Felipe97/llama-cpp-compiled
01.2k
1name: Server (sanitize)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-sanitize.yml',19 '**/CMakeLists.txt',20 '**/Makefile',21 '**/*.h',22 '**/*.hpp',23 '**/*.c',24 '**/*.cpp',25 'tools/server/**.*'26 ]27 28 pull_request:29 types: [opened, synchronize, reopened]30 paths: [31 '.github/workflows/server-sanitize.yml'32 ]33 34env:35 # note: this is dud token to avoid rate limiting (https://github.com/ggml-org/llama.cpp/pull/25706#issuecomment-4979941302)36 HF_TOKEN: ${{ secrets.HF_TOKEN_CI }}37 LLAMA_ARG_LOG_COLORS: 138 LLAMA_ARG_LOG_PREFIX: 139 LLAMA_ARG_LOG_TIMESTAMPS: 140 LLAMA_ARG_LOG_VERBOSITY: 1041 42concurrency:43 group: ${{ github.workflow }}-${{ github.ref }}-${{ github.head_ref || github.run_id }}44 cancel-in-progress: true45 46jobs:47 server:48 runs-on: hf-jobs-cpu-upgrade49 50 strategy:51 matrix:52 sanitizer: [ADDRESS, UNDEFINED] # THREAD is very slow53 build_type: [RelWithDebInfo]54 fail-fast: false55 56 steps:57 - name: Clone58 id: checkout59 uses: actions/checkout@v660 with:61 fetch-depth: 062 ref: ${{ github.event.inputs.sha || github.event.pull_request.head.sha || github.sha || github.head_ref || github.ref_name }}63 64 - name: Install dependencies65 run: |66 sudo apt update67 sudo apt install -y build-essential cmake python3-full68 69 - name: ccache70 uses: ggml-org/ccache-action@v1.2.2471 with:72 restore: false73 save: false74 75 - name: ccache-buckets-restore76 uses: ./.github/actions/ccache-buckets77 with:78 key: server-sanitize-${{ matrix.sanitizer }}79 folder: llama.cpp80 hf_bucket: ggml-org/cache81 82 - name: Build83 id: cmake_build84 run: |85 cmake -B build \86 -DLLAMA_BUILD_BORINGSSL=ON \87 -DGGML_SCHED_NO_REALLOC=ON \88 -DGGML_SANITIZE_ADDRESS=${{ matrix.sanitizer == 'ADDRESS' }} \89 -DGGML_SANITIZE_THREAD=${{ matrix.sanitizer == 'THREAD' }} \90 -DGGML_SANITIZE_UNDEFINED=${{ matrix.sanitizer == 'UNDEFINED' }} \91 -DLLAMA_SANITIZE_ADDRESS=${{ matrix.sanitizer == 'ADDRESS' }} \92 -DLLAMA_SANITIZE_THREAD=${{ matrix.sanitizer == 'THREAD' }} \93 -DLLAMA_SANITIZE_UNDEFINED=${{ matrix.sanitizer == 'UNDEFINED' }}94 cmake --build build --config ${{ matrix.build_type }} -j $(nproc) --target llama-server95 96 - name: ccache-buckets-save97 if: ${{ github.event_name == 'push' && github.ref == 'refs/heads/master' }}98 uses: ./.github/actions/ccache-buckets99 env:100 HF_TOKEN: ${{ secrets.HF_TOKEN_CACHE_OUTPUT }}101 with:102 key: server-sanitize-${{ matrix.sanitizer }}103 folder: llama.cpp104 evict-old-files: 1d105 hf_bucket: ggml-org/cache106 save: true107 108 - name: Install Python dependencies109 run: |110 python3 -m venv .venv111 .venv/bin/pip install -r tools/server/tests/requirements.txt112 113 - name: Tests114 id: server_integration_tests115 if: ${{ (!matrix.disabled_on_pr || !github.event.pull_request) }}116 run: |117 source .venv/bin/activate118 cd tools/server/tests119 PYTEST_WORKERS=1 ./tests.sh120 121 - name: Slow tests122 id: server_integration_tests_slow123 if: ${{ (github.event.schedule || github.event.inputs.slow_tests == 'true') && matrix.build_type == 'Release' }}124 run: |125 source .venv/bin/activate126 cd tools/server/tests127 PYTEST_WORKERS=1 SLOW_TESTS=1 ./tests.sh128 