llamacpp.yaml

name: llama.cpp

defaults:
  run:
    shell: bash -el {0}
on:
  workflow_dispatch:
  pull_request:
    paths:
      - '.github/workflows/llamacpp.yaml'
      - 'requirements/**'

  push:
    paths:
        - '.github/workflows/llamacpp.yaml'

concurrency:
  group: ${{ github.workflow }}-${{ github.ref }}
  cancel-in-progress: true

permissions:
  contents: read
  issues: write

jobs:
    # ubuntu-arm64-test:
    #   runs-on: ubuntu-latest
    #   steps:
    #   - uses: actions/checkout@v4
      
    #   - uses: uraimo/run-on-arch-action@v2
    #     name: Run Docker on ARM64
    #     with:
    #       arch: aarch64
    #       distro: ubuntu22.04
    #       base_image: --platform=linux/arm64 ascendai/cann:openeuler-python3.10-cann8.0.rc2.beta1
    #       run: |
    #         yum update -y
    #         yum install git cmake gcc gcc-c++ make -y
    #         export LD_LIBRARY_PATH=/usr/local/Ascend/ascend-toolkit/latest/lib64:/usr/local/Ascend/ascend-toolkit/latest/aarch64-linux/devlib/:${LD_LIBRARY_PATH}
            
    #         git clone https://github.com/ggerganov/llama.cpp.git
    #         cd llama.cpp
    #         cmake -B build -DCMAKE_BUILD_TYPE=release -DGGML_CANN=on -DSOC_TYPE=ascend910b3 -DGGML_NATIVE=OFF -DGGML_CPU_ARM_ARCH=armv8-a
    #         cmake --build build

    unit-tests:
      runs-on: ubuntu-24.04-arm
      strategy:
        matrix:
          cann:
            - 'openeuler-python3.10-cann8.0.rc3.beta1'
          build:
            - 'Release'
          device:
            - 'ascend910b3'
      container:
        image: ascendai/cann:${{ matrix.cann }}
      steps:
      - name: Dependencies
        run: |
            yum update -y
            yum install git cmake gcc gcc-c++ make -y  

      - name: Checkout llama.cpp
        uses: actions/checkout@v4
        with:
          repository: ggerganov/llama.cpp
          path: llama.cpp
            
      - name: Build
        run: |
            cd llama.cpp
            mkdir build
            cd build
            export LD_LIBRARY_PATH=${ASCEND_TOOLKIT_HOME}/lib64:${ASCEND_TOOLKIT_HOME}/$(uname -m)-linux/devlib/:${LD_LIBRARY_PATH}
            cmake .. -DCMAKE_BUILD_TYPE=${{ matrix.build }} -DGGML_CANN=on -DSOC_TYPE=${{ matrix.device }} && cmake --build . -j $(nproc)

    ubuntu-unit-tests:
      runs-on: ubuntu-latest
  
      container:
        image: ascendai/cann
  
      steps:
      - uses: actions/checkout@v4
          
      - name: Install llamacpp
        uses: nick-fields/retry@v3
        with:
          timeout_minutes: 30
          max_attempts: 3
          retry_on: error
          command: |
            uname -a
            apt-get update -y
            apt-get install git cmake build-essential make -y   
            git clone https://github.com/ggerganov/llama.cpp.git
      - name: Build
        run: |
            cd llama.cpp
            export LD_LIBRARY_PATH=${ASCEND_TOOLKIT_HOME}/$(uname -m)-linux/devlib/:${LD_LIBRARY_PATH}
            cmake -B build -DCMAKE_BUILD_TYPE=release -DGGML_CANN=on -DSOC_TYPE=ascend910b3 && cmake --build build -j $(nproc)