sgl-project--sglang/.github/workflows/stress-test.yml

name: Stress Test

on:
  workflow_dispatch:
    inputs:
      num_prompts:
        description: 'Number of prompts per model'
        required: true
        default: '50000'
        type: string
      duration_minutes:
        description: 'Timeout per model in minutes'
        required: true
        default: '45'
        type: string

jobs:
  stress-test:
    if: github.repository == 'sgl-project/sglang'
    runs-on: 8-gpu-h200
    steps:
      - name: Checkout code
        uses: actions/checkout@v4

      - name: Install dependencies
        run: |
          bash scripts/ci/cuda/ci_install_dependency.sh

      - name: Run stress tests
        timeout-minutes: 210
        env:
          NUM_PROMPTS: ${{ inputs.num_prompts }}
          DURATION_MINUTES: ${{ inputs.duration_minutes }}
        run: |
          cd test
          python3 run_suite.py --hw cuda --suite stress

      - name: Upload results
        if: always()
        uses: actions/upload-artifact@v4
        with:
          name: stress-test-results
          path: |
            stress_test_*.jsonl