echodict/llama.cpp

1name: Server (self-hosted)2 3on:4  workflow_dispatch: # allows manual triggering5    inputs:6      sha:7        description: 'Commit SHA1 to build'8        required: false9        type: string10      slow_tests:11        description: 'Run slow tests'12        required: true13        type: boolean14  push:15    branches:16      - master17    paths: [18      '.github/workflows/server-self-hosted.yml',19      '**/CMakeLists.txt',20      '**/Makefile',21      '**/*.h',22      '**/*.hpp',23      '**/*.c',24      '**/*.cpp',25      '**/*.cu',26      '**/*.swift',27      '**/*.m',28      'tools/server/**.*'29    ]30 31env:32  LLAMA_LOG_COLORS: 133  LLAMA_LOG_PREFIX: 134  LLAMA_LOG_TIMESTAMPS: 135  LLAMA_LOG_VERBOSITY: 1036 37concurrency:38  group: ${{ github.workflow }}-${{ github.ref }}-${{ github.head_ref || github.run_id }}39  cancel-in-progress: true40 41jobs:42  server-metal:43    runs-on: [self-hosted, llama-server, macOS, ARM64]44 45    name: server-metal (${{ matrix.wf_name }})46    strategy:47      matrix:48        build_type: [Release]49        wf_name: ["GPUx1"]50        include:51          - build_type: Release52            extra_args: "LLAMA_ARG_BACKEND_SAMPLING=1"53            wf_name:    "GPUx1, backend-sampling"54          - build_type: Release55            extra_args: "GGML_METAL_DEVICES=2"56            wf_name:    "GPUx2"57          - build_type: Release58            extra_args: "GGML_METAL_DEVICES=2 LLAMA_ARG_BACKEND_SAMPLING=1"59            wf_name:    "GPUx2, backend-sampling"60      fail-fast: false61 62    steps:63      - name: Clone64        id: checkout65        uses: actions/checkout@v666        with:67          fetch-depth: 068          ref: ${{ github.event.inputs.sha || github.event.pull_request.head.sha || github.sha || github.head_ref || github.ref_name }}69 70      - name: Build71        id: cmake_build72        run: |73          cmake -B build -DGGML_SCHED_NO_REALLOC=ON74          cmake --build build --config ${{ matrix.build_type }} -j $(sysctl -n hw.logicalcpu) --target llama-server75 76      - name: Tests77        id: server_integration_tests78        if: ${{ (!matrix.disabled_on_pr || !github.event.pull_request) }}79        run: |80          cd tools/server/tests81          python3 -m venv venv82          source venv/bin/activate83          pip install -r requirements.txt84          export ${{ matrix.extra_args }}85          pytest -v -x -m "not slow"86 87  # TODO: provision CUDA runner88  #  server-cuda:89  #    runs-on: [self-hosted, llama-server, Linux, NVIDIA]90  #91  #    name: server-cuda (${{ matrix.wf_name }})92  #    strategy:93  #      matrix:94  #        build_type: [Release]95  #        wf_name: ["GPUx1"]96  #        include:97  #          - build_type: Release98  #            extra_args: "LLAMA_ARG_BACKEND_SAMPLING=1"99  #            wf_name:    "GPUx1, backend-sampling"100  #      fail-fast: false101  #102  #    steps:103  #      - name: Clone104  #        id: checkout105  #        uses: actions/checkout@v6106  #        with:107  #          fetch-depth: 0108  #          ref: ${{ github.event.inputs.sha || github.event.pull_request.head.sha || github.sha || github.head_ref || github.ref_name }}109  #110  #      - name: Build111  #        id: cmake_build112  #        run: |113  #          cmake -B build -DGGML_SCHED_NO_REALLOC=ON114  #          cmake --build build --config ${{ matrix.build_type }} -j $(sysctl -n hw.logicalcpu) --target llama-server115  #116  #      - name: Tests117  #        id: server_integration_tests118  #        if: ${{ (!matrix.disabled_on_pr || !github.event.pull_request) }}119  #        run: |120  #          cd tools/server/tests121  #          python3 -m venv venv122  #          source venv/bin/activate123  #          pip install -r requirements.txt124  #          export ${{ matrix.extra_args }}125  #          pytest -v -x -m "not slow"126