text-generation-inference/.github/workflows/load_test.yaml

name: Nightly load test

on:
  schedule:
    - cron: '0 0 * * 1-5'
  workflow_call:
  workflow_dispatch:

  pull_request:
    paths:
      - ".github/workflows/load_test.yaml"
    branches:
      - 'main'

env:
  AWS_DEFAULT_REGION: us-east-1
  AWS_ACCESS_KEY_ID: ${{ secrets.S3_AWS_ACCESS_KEY_ID }}
  AWS_SECRET_ACCESS_KEY: ${{ secrets.S3_AWS_SECRET_ACCESS_KEY }}
  LOAD_TEST_ISSUE: 2235

jobs:
  load-tests:
    permissions:
      issues: write
    concurrency:
      group: ${{ github.workflow }}-${{ github.job }}-${{ github.head_ref || github.run_id }}
      cancel-in-progress: true
    runs-on:
      group: aws-g6-12xlarge-plus-priv
    env:
      DOCKER_VOLUME: /cache
    steps:
      - name: Checkout repository
        uses: actions/checkout@v3

      - name: Install awscli
        run: |
          sudo apt-get update
          sudo apt-get install -y awscli          

      - name: Install poetry
        run: |
          curl -sSL https://install.python-poetry.org | python3 -
          export PATH="$HOME/.local/bin:$PATH"
          poetry --version          

      - name: Install texlive minimal
        run: |
          sudo apt-get update
          sudo apt-get install -y texlive-latex-extra texlive-fonts-recommended dvipng cm-super          

      - name: Install Go 1.21
        uses: actions/setup-go@v2
        with:
          go-version: 1.21

      - name: Install Python 3.11
        uses: actions/setup-python@v2
        with:
          python-version: 3.11

      - name: Download artifacts from previous runs
        uses: actions/github-script@v6
        continue-on-error: true
        env:
          WORKFLOW_FILENAME: load_test.yaml
          ARTIFACT_NAME: benchmark_results_csv
          ARTIFACT_FILENAME: benchmark_results_csv.zip
          UNZIP_DIR: /tmp/artifacts
        with:
          script: |
            const script = require('./load_tests/download_artifact.js')
            await script({github, context, core})            

      - name: Run load test
        run: |
          export PATH="/home/runner/.local/bin:$PATH"
          cd load_tests
          make build-k6
          make load-test          
        shell: bash

      - name: Archive test results artifacts
        uses: actions/upload-artifact@v4
        with:
          name: benchmark_results
          path: |
            load_tests/output/*            

      - name: Upload to S3
        run: |
          aws s3 cp load_tests/output/ s3://text-generation-inference-ci/${{ github.sha }} --recursive          

      - uses: actions/github-script@v6
        if: github.event_name == 'pull_request'
        with:
          script: |
            let content=require('fs').readFileSync('load_tests/output/benchmark_avg_delta.md', 'utf-8');
            github.rest.issues.createComment({
              issue_number: process.env.LOAD_TEST_ISSUE,
              owner: context.repo.owner,
              repo: context.repo.repo,
              body: '🚀 Load test results are in for commit [${{ github.sha }}](https://github.com/huggingface/text-generation-inference/commit/${{ github.sha }})\n\n'+
                '## Variable length prompts\n'+
                '<p float="left">\n'+
                '<img src="http://text-generation-inference-ci.s3-website-us-east-1.amazonaws.com/${{github.sha}}/sharegpt_conversations_constant_arrival_rate.png" width=200>\n' +
                '<img src="http://text-generation-inference-ci.s3-website-us-east-1.amazonaws.com/${{github.sha}}/sharegpt_conversations_constant_vus.png" width=200>\n' +
                '</p>\n\n' +
                '## Constant length prompts\n'+
                '<p float="left">\n'+
                '<img src="http://text-generation-inference-ci.s3-website-us-east-1.amazonaws.com/${{github.sha}}/constant_tokens_constant_vus.png" width=200>\n' +
                '<img src="http://text-generation-inference-ci.s3-website-us-east-1.amazonaws.com/${{github.sha}}/constant_tokens_constant_arrival_rate.png" width=200>\n'+
                '</p>\n\n' +
                '## Delta to last release\n\n'+
                content
            })