Felipe97/llama-cpp-compiled

1ARG ONEAPI_VERSION=2025.3.3-0-devel-ubuntu24.042ARG BUILD_DATE=N/A3ARG APP_VERSION=N/A4ARG APP_REVISION=N/A5 6## Build Image7 8ARG NODE_VERSION=249 10FROM docker.io/node:$NODE_VERSION AS web11 12ARG APP_VERSION13 14WORKDIR /app/tools/ui15 16COPY tools/ui/package.json tools/ui/package-lock.json ./17RUN npm ci18 19COPY tools/ui/ ./20RUN LLAMA_BUILD_NUMBER="$APP_VERSION" npm run build21 22FROM docker.io/intel/deep-learning-essentials:$ONEAPI_VERSION AS build23 24ARG GGML_SYCL_F16=ON25ARG LEVEL_ZERO_VERSION=1.28.226ARG LEVEL_ZERO_UBUNTU_VERSION=u24.0427RUN apt-get update && \28    apt-get install -y git libssl-dev wget ca-certificates && \29    cd /tmp && \30    wget -q "https://github.com/oneapi-src/level-zero/releases/download/v${LEVEL_ZERO_VERSION}/level-zero_${LEVEL_ZERO_VERSION}%2B${LEVEL_ZERO_UBUNTU_VERSION}_amd64.deb" -O level-zero.deb && \31    wget -q "https://github.com/oneapi-src/level-zero/releases/download/v${LEVEL_ZERO_VERSION}/level-zero-devel_${LEVEL_ZERO_VERSION}%2B${LEVEL_ZERO_UBUNTU_VERSION}_amd64.deb" -O level-zero-devel.deb && \32    apt-get -o Dpkg::Options::="--force-overwrite" install -y ./level-zero.deb ./level-zero-devel.deb && \33    rm -f /tmp/level-zero.deb /tmp/level-zero-devel.deb34 35WORKDIR /app36 37COPY . .38 39COPY --from=web /app/tools/ui/dist tools/ui/dist40 41RUN if [ "${GGML_SYCL_F16}" = "ON" ]; then \42        echo "GGML_SYCL_F16 is set" \43        && export OPT_SYCL_F16="-DGGML_SYCL_F16=ON" \44        && export SYCL_PROGRAM_COMPILE_OPTIONS="-cl-fp32-correctly-rounded-divide-sqrt"; \45    fi && \46    echo "Building with dynamic libs" && \47    cmake -B build -DGGML_NATIVE=OFF -DGGML_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx -DGGML_BACKEND_DL=ON -DGGML_CPU_ALL_VARIANTS=ON -DLLAMA_BUILD_TESTS=OFF ${OPT_SYCL_F16} && \48    cmake --build build --config Release -j$(nproc)49 50RUN mkdir -p /app/lib && \51    find build -name "*.so*" -exec cp -P {} /app/lib \;52 53RUN mkdir -p /app/full \54    && cp build/bin/* /app/full \55    && cp *.py /app/full \56    && cp -r conversion /app/full \57    && cp -r gguf-py /app/full \58    && cp -r requirements /app/full \59    && cp requirements.txt /app/full \60    && cp .devops/tools.sh /app/full/tools.sh61 62FROM docker.io/intel/deep-learning-essentials:$ONEAPI_VERSION AS base63 64ARG BUILD_DATE=N/A65ARG APP_VERSION=N/A66ARG APP_REVISION=N/A67ARG IMAGE_URL=https://github.com/ggml-org/llama.cpp68ARG IMAGE_SOURCE=https://github.com/ggml-org/llama.cpp69LABEL org.opencontainers.image.created=$BUILD_DATE \70      org.opencontainers.image.version=$APP_VERSION \71      org.opencontainers.image.revision=$APP_REVISION \72      org.opencontainers.image.title="llama.cpp" \73      org.opencontainers.image.description="LLM inference in C/C++" \74      org.opencontainers.image.url=$IMAGE_URL \75      org.opencontainers.image.source=$IMAGE_SOURCE76 77#Following versions are for multiple GPUs, since 26.x has known issue:78#   https://github.com/ggml-org/llama.cpp/issues/21747,79#   https://github.com/intel/compute-runtime/issues/921.80#ARG IGC_VERSION=v2.20.581#ARG IGC_VERSION_FULL=2_2.20.5+1997282#ARG COMPUTE_RUNTIME_VERSION=25.40.35563.1083#ARG COMPUTE_RUNTIME_VERSION_FULL=25.40.35563.10-084#ARG IGDGMM_VERSION=22.8.285 86 87ARG IGC_VERSION=v2.34.488ARG IGC_VERSION_FULL=2_2.34.4+2142889ARG COMPUTE_RUNTIME_VERSION=26.18.38308.190ARG COMPUTE_RUNTIME_VERSION_FULL=26.18.38308.1-091ARG IGDGMM_VERSION=22.10.092RUN mkdir /tmp/neo/ && cd /tmp/neo/ \93  && wget https://github.com/intel/intel-graphics-compiler/releases/download/$IGC_VERSION/intel-igc-core-${IGC_VERSION_FULL}_amd64.deb \94  && wget https://github.com/intel/intel-graphics-compiler/releases/download/$IGC_VERSION/intel-igc-opencl-${IGC_VERSION_FULL}_amd64.deb \95  && wget https://github.com/intel/compute-runtime/releases/download/$COMPUTE_RUNTIME_VERSION/intel-ocloc-dbgsym_${COMPUTE_RUNTIME_VERSION_FULL}_amd64.ddeb \96  && wget https://github.com/intel/compute-runtime/releases/download/$COMPUTE_RUNTIME_VERSION/intel-ocloc_${COMPUTE_RUNTIME_VERSION_FULL}_amd64.deb \97  && wget https://github.com/intel/compute-runtime/releases/download/$COMPUTE_RUNTIME_VERSION/intel-opencl-icd-dbgsym_${COMPUTE_RUNTIME_VERSION_FULL}_amd64.ddeb \98  && wget https://github.com/intel/compute-runtime/releases/download/$COMPUTE_RUNTIME_VERSION/intel-opencl-icd_${COMPUTE_RUNTIME_VERSION_FULL}_amd64.deb \99  && wget https://github.com/intel/compute-runtime/releases/download/$COMPUTE_RUNTIME_VERSION/libigdgmm12_${IGDGMM_VERSION}_amd64.deb \100  && wget https://github.com/intel/compute-runtime/releases/download/$COMPUTE_RUNTIME_VERSION/libze-intel-gpu1-dbgsym_${COMPUTE_RUNTIME_VERSION_FULL}_amd64.ddeb \101  && wget https://github.com/intel/compute-runtime/releases/download/$COMPUTE_RUNTIME_VERSION/libze-intel-gpu1_${COMPUTE_RUNTIME_VERSION_FULL}_amd64.deb \102  && dpkg --install *.deb103 104RUN apt-get update \105    && apt-get install -y libgomp1 curl ffmpeg \106    && apt autoremove -y \107    && apt clean -y \108    && rm -rf /tmp/* /var/tmp/* \109    && find /var/cache/apt/archives /var/lib/apt/lists -not -name lock -type f -delete \110    && find /var/cache -type f -delete111 112### Full113FROM base AS full114 115COPY --from=build /app/lib/ /app116COPY --from=build /app/full /app117 118WORKDIR /app119 120RUN apt-get update && \121    apt-get install -y \122        git \123        python3 \124        python3-pip \125        python3-venv && \126    python3 -m venv /opt/venv && \127    . /opt/venv/bin/activate && \128    pip install --upgrade pip setuptools wheel && \129    pip install -r requirements.txt && \130    apt autoremove -y && \131    apt clean -y && \132    rm -rf /tmp/* /var/tmp/* && \133    find /var/cache/apt/archives /var/lib/apt/lists -not -name lock -type f -delete && \134    find /var/cache -type f -delete135 136ENV PATH="/opt/venv/bin:$PATH"137 138ENTRYPOINT ["/app/tools.sh"]139 140### Light, CLI only141FROM base AS light142 143COPY --from=build /app/lib/ /app144COPY --from=build /app/full/llama /app/full/llama-cli /app/full/llama-completion /app145 146WORKDIR /app147 148ENTRYPOINT [ "/app/llama-cli" ]149 150### Server, Server only151FROM base AS server152 153ENV LLAMA_ARG_HOST=0.0.0.0154 155COPY --from=build /app/lib/ /app156COPY --from=build /app/full/llama /app/full/llama-server /app157 158WORKDIR /app159 160HEALTHCHECK CMD [ "curl", "-f", "http://localhost:8080/health" ]161 162ENTRYPOINT [ "/app/llama-server" ]163