condition=complete job/tuning-example --timeout=0 2>&1) | |
| retval_complete=$? | |
| count=$((count + 1)) | |
| done | |
| if [ $retval_failed -eq 0 ]; then | |
| echo "Job failed. Please check logs." | |
| exit 1 | |
| elif [ $retval_complete -ne 0 ]; then | |
| echo "Job timeout." | |
| exit 1 | |
| else | |
| echo "Job succeeded." | |
| fi | |
| - name: Cleanup | |
| if: always() | |
| run: | | |
| kubectl delete --wait=true -f presets/workspace/test/tuning/tuning-job.yaml | |
| # Check and Delete AKS Nodepool if it exists | |
| NODEPOOL_EXIST=$(az aks nodepool show \ | |
| --name ${{ steps.get_test_meta.outputs.model.name }} \ | |
| --cluster-name GitRunner \ | |
| --resource-group llm-test \ | |
| --query 'name' -o tsv || echo "") | |
| if [ -n "$NODEPOOL_EXIST" ]; then | |
| echo "deleting nodepool" | |
| az aks nodepool delete \ | |
| --name ${{ steps.get_test_meta.outputs.model.name }} \ | |
| --cluster-name GitRunner \ | |
| --resource-group llm-test | |
| fi |