e2e: replace old small job with new medium job (backport #277)

.github/mergify.yml

@@ -28,18 +28,18 @@ pull_request_rules:
     # e2e workflow
     - or:
       - and:
-        # note this should match the triggering criteria in 'e2e-nvidia-t4-x1.yml'
-        - check-success=e2e-workflow-complete
+        # note this should match the triggering criteria in 'e2e-nvidia-a10g-x1.yml'
+        - check-success=e2e-medium-workflow-complete
         - or:
           - files~=\.py$
           - files=pyproject.toml
           - files~=^requirements.*\.txt$
-          - files=.github/workflows/e2e-nvidia-t4-x1.yml
+          - files=.github/workflows/e2e-nvidia-a10g-x1.yml
       - and:
         - -files~=\.py$
         - -files=pyproject.toml
         - -files~=^requirements.*\.txt$
-        - -files=.github/workflows/e2e-nvidia-t4-x1.yml
+        - -files=.github/workflows/e2e-nvidia-a10g-x1.yml
 
     # code lint workflow
     - or:

.github/workflows/e2e-nvidia-t4-x1.yml → .github/workflows/e2e-nvidia-a10g-x1.yml

@@ -1,17 +1,15 @@
 # SPDX-License-Identifier: Apache-2.0
 
-name: E2E (NVIDIA Tesla T4 x1)
+name: E2E (NVIDIA A10G x1)
 
 on:
+  # run against every merge commit to 'main' and release branches
   push:
     branches:
       - main
       - release-*
+  # only run on PRs that touch certain regex paths
   pull_request_target:
-    types:
-      - opened
-      - synchronize
-      - reopened
     branches:
       - main
       - release-*
@@ -20,15 +18,24 @@ on:
       - '**.py'
       - 'pyproject.toml'
       - 'requirements**.txt'
-      - '.github/workflows/e2e-nvidia-t4-x1.yml' # Follow-on workflow
+      - '.github/workflows/e2e-nvidia-a10g-x1.yml' # This workflow
 
 concurrency:
   group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
   cancel-in-progress: true
 
+env:
+  LC_ALL: en_US.UTF-8
+
+defaults:
+  run:
+    shell: bash
+
+permissions:
+  contents: read
+
 jobs:
-  start-runner:
-    name: Start external EC2 runner
+  start-medium-ec2-runner:
     runs-on: ubuntu-latest
     outputs:
       label: ${{ steps.start-ec2-runner.outputs.label }}
@@ -40,6 +47,7 @@ jobs:
           aws-access-key-id: ${{ secrets.AWS_ACCESS_KEY_ID }}
           aws-secret-access-key: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
           aws-region: ${{ secrets.AWS_REGION }}
+
       - name: Start EC2 runner
         id: start-ec2-runner
         uses: machulav/ec2-github-runner@fcfb31a5760dad1314a64a0e172b78ec6fc8a17e # v2.3.6
@@ -59,23 +67,21 @@ jobs:
               {"Key": "GitHubPR", "Value": "${{ github.event.number }}"}
             ]
 
-  e2e:
-    name: E2E Test
-    needs: start-runner
-    runs-on: ${{ needs.start-runner.outputs.label }}
+  e2e-medium-test:
+    needs:
+      - start-medium-ec2-runner
+    runs-on: ${{ needs.start-medium-ec2-runner.outputs.label }}
 
     # It is important that this job has no write permissions and has
     # no access to any secrets. This part (e2e) is where we are running
     # untrusted code from PRs.
     permissions: {}
 
     steps:
-      # for debugging
-      - name: Print environment state
+      - name: Install Packages
         run: |
-          echo "Current Working Directory: $PWD"
-          echo "Files in Local Directory:"
-          ls -l
+          cat /etc/os-release
+          sudo dnf install -y gcc gcc-c++ make git python3.11 python3.11-devel
 
       - name: Checkout instructlab/instructlab
         uses: actions/checkout@d632683dd7b4114ad314bca15554477dd762a938 # v4.2.0
@@ -93,30 +99,19 @@ jobs:
           # https://github.com/actions/checkout/issues/249
           fetch-depth: 0
 
-      # for debugging
-      - name: Print environment state
-        run: |
-          echo "Current Working Directory: $PWD"
-          echo "Files in Local Directory:"
-          ls -l
-
       - name: Fetch and checkout PR
-        id: fetch_pr
-        if: github.event_name == 'pull_request_target'
+        if: ${{ github.event_name == 'pull_request_target' }}
         working-directory: ./training
         run: |
           git fetch origin pull/${{ github.event.pull_request.number }}/head:pr-${{ github.event.pull_request.number }}
           git checkout pr-${{ github.event.pull_request.number }}
 
-      - name: Install system packages
-        run: |
-          cat /etc/os-release
-          sudo dnf install -y gcc gcc-c++ make git python3.11 python3.11-devel
-
-      - name: Install instructlab
+      - name: Install ilab
         working-directory: ./instructlab
         run: |
-          export PATH="/home/ec2-user/.local/bin:/usr/local/cuda/bin:$PATH"
+          export CUDA_HOME="/usr/local/cuda"
+          export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/usr/local/cuda/lib64:/usr/local/cuda/extras/CUPTI/lib64"
+          export PATH="$PATH:$CUDA_HOME/bin"
           python3.11 -m venv --upgrade-deps venv
           . venv/bin/activate
           nvidia-smi
@@ -127,7 +122,7 @@ jobs:
           # https://github.com/instructlab/instructlab/issues/1821
           # install with Torch and build dependencies installed
           python3.11 -m pip install packaging wheel setuptools-scm
-          python3.11 -m pip install .[cuda]
+          python3.11 -m pip install .[cuda] -r requirements-vllm-cuda.txt
 
       - name: Update instructlab-training library
         working-directory: ./training
@@ -136,17 +131,21 @@ jobs:
           pip install .
           pip install .[cuda]
 
+      - name: Check disk
+        run: |
+          df -h
+
       - name: Run e2e test
         working-directory: ./instructlab
         run: |
           . venv/bin/activate
-          ./scripts/basic-workflow-tests.sh -a
+          ./scripts/e2e-ci.sh -m
 
-  stop-runner:
+  stop-medium-ec2-runner:
     name: Stop external EC2 runner
     needs:
-      - start-runner
-      - e2e
+      - start-medium-ec2-runner
+      - e2e-medium-test
     runs-on: ubuntu-latest
     if: ${{ always() }}
     steps:
@@ -161,13 +160,13 @@ jobs:
         with:
           mode: stop
           github-token: ${{ secrets.GH_PERSONAL_ACCESS_TOKEN }}
-          label: ${{ needs.start-runner.outputs.label }}
-          ec2-instance-id: ${{ needs.start-runner.outputs.ec2-instance-id }}
+          label: ${{ needs.start-medium-ec2-runner.outputs.label }}
+          ec2-instance-id: ${{ needs.start-medium-ec2-runner.outputs.ec2-instance-id }}
 
-  e2e-workflow-complete:
+  e2e-medium-workflow-complete:
     # we don't want to block PRs on failed EC2 cleanup
     # so not requiring "stop-runner" as well
-    needs: ["start-runner", "e2e"]
+    needs: ["start-medium-ec2-runner", "e2e-medium-test"]
     runs-on: ubuntu-latest
     steps:
       - name: E2E Workflow Complete

README.md

@@ -1,6 +1,7 @@
 # InstructLab Training Library
 
 ![Lint](https://github.com/instructlab/training/actions/workflows/lint.yml/badge.svg?branch=main)
+![`e2e-nvidia-a10g-x1.yaml` on `main`](https://github.com/instructlab/training/actions/workflows/e2e-nvidia-a10g-x1.yml/badge.svg?branch=main)
 ![Build](https://github.com/instructlab/training/actions/workflows/pypi.yaml/badge.svg?branch=main)
 ![Release](https://img.shields.io/github/v/release/instructlab/training)
 ![License](https://img.shields.io/github/license/instructlab/training)