E2E (NVIDIA L40S x4) Release Branch #26
This file contains hidden or 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
| # SPDX-License-Identifier: Apache-2.0 | |
| name: E2E (NVIDIA L40S x4) Release Branch | |
| on: | |
| schedule: | |
| - cron: '0 19 * * 1,3' # Runs at 7PM UTC every Mon/Wed | |
| workflow_dispatch: {} | |
| env: | |
| TMPDIR: /home/tmp | |
| LATEST_SDG_RELEASE_BRANCH: release-v0.8 | |
| LATEST_ILAB_RELEASE_BRANCH: release-v0.26 | |
| jobs: | |
| start-large-ec2-runner: | |
| runs-on: ubuntu-latest | |
| outputs: | |
| label: ${{ steps.launch-ec2-instance-with-fallback.outputs.label }} | |
| ec2-instance-id: ${{ steps.launch-ec2-instance-with-fallback.outputs.ec2-instance-id }} | |
| ec2-instance-region: ${{ steps.launch-ec2-instance-with-fallback.outputs.ec2-instance-region }} | |
| steps: | |
| - name: Checkout "launch-ec2-runner-with-fallback" in-house CI action | |
| uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0 | |
| with: | |
| repository: instructlab/ci-actions | |
| # clone the "ci-actions" repo to a local directory called "ci-actions", instead of | |
| # overwriting the current WORKDIR contents | |
| path: ci-actions | |
| ref: release-v0.1 | |
| sparse-checkout: | | |
| actions/launch-ec2-runner-with-fallback | |
| - name: Launch EC2 Runner with Fallback | |
| id: launch-ec2-instance-with-fallback | |
| uses: ./ci-actions/actions/launch-ec2-runner-with-fallback | |
| env: | |
| TMPDIR: "/tmp" | |
| with: | |
| aws_access_key_id: ${{ secrets.AWS_ACCESS_KEY_ID }} | |
| aws_secret_access_key: ${{ secrets.AWS_SECRET_ACCESS_KEY }} | |
| github_token: ${{ secrets.GH_PERSONAL_ACCESS_TOKEN }} | |
| regions_config: > | |
| [ | |
| { | |
| "region": "us-east-2", | |
| "subnets": { | |
| "us-east-2a": "${{ vars.SUBNET_US_EAST_2A }}", | |
| "us-east-2b": "${{ vars.SUBNET_US_EAST_2B }}", | |
| "us-east-2c": "${{ vars.SUBNET_US_EAST_2C }}" | |
| }, | |
| "ec2-ami": "${{ vars.AWS_EC2_AMI_US_EAST_2 }}", | |
| "security-group-id": "${{ vars.SECURITY_GROUP_ID_US_EAST_2 }}" | |
| }, | |
| { | |
| "region": "us-east-1", | |
| "subnets": { | |
| "us-east-1a": "${{ vars.SUBNET_US_EAST_1A }}", | |
| "us-east-1b": "${{ vars.SUBNET_US_EAST_1B }}", | |
| "us-east-1c": "${{ vars.SUBNET_US_EAST_1C }}", | |
| "us-east-1d": "${{ vars.SUBNET_US_EAST_1D }}", | |
| "us-east-1e": "${{ vars.SUBNET_US_EAST_1E }}", | |
| "us-east-1f": "${{ vars.SUBNET_US_EAST_1F }}" | |
| }, | |
| "ec2-ami": "${{ vars.AWS_EC2_AMI_US_EAST_1 }}", | |
| "security-group-id": "${{ vars.SECURITY_GROUP_ID_US_EAST_1 }}" | |
| } | |
| ] | |
| try_spot_instance_first: false | |
| ec2_instance_type: g6e.12xlarge | |
| aws_resource_tags: > | |
| [ | |
| {"Key": "Name", "Value": "instructlab-ci-github-large-runner"}, | |
| {"Key": "GitHubRepository", "Value": "${{ github.repository }}"}, | |
| {"Key": "GitHubRef", "Value": "${{ github.ref }}"}, | |
| {"Key": "GitHubPR", "Value": "${{ github.event.number }}"} | |
| ] | |
| e2e-large-test: | |
| needs: | |
| - start-large-ec2-runner | |
| runs-on: ${{ needs.start-large-ec2-runner.outputs.label }} | |
| permissions: | |
| pull-requests: write | |
| steps: | |
| - name: Install Packages | |
| run: | | |
| cat /etc/os-release | |
| mkdir -p /home/tmp | |
| sudo dnf install -y gcc gcc-c++ make git-core python3.11 python3.11-devel | |
| - name: Checkout instructlab/instructlab | |
| uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0 | |
| with: | |
| repository: "instructlab/instructlab" | |
| ref: ${{ env.LATEST_ILAB_RELEASE_BRANCH }} | |
| path: "instructlab" | |
| # https://github.com/actions/checkout/issues/249 | |
| fetch-depth: 0 | |
| - name: Checkout instructlab/sdg | |
| uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0 | |
| with: | |
| path: "sdg" | |
| ref: ${{ env.LATEST_SDG_RELEASE_BRANCH }} | |
| # https://github.com/actions/checkout/issues/249 | |
| fetch-depth: 0 | |
| - name: Install ilab | |
| working-directory: ./instructlab | |
| run: | | |
| PYTHON="python3.11" ./scripts/install-ilab-with-cuda.sh | |
| - name: Update instructlab-sdg library | |
| working-directory: ./sdg | |
| run: | | |
| . ../instructlab/venv/bin/activate | |
| # Patch out our own pin from the ilab repo constraints file | |
| ilab_constraints=../instructlab/constraints-dev.txt | |
| sed -i '/instructlab-sdg==/d' $ilab_constraints | |
| # Since we reuse the virtual environment prepared using ilab | |
| # constraints, we should stick to the same constraints when | |
| # installing latest sdg. | |
| # | |
| # FIX: this is not ideal; a proper fix would require decoupling the | |
| # two repos in CI: either by removing the job completely and relying | |
| # on "sdk" (no ilab) test runs; or by preparing a separate | |
| # constraints file that would consider both the requirements files | |
| # for the sdg library AND for the ilab - so that they are | |
| # consistent. | |
| pip_install="pip install -c $ilab_constraints" | |
| $pip_install . | |
| - name: Check disk before tests | |
| run: | | |
| df -h | |
| df -i | |
| - name: Run e2e test | |
| working-directory: ./instructlab | |
| env: | |
| HF_TOKEN: ${{ secrets.HF_TOKEN }} | |
| OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }} | |
| run: | | |
| . venv/bin/activate | |
| ./scripts/e2e-ci.sh -l -p | |
| - name: Check disk after tests | |
| if: ${{ always() }} | |
| run: | | |
| df -h | |
| df -i | |
| - name: Send Discord notification for failure | |
| if: failure() | |
| uses: sarisia/actions-status-discord@11a0bfe3b50977e38aa2bd4a4ebd296415e83c19 # v1.15.4 | |
| with: | |
| webhook: ${{ secrets.SON_OF_JEEVES_DISCORD_WEBHOOK }} | |
| status: ${{ job.status }} | |
| title: "e2e-nvidia-l40s-x4" | |
| description: | | |
| Job in **${{ github.repository }}** running on branch `${{ env.LATEST_SDG_RELEASE_BRANCH }}` completed **with failures** ❌ | |
| Click [here](${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}) to view details. | |
| color: 0xCB2431 # Red color for failure | |
| - name: Send Discord notification for success | |
| if: success() | |
| uses: sarisia/actions-status-discord@11a0bfe3b50977e38aa2bd4a4ebd296415e83c19 # v1.15.4 | |
| with: | |
| webhook: ${{ secrets.SON_OF_JEEVES_DISCORD_WEBHOOK }} | |
| status: ${{ job.status }} | |
| title: "e2e-nvidia-l40s-x4" | |
| description: | | |
| Job in **${{ github.repository }}** running on branch `${{ env.LATEST_SDG_RELEASE_BRANCH }}` completed **successfully** ✅ | |
| Click [here](${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}) to view details. | |
| color: 0x28A745 # Green color for success | |
| stop-large-ec2-runner: | |
| needs: | |
| - start-large-ec2-runner | |
| - e2e-large-test | |
| runs-on: ubuntu-latest | |
| if: ${{ always() }} | |
| steps: | |
| - name: Configure AWS credentials | |
| uses: aws-actions/configure-aws-credentials@7474bc4690e29a8392af63c5b98e7449536d5c3a # v4.3.1 | |
| with: | |
| aws-access-key-id: ${{ secrets.AWS_ACCESS_KEY_ID }} | |
| aws-secret-access-key: ${{ secrets.AWS_SECRET_ACCESS_KEY }} | |
| aws-region: ${{ needs.start-large-ec2-runner.outputs.ec2-instance-region }} | |
| - name: Stop EC2 runner | |
| uses: machulav/ec2-github-runner@fb91019e71385fb10dfcbec812b4de8c61589f7b # v2.4.1 | |
| with: | |
| mode: stop | |
| github-token: ${{ secrets.GH_PERSONAL_ACCESS_TOKEN }} | |
| label: ${{ needs.start-large-ec2-runner.outputs.label }} | |
| ec2-instance-id: ${{ needs.start-large-ec2-runner.outputs.ec2-instance-id }} |