vllm-project/vllm
1
0
Fork 0
mirror of https://github.com/vllm-project/vllm.git synced 2025-12-06 06:53:12 +08:00
Code Issues Packages Projects Releases Wiki Activity
Files
main
vllm/docker/Dockerfile.s390x
R3hankhan e04492449e [Hardware][IBM Z] Optimize s390x Dockerfile (#28023) 
Signed-off-by: Rehan Khan <Rehan.Khan7@ibm.com>
2025-11-05 11:25:44 -08:00

267 lines
11 KiB
Docker
Raw Permalink Blame History

# Base UBI image for s390x architecture
ARG BASE_UBI_IMAGE_TAG=9.5-1736404155
ARG PYTHON_VERSION=3.12
FROM registry.access.redhat.com/ubi9/ubi-minimal:${BASE_UBI_IMAGE_TAG} AS base
# Install basic dependencies
ARG PYTHON_VERSION
ENV PYTHON_VERSION=${PYTHON_VERSION}
WORKDIR /workspace
ENV LANG=C.UTF-8 \
LC_ALL=C.UTF-8
# Install development utilities
RUN microdnf install -y \
which procps findutils tar vim git gcc-toolset-14 gcc-toolset-14-libatomic-devel patch zlib-devel \
libjpeg-turbo-devel libtiff-devel libpng-devel libwebp-devel freetype-devel harfbuzz-devel \
openssl-devel openblas openblas-devel autoconf automake libtool cmake numpy libsndfile \
clang llvm-devel llvm-static clang-devel && \
microdnf clean all
# Python Installation
FROM base AS python-install
ARG PYTHON_VERSION
ENV VIRTUAL_ENV=/opt/vllm
ENV PATH="$VIRTUAL_ENV/bin:$PATH"
ENV PYTHON_VERSION=${PYTHON_VERSION}
RUN microdnf install -y \
python${PYTHON_VERSION}-devel python${PYTHON_VERSION}-pip python${PYTHON_VERSION}-wheel && \
python${PYTHON_VERSION} -m venv $VIRTUAL_ENV && pip install --no-cache -U pip wheel uv && microdnf clean all
FROM python-install AS pyarrow
# Build Apache Arrow
WORKDIR /tmp
RUN --mount=type=cache,target=/root/.cache/uv \
git clone https://github.com/apache/arrow.git && \
cd arrow/cpp && \
mkdir release && cd release && \
cmake -DCMAKE_BUILD_TYPE=Release \
-DCMAKE_INSTALL_PREFIX=/usr/local \
-DARROW_PYTHON=ON \
-DARROW_PARQUET=ON \
-DARROW_ORC=ON \
-DARROW_FILESYSTEM=ON \
-DARROW_WITH_LZ4=ON \
-DARROW_WITH_ZSTD=ON \
-DARROW_WITH_SNAPPY=ON \
-DARROW_JSON=ON \
-DARROW_CSV=ON \
-DARROW_DATASET=ON \
-DPROTOBUF_PROTOC_EXECUTABLE=/usr/bin/protoc \
-DARROW_DEPENDENCY_SOURCE=BUNDLED \
.. && \
make -j$(nproc) && \
make install && \
cd ../../python && \
export PYARROW_PARALLEL=4 && \
export ARROW_BUILD_TYPE=release && \
uv pip install -r requirements-build.txt && \
python setup.py build_ext --build-type=$ARROW_BUILD_TYPE --bundle-arrow-cpp bdist_wheel
FROM python-install AS numa-build
# Install numactl (needed for numa.h dependency)
WORKDIR /tmp
RUN curl -LO https://github.com/numactl/numactl/archive/refs/tags/v2.0.16.tar.gz && \
tar -xvzf v2.0.16.tar.gz && \
cd numactl-2.0.16 && \
./autogen.sh && \
./configure && \
make
# Set include path
ENV C_INCLUDE_PATH="/usr/local/include:$C_INCLUDE_PATH"
FROM python-install AS rust
ENV CARGO_HOME=/root/.cargo
ENV RUSTUP_HOME=/root/.rustup
ENV PATH="$CARGO_HOME/bin:$RUSTUP_HOME/bin:$PATH"
RUN curl https://sh.rustup.rs -sSf | sh -s -- -y && \
. "$CARGO_HOME/env" && \
rustup default stable && \
rustup show
FROM python-install AS torch-vision
# Install torchvision
ARG TORCH_VISION_VERSION=v0.23.0
WORKDIR /tmp
RUN --mount=type=cache,target=/root/.cache/uv \
git clone https://github.com/pytorch/vision.git && \
cd vision && \
git checkout $TORCH_VISION_VERSION && \
uv pip install torch==2.8.0 --index-url https://download.pytorch.org/whl/cpu && \
python setup.py bdist_wheel
FROM python-install AS hf-xet-builder
# Install hf-xet
WORKDIR /tmp
ENV CARGO_HOME=/root/.cargo
ENV RUSTUP_HOME=/root/.rustup
ENV PATH="$CARGO_HOME/bin:$RUSTUP_HOME/bin:$PATH"
RUN --mount=type=cache,target=/root/.cache/uv \
--mount=type=bind,from=rust,source=/root/.cargo,target=/root/.cargo,rw \
--mount=type=bind,from=rust,source=/root/.rustup,target=/root/.rustup,rw \
git clone https://github.com/huggingface/xet-core.git && \
cd xet-core/hf_xet/ && \
uv pip install maturin patchelf && \
python -m maturin build --release --out dist && \
mkdir -p /tmp/hf-xet/dist && \
cp dist/*.whl /tmp/hf-xet/dist/
# Build numba
FROM python-install AS numba-builder
ARG MAX_JOBS
ARG NUMBA_VERSION=0.61.2
WORKDIR /tmp
# Clone all required dependencies
RUN --mount=type=cache,target=/root/.cache/uv \
microdnf install ninja-build gcc gcc-c++ -y && \
git clone --recursive https://github.com/llvm/llvm-project.git -b llvmorg-15.0.7 && \
git clone --recursive https://github.com/numba/llvmlite.git -b v0.44.0 && \
git clone --recursive https://github.com/numba/numba.git -b ${NUMBA_VERSION} && \
cd llvm-project && mkdir build && cd build && \
uv pip install 'cmake<4' setuptools numpy && \
export PREFIX=/usr/local && CMAKE_ARGS="${CMAKE_ARGS} -DLLVM_ENABLE_PROJECTS=lld;libunwind;compiler-rt" \
CFLAGS="$(echo $CFLAGS | sed 's/-fno-plt //g')" \
CXXFLAGS="$(echo $CXXFLAGS | sed 's/-fno-plt //g')" \
CMAKE_ARGS="${CMAKE_ARGS} -DFFI_INCLUDE_DIR=$PREFIX/include" \
CMAKE_ARGS="${CMAKE_ARGS} -DFFI_LIBRARY_DIR=$PREFIX/lib" \
cmake -DCMAKE_INSTALL_PREFIX="${PREFIX}" \
-DCMAKE_BUILD_TYPE=Release \
-DCMAKE_LIBRARY_PATH="${PREFIX}" \
-DLLVM_ENABLE_LIBEDIT=OFF \
-DLLVM_ENABLE_LIBXML2=OFF \
-DLLVM_ENABLE_RTTI=ON \
-DLLVM_ENABLE_TERMINFO=OFF \
-DLLVM_INCLUDE_BENCHMARKS=OFF \
-DLLVM_INCLUDE_DOCS=OFF \
-DLLVM_INCLUDE_EXAMPLES=OFF \
-DLLVM_INCLUDE_GO_TESTS=OFF \
-DLLVM_INCLUDE_TESTS=OFF \
-DLLVM_INCLUDE_UTILS=ON \
-DLLVM_INSTALL_UTILS=ON \
-DLLVM_UTILS_INSTALL_DIR=libexec/llvm \
-DLLVM_BUILD_LLVM_DYLIB=OFF \
-DLLVM_LINK_LLVM_DYLIB=OFF \
-DLLVM_EXPERIMENTAL_TARGETS_TO_BUILD=WebAssembly \
-DLLVM_ENABLE_FFI=ON \
-DLLVM_ENABLE_Z3_SOLVER=OFF \
-DLLVM_OPTIMIZED_TABLEGEN=ON \
-DCMAKE_POLICY_DEFAULT_CMP0111=NEW \
-DCOMPILER_RT_BUILD_BUILTINS=ON \
-DCOMPILER_RT_BUILTINS_HIDE_SYMBOLS=OFF \
-DCOMPILER_RT_BUILD_LIBFUZZER=OFF \
-DCOMPILER_RT_BUILD_CRT=OFF \
-DCOMPILER_RT_BUILD_MEMPROF=OFF \
-DCOMPILER_RT_BUILD_PROFILE=OFF \
-DCOMPILER_RT_BUILD_SANITIZERS=OFF \
-DCOMPILER_RT_BUILD_XRAY=OFF \
-DCOMPILER_RT_BUILD_GWP_ASAN=OFF \
-DCOMPILER_RT_BUILD_ORC=OFF \
-DCOMPILER_RT_INCLUDE_TESTS=OFF \
${CMAKE_ARGS} -GNinja ../llvm \
&& ninja install . && \
# build llvmlite
cd ../../llvmlite && python setup.py bdist_wheel && \
cd ../numba && \
if ! grep '#include "dynamic_annotations.h"' numba/_dispatcher.cpp; then \
sed -i '/#include "internal\/pycore_atomic.h"/i\#include "dynamic_annotations.h"' numba/_dispatcher.cpp; \
fi && python setup.py bdist_wheel
# Build Outlines Core
FROM python-install AS outlines-core-builder
WORKDIR /tmp
ENV CARGO_HOME=/root/.cargo
ENV RUSTUP_HOME=/root/.rustup
ENV PATH="$CARGO_HOME/bin:$RUSTUP_HOME/bin:$PATH"
COPY requirements/common.txt /tmp/requirements/common.txt
ARG OUTLINES_CORE_VERSION
RUN --mount=type=cache,target=/root/.cache/uv \
--mount=type=bind,from=rust,source=/root/.cargo,target=/root/.cargo,rw \
--mount=type=bind,from=rust,source=/root/.rustup,target=/root/.rustup,rw \
OUTLINES_CORE_VERSION=${OUTLINES_CORE_VERSION:-$(grep -E '^outlines_core\s*==\s*[0-9.]+' /tmp/requirements/common.txt | grep -Eo '[0-9.]+')} && \
if [ -z "${OUTLINES_CORE_VERSION}" ]; then echo "ERROR: Could not determine outlines_core version"; exit 1; fi && \
git clone https://github.com/dottxt-ai/outlines-core.git && \
cd outlines-core && \
git checkout tags/${OUTLINES_CORE_VERSION} && \
sed -i "s/version = \"0.0.0\"/version = \"${OUTLINES_CORE_VERSION}\"/" Cargo.toml && \
uv pip install maturin && \
python -m maturin build --release --out dist
# Final build stage
FROM python-install AS vllm-cpu
ARG PYTHON_VERSION
# Set correct library path for torch and numactl
ENV LD_LIBRARY_PATH="/opt/vllm/lib64/python${PYTHON_VERSION}/site-packages/torch/lib:/usr/local/lib:/opt/rh/gcc-toolset-14/root/usr/lib64:$LD_LIBRARY_PATH"
ENV C_INCLUDE_PATH="/usr/local/include:$C_INCLUDE_PATH"
ENV UV_LINK_MODE=copy
ENV CARGO_HOME=/root/.cargo
ENV RUSTUP_HOME=/root/.rustup
ENV GRPC_PYTHON_BUILD_SYSTEM_OPENSSL=1
ENV PCP_DIR=/opt/rh/gcc-toolset-14/root
ENV PKG_CONFIG_PATH="/opt/rh/gcc-toolset-14/root/usr/lib64/pkgconfig:/usr/local/lib/pkgconfig/"
ENV PATH="${VIRTUAL_ENV:+${VIRTUAL_ENV}/bin}:/opt/rh/gcc-toolset-14/root/usr/bin:/usr/local/bin:$CARGO_HOME/bin:$RUSTUP_HOME/bin:$PATH"
COPY . /workspace/vllm
WORKDIR /workspace/vllm
RUN --mount=type=bind,from=numa-build,src=/tmp/numactl-2.0.16,target=/numactl \
make -C /numactl install
# Install dependencies, including PyTorch and Apache Arrow
RUN --mount=type=cache,target=/root/.cache/uv \
--mount=type=bind,from=rust,source=/root/.cargo,target=/root/.cargo,rw \
--mount=type=bind,from=rust,source=/root/.rustup,target=/root/.rustup,rw \
--mount=type=bind,from=pyarrow,source=/tmp/arrow/python/dist,target=/tmp/arrow-wheels \
--mount=type=bind,from=torch-vision,source=/tmp/vision/dist,target=/tmp/vision-wheels/ \
--mount=type=bind,from=hf-xet-builder,source=/tmp/hf-xet/dist,target=/tmp/hf-xet-wheels/ \
--mount=type=bind,from=numba-builder,source=/tmp/llvmlite/dist,target=/tmp/llvmlite-wheels/ \
--mount=type=bind,from=numba-builder,source=/tmp/numba/dist,target=/tmp/numba-wheels/ \
--mount=type=bind,from=outlines-core-builder,source=/tmp/outlines-core/dist,target=/tmp/outlines-core/dist/ \
sed -i '/^torch/d' requirements/build.txt && \
ARROW_WHL_FILE=$(ls /tmp/arrow-wheels/pyarrow-*.whl) && \
VISION_WHL_FILE=$(ls /tmp/vision-wheels/*.whl) && \
HF_XET_WHL_FILE=$(ls /tmp/hf-xet-wheels/*.whl) && \
LLVM_WHL_FILE=$(ls /tmp/llvmlite-wheels/*.whl) && \
NUMBA_WHL_FILE=$(ls /tmp/numba-wheels/*.whl) && \
OUTLINES_CORE_WHL_FILE=$(ls /tmp/outlines-core/dist/*.whl) && \
uv pip install -v \
$ARROW_WHL_FILE \
$VISION_WHL_FILE \
$HF_XET_WHL_FILE \
$LLVM_WHL_FILE \
$NUMBA_WHL_FILE \
$OUTLINES_CORE_WHL_FILE \
--index-strategy unsafe-best-match \
-r requirements/build.txt \
-r requirements/cpu.txt
# Build and install vllm
RUN --mount=type=cache,target=/root/.cache/uv \
VLLM_TARGET_DEVICE=cpu VLLM_CPU_MOE_PREPACK=0 python setup.py bdist_wheel && \
uv pip install "$(echo dist/*.whl)[tensorizer]"
# setup non-root user for vllm
RUN umask 002 && \
useradd --uid 2000 --gid 0 vllm && \
mkdir -p /home/vllm && \
chmod g+rwx /home/vllm
COPY LICENSE /licenses/vllm.md
COPY examples/*.jinja /app/data/template/
USER 2000
WORKDIR /home/vllm
# Set the default entrypoint
ENTRYPOINT ["vllm", "serve"]
Reference in New Issue View Git Blame Copy Permalink