LLamaSharp/.github/workflows/benchmark.yml

name: Benchmark Test
on:
  push:
    branches: [master]
  pull_request:
    branches: [master]
concurrency:
  group: ${{ github.workflow }}-${{ github.ref }}-benchmark
  cancel-in-progress: true

jobs:
  linux-benchmark-cuda:
    if: contains(github.event.pull_request.labels.*.name, 'benchmark')
    runs-on: [self-hosted, linux, gpu]

    strategy:
      fail-fast: false
      matrix:
        build: [cuda11]
        include:
          - build: cuda11
            image: nvidia/cuda:11.7.1-devel-ubuntu22.04
          # - build: cuda12
          #   image: nvidia/cuda:12.1.1-runtime-ubuntu22.04

    container:
      image: ${{ matrix.image }}
      env:
        NODE_ENV: development
      ports:
        - 80
      volumes:
        - /llamasharp_ci:/llamasharp_ci
      options: --gpus=all --ipc=host --runtime=nvidia

    steps:
    - name: DEBUG
      run: | 
        whoami
        nvidia-smi
        nvcc -V
    
    - name: Install libraries
      run: |
        apt update
        apt install -y curl libicu-dev

    - uses: actions/checkout@v4
    - uses: actions/setup-dotnet@v4
      with:
        dotnet-version: |
          8.0.x

    - name: Cache Packages
      uses: actions/cache@v4
      with:
        key: "benchmark_models"
        path: LLama.Benchmark/Models

    - name: Clear package cache
      run: dotnet clean LLamaSharp.sln && dotnet nuget locals all --clear
    - name: Restore packages
      run: dotnet restore LLamaSharp.sln
    - name: Build
      run: | 
        dotnet build LLama/LLamaSharp.csproj -c release --no-restore
        dotnet build LLama.Benchmark/LLama.Benchmark.csproj -c release --no-restore
    - name: Run benchmark test
      run: dotnet run LLama.Benchmark -c release
    - name: Upload artifacts
      if: always()
      uses: actions/upload-artifact@v3
      with:
        path: logs/ # TODO: change it
        name: logs