EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity

feat: unify dockerfiles (#10705)

Browse Source 
Co-authored-by: Yineng Zhang <me@zhyncs.com>
This commit is contained in:
ishandhanani
2025-09-23 14:23:48 +08:00
committed by GitHub
parent 9241f4fd20
commit 1c82d9db28
5 changed files with 216 additions and 62 deletions
Download Patch File Download Diff File Expand all files Collapse all files

114
.github/workflows/release-docker-dev.yml vendored
View File

@@ -6,7 +6,7 @@ on:
- cron: '0 0 * * *'
jobs:
build-dev:
build-dev-x86:
if: ${{ github.repository == 'sgl-project/sglang' }}
runs-on: ubuntu-22.04
strategy:
@@ -15,6 +15,46 @@ jobs:
- version: 12.9.1
type: all
tag: dev
steps:
- name: Checkout repository
uses: actions/checkout@v4
- name: Free disk space
uses: jlumbroso/free-disk-space@main
with:
tool-cache: false
docker-images: false
android: true
dotnet: true
haskell: true
large-packages: true
swap-storage: false
- name: Set up Docker Buildx
uses: docker/setup-buildx-action@v3
- name: Login to Docker Hub
uses: docker/login-action@v2
with:
username: ${{ secrets.DOCKERHUB_USERNAME }}
password: ${{ secrets.DOCKERHUB_TOKEN }}
- name: Build and Push Dev Image (x86)
run: |
docker buildx build --platform linux/amd64 --push -f docker/Dockerfile --build-arg CUDA_VERSION=${{ matrix.variant.version }} --build-arg BUILD_TYPE=${{ matrix.variant.type }} --build-arg CMAKE_BUILD_PARALLEL_LEVEL=$(nproc) -t lmsysorg/sglang:${{ matrix.variant.tag }} --no-cache .
build-blackwell-x86:
if: ${{ github.repository == 'sgl-project/sglang' }}
runs-on: ubuntu-latest
strategy:
matrix:
variant:
- version: 12.8.1
type: blackwell
tag: blackwell
- version: 12.9.1
type: blackwell
tag: blackwell-cu129
steps:
- name: Checkout repository
@@ -31,13 +71,79 @@ jobs:
large-packages: true
swap-storage: false
- name: Set up Docker Buildx
uses: docker/setup-buildx-action@v3
- name: Login to Docker Hub
uses: docker/login-action@v2
with:
username: ${{ secrets.DOCKERHUB_USERNAME }}
password: ${{ secrets.DOCKERHUB_TOKEN }}
- name: Build and Push Dev Image
- name: Build and Push Blackwell Image (x86)
run: |
docker buildx build --output type=image,compression=zstd . -f docker/Dockerfile --build-arg CUDA_VERSION=${{ matrix.variant.version }} --build-arg BUILD_TYPE=${{ matrix.variant.type }} --build-arg CMAKE_BUILD_PARALLEL_LEVEL=$(nproc) -t lmsysorg/sglang:${{ matrix.variant.tag }} --no-cache
docker push lmsysorg/sglang:${{ matrix.variant.tag }}
if [ "${{ matrix.variant.version }}" = "12.9.1" ]; then
docker buildx build --platform linux/amd64 --push -f docker/Dockerfile --build-arg CUDA_VERSION=${{ matrix.variant.version }} --build-arg BUILD_TYPE=${{ matrix.variant.type }} --build-arg CMAKE_BUILD_PARALLEL_LEVEL=$(nproc) -t lmsysorg/sglang:${{ matrix.variant.tag }}-x86 --no-cache .
else
docker buildx build --platform linux/amd64 --push -f docker/Dockerfile --build-arg CUDA_VERSION=${{ matrix.variant.version }} --build-arg BUILD_TYPE=${{ matrix.variant.type }} --build-arg CMAKE_BUILD_PARALLEL_LEVEL=$(nproc) -t lmsysorg/sglang:${{ matrix.variant.tag }} --no-cache .
fi
build-blackwell-arm:
if: ${{ github.repository == 'sgl-project/sglang' }}
runs-on: ubuntu-22.04-arm
strategy:
matrix:
variant:
- version: 12.9.1
type: blackwell_aarch
tag: blackwell-cu129
steps:
- name: Checkout repository
uses: actions/checkout@v4
- name: Free disk space
uses: jlumbroso/free-disk-space@main
with:
tool-cache: false
docker-images: false
android: true
dotnet: true
haskell: true
large-packages: true
swap-storage: false
- name: Set up Docker Buildx
uses: docker/setup-buildx-action@v3
- name: Login to Docker Hub
uses: docker/login-action@v2
with:
username: ${{ secrets.DOCKERHUB_USERNAME }}
password: ${{ secrets.DOCKERHUB_TOKEN }}
- name: Build and Push Blackwell Image (ARM)
run: |
docker buildx build --platform linux/arm64 --push -f docker/Dockerfile --build-arg CUDA_VERSION=${{ matrix.variant.version }} --build-arg BUILD_TYPE=${{ matrix.variant.type }} --build-arg CMAKE_BUILD_PARALLEL_LEVEL=$(nproc) -t lmsysorg/sglang:${{ matrix.variant.tag }}-arm64 --no-cache .
create-manifests:
runs-on: ubuntu-22.04
needs: [build-blackwell-x86, build-blackwell-arm]
if: ${{ github.repository == 'sgl-project/sglang' }}
strategy:
matrix:
variant:
- tag: blackwell-cu129
x86_tag: blackwell-cu129-x86
arm64_tag: blackwell-cu129-arm64
steps:
- uses: docker/setup-buildx-action@v3
- uses: docker/login-action@v2
with:
username: ${{ secrets.DOCKERHUB_USERNAME }}
password: ${{ secrets.DOCKERHUB_TOKEN }}
- run: |
docker buildx imagetools create \
-t lmsysorg/sglang:${{ matrix.variant.tag }} \
lmsysorg/sglang:${{ matrix.variant.x86_tag }} \
lmsysorg/sglang:${{ matrix.variant.arm64_tag }}

5
.github/workflows/release-docker-gb200.yml vendored
View File

@@ -1,10 +1,5 @@
name: Release Docker Images (GB200)
on:
push:
branches:
- main
paths:
- "python/sglang/version.py"
workflow_dispatch:
jobs:

89
.github/workflows/release-docker.yml vendored
View File

@@ -8,14 +8,19 @@ on:
workflow_dispatch:
jobs:
publish:
publish-x86:
if: github.repository == 'sgl-project/sglang'
runs-on: ubuntu-latest
environment: 'prod'
strategy:
matrix:
cuda_version: ['12.6.1', '12.9.1']
build_type: ['all']
variant:
- cuda_version: '12.6.1'
build_type: 'all'
- cuda_version: '12.8.1'
build_type: 'blackwell'
- cuda_version: '12.9.1'
build_type: 'blackwell'
runs-on: ubuntu-latest
steps:
- name: Delete huge unnecessary tools folder
run: rm -rf /opt/hostedtoolcache
@@ -34,29 +39,24 @@ jobs:
large-packages: true
swap-storage: false
- name: Set up Docker Buildx
uses: docker/setup-buildx-action@v3
- name: Login to Docker Hub
uses: docker/login-action@v2
with:
username: ${{ secrets.DOCKERHUB_USERNAME }}
password: ${{ secrets.DOCKERHUB_TOKEN }}
- name: Build and Push
- name: Build and Push AMD64
run: |
version=$(cat python/sglang/version.py | cut -d'"' -f2)
if [ "${{ matrix.cuda_version }}" = "11.8.0" ]; then
cuda_tag="cu118"
elif [ "${{ matrix.cuda_version }}" = "12.1.1" ]; then
cuda_tag="cu121"
elif [ "${{ matrix.cuda_version }}" = "12.4.1" ]; then
cuda_tag="cu124"
elif [ "${{ matrix.cuda_version }}" = "12.5.1" ]; then
cuda_tag="cu125"
elif [ "${{ matrix.cuda_version }}" = "12.6.1" ]; then
if [ "${{ matrix.variant.cuda_version }}" = "12.6.1" ]; then
cuda_tag="cu126"
elif [ "${{ matrix.cuda_version }}" = "12.8.1" ]; then
elif [ "${{ matrix.variant.cuda_version }}" = "12.8.1" ]; then
cuda_tag="cu128"
elif [ "${{ matrix.cuda_version }}" = "12.9.1" ]; then
elif [ "${{ matrix.variant.cuda_version }}" = "12.9.1" ]; then
cuda_tag="cu129"
else
echo "Unsupported CUDA version"
@@ -65,25 +65,58 @@ jobs:
tag=v${version}-${cuda_tag}
if [ "${{ matrix.build_type }}" = "all" ]; then
if [ "${{ matrix.variant.build_type }}" = "all" ]; then
tag_suffix=""
elif [ "${{ matrix.build_type }}" = "srt" ]; then
tag_suffix="-srt"
elif [ "${{ matrix.build_type }}" = "blackwell" ]; then
elif [ "${{ matrix.variant.build_type }}" = "blackwell" ]; then
tag_suffix="-b200"
else
echo "Unsupported build type"
exit 1
fi
docker buildx build --output type=image,compression=zstd . -f docker/Dockerfile --build-arg CUDA_VERSION=${{ matrix.cuda_version }} --build-arg BUILD_TYPE=${{ matrix.build_type }} -t lmsysorg/sglang:${tag}${tag_suffix} --no-cache
if [ "${{ matrix.cuda_version }}" = "12.6.1" ]; then
docker tag lmsysorg/sglang:${tag}${tag_suffix} lmsysorg/sglang:latest${tag_suffix}
docker push lmsysorg/sglang:latest${tag_suffix}
if [ "${{ matrix.variant.cuda_version }}" = "12.9.1" ]; then
docker buildx build --platform linux/amd64 --push -f docker/Dockerfile --build-arg CUDA_VERSION=${{ matrix.variant.cuda_version }} --build-arg BUILD_TYPE=${{ matrix.variant.build_type }} -t lmsysorg/sglang:${tag}${tag_suffix} -t lmsysorg/sglang:latest --no-cache .
else
docker buildx build --platform linux/amd64 --push -f docker/Dockerfile --build-arg CUDA_VERSION=${{ matrix.variant.cuda_version }} --build-arg BUILD_TYPE=${{ matrix.variant.build_type }} -t lmsysorg/sglang:${tag}${tag_suffix} --no-cache .
fi
if [ "${{ matrix.cuda_version }}" = "12.9.1" ]; then
docker tag lmsysorg/sglang:${tag}${tag_suffix} lmsysorg/sglang:v${version}
docker push lmsysorg/sglang:v${version}
publish-arm64:
if: github.repository == 'sgl-project/sglang'
environment: 'prod'
strategy:
matrix:
variant:
- cuda_version: '12.9.1'
build_type: 'blackwell_aarch'
runs-on: ubuntu-22.04-arm
steps:
- name: Delete huge unnecessary tools folder
run: rm -rf /opt/hostedtoolcache
- name: Checkout repository
uses: actions/checkout@v4
- name: Set up Docker Buildx
uses: docker/setup-buildx-action@v3
- name: Login to Docker Hub
uses: docker/login-action@v2
with:
username: ${{ secrets.DOCKERHUB_USERNAME }}
password: ${{ secrets.DOCKERHUB_TOKEN }}
- name: Build and Push ARM64
run: |
version=$(cat python/sglang/version.py | cut -d'"' -f2)
if [ "${{ matrix.variant.cuda_version }}" = "12.9.1" ]; then
cuda_tag="cu129"
else
echo "Unsupported CUDA version"
exit 1
fi
tag=v${version}-${cuda_tag}
tag_suffix="-gb200"
docker buildx build --platform linux/arm64 --push -f docker/Dockerfile --build-arg CUDA_VERSION=${{ matrix.variant.cuda_version }} --build-arg BUILD_TYPE=${{ matrix.variant.build_type }} -t lmsysorg/sglang:${tag}${tag_suffix} --no-cache .

8
.github/workflows/release-whl-kernel.yml vendored
View File

@@ -44,7 +44,7 @@ jobs:
working-directory: sgl-kernel
run: |
pip install twine
python3 -m twine upload dist/* -u __token__ -p ${{ secrets.PYPI_TOKEN }}
python3 -m twine upload --skip-existing dist/* -u __token__ -p ${{ secrets.PYPI_TOKEN }}
build-cu124:
if: github.repository == 'sgl-project/sglang'
@@ -227,6 +227,12 @@ jobs:
chmod +x ./build.sh
./build.sh "${{ matrix.python-version }}" "${{ matrix.cuda-version }}" aarch64
- name: Upload to PyPI
working-directory: sgl-kernel
run: |
pip install twine
python3 -m twine upload --skip-existing dist/* -u __token__ -p ${{ secrets.PYPI_TOKEN }}
- name: Upload artifacts
uses: actions/upload-artifact@v4
with:

62
docker/Dockerfile
View File

@@ -1,10 +1,11 @@
ARG CUDA_VERSION=12.9.1
FROM nvidia/cuda:${CUDA_VERSION}-cudnn-devel-ubuntu22.04 as base
FROM nvidia/cuda:${CUDA_VERSION}-cudnn-devel-ubuntu22.04 AS base
ARG BUILD_TYPE=all
ARG BRANCH_TYPE=remote
ARG DEEPEP_COMMIT=9af0e0d0e74f3577af1979c9b9e1ac2cad0104ee
ARG CMAKE_BUILD_PARALLEL_LEVEL=2
ARG SGL_KERNEL_VERSION=0.3.12
ENV DEBIAN_FRONTEND=noninteractive \
CUDA_HOME=/usr/local/cuda \
GDRCOPY_HOME=/usr/src/gdrdrv-2.4.4/ \
@@ -57,7 +58,7 @@ RUN mkdir -p /tmp/gdrcopy && cd /tmp \
&& cd / && rm -rf /tmp/gdrcopy
# Fix DeepEP IBGDA symlink
RUN ln -sf /usr/lib/x86_64-linux-gnu/libmlx5.so.1 /usr/lib/x86_64-linux-gnu/libmlx5.so
RUN ln -sf /usr/lib/$(uname -m)-linux-gnu/libmlx5.so.1 /usr/lib/$(uname -m)-linux-gnu/libmlx5.so
FROM scratch AS local_src
COPY . /src
@@ -81,24 +82,33 @@ RUN python3 -m pip install --no-cache-dir --upgrade pip setuptools wheel html5li
12.9.1) CUINDEX=129 ;; \
*) echo "Unsupported CUDA version: $CUDA_VERSION" && exit 1 ;; \
esac \
&& if [ "$CUDA_VERSION" = "12.6.1" ]; then \
python3 -m pip install --no-cache-dir https://github.com/sgl-project/whl/releases/download/v${SGL_KERNEL_VERSION}/sgl_kernel-${SGL_KERNEL_VERSION}+cu124-cp310-abi3-manylinux2014_$(uname -m).whl --force-reinstall --no-deps ; \
fi \
&& if [ "$CUDA_VERSION" = "12.8.1" ] || [ "$CUDA_VERSION" = "12.9.1" ]; then \
python3 -m pip install --no-cache-dir sgl-kernel==${SGL_KERNEL_VERSION} ; \
fi \
&& python3 -m pip install --no-cache-dir -e "python[${BUILD_TYPE}]" --extra-index-url https://download.pytorch.org/whl/cu${CUINDEX} \
&& python3 -m pip install --no-cache-dir nvidia-nccl-cu12==2.27.6 --force-reinstall --no-deps \
&& python3 -m flashinfer --download-cubin \
&& if [ "$CUDA_VERSION" = "12.6.1" ]; then \
python3 -m pip install --no-cache-dir https://github.com/sgl-project/whl/releases/download/v0.3.11/sgl_kernel-0.3.11+cu124-cp310-abi3-manylinux2014_x86_64.whl --force-reinstall --no-deps ; \
fi
&& python3 -m flashinfer --download-cubin
# Download source files
RUN wget https://developer.download.nvidia.com/compute/redist/nvshmem/3.3.9/source/nvshmem_src_cuda12-all-all-3.3.9.tar.gz && \
git clone https://github.com/deepseek-ai/DeepEP.git && \
cd DeepEP && git checkout ${DEEPEP_COMMIT} && sed -i 's/#define NUM_CPU_TIMEOUT_SECS 100/#define NUM_CPU_TIMEOUT_SECS 1000/' csrc/kernels/configs.cuh && \
cd .. && \
tar -xf nvshmem_src_cuda12-all-all-3.3.9.tar.gz && \
mv nvshmem_src nvshmem && \
rm -f /sgl-workspace/nvshmem_src_cuda12-all-all-3.3.9.tar.gz
if [ "$BUILD_TYPE" = "blackwell" ] && [ "$(uname -m)" = "aarch64" ]; then \
git clone https://github.com/fzyzcjy/DeepEP.git \
&& cd DeepEP && git checkout 1b14ad661c7640137fcfe93cccb2694ede1220b0 && sed -i 's/#define NUM_CPU_TIMEOUT_SECS 100/#define NUM_CPU_TIMEOUT_SECS 1000/' csrc/kernels/configs.cuh && cd .. ; \
else \
git clone https://github.com/deepseek-ai/DeepEP.git \
&& cd DeepEP && git checkout ${DEEPEP_COMMIT} && sed -i 's/#define NUM_CPU_TIMEOUT_SECS 100/#define NUM_CPU_TIMEOUT_SECS 1000/' csrc/kernels/configs.cuh && cd .. ; \
fi \
&& tar -xf nvshmem_src_cuda12-all-all-3.3.9.tar.gz \
&& mv nvshmem_src nvshmem \
&& rm -f /sgl-workspace/nvshmem_src_cuda12-all-all-3.3.9.tar.gz
# Build and install NVSHMEM
RUN cd /sgl-workspace/nvshmem && \
if [ "$BUILD_TYPE" = "blackwell" ] || [ "$BUILD_TYPE" = "blackwell_aarch" ]; then CUDA_ARCH="90;100;120"; else CUDA_ARCH="90"; fi && \
NVSHMEM_SHMEM_SUPPORT=0 \
NVSHMEM_UCX_SUPPORT=0 \
NVSHMEM_USE_NCCL=0 \
@@ -107,7 +117,7 @@ RUN cd /sgl-workspace/nvshmem && \
NVSHMEM_PMIX_SUPPORT=0 \
NVSHMEM_TIMEOUT_DEVICE_POLLING=0 \
NVSHMEM_USE_GDRCOPY=1 \
cmake -S . -B build/ -DCMAKE_INSTALL_PREFIX=${NVSHMEM_DIR} -DCMAKE_CUDA_ARCHITECTURES="90" && \
cmake -S . -B build/ -DCMAKE_INSTALL_PREFIX=${NVSHMEM_DIR} -DCMAKE_CUDA_ARCHITECTURES=${CUDA_ARCH} && \
cmake --build build --target install -j${CMAKE_BUILD_PARALLEL_LEVEL}
# Install DeepEP
@@ -125,6 +135,7 @@ RUN cd /sgl-workspace/DeepEP && \
esac && \
NVSHMEM_DIR=${NVSHMEM_DIR} TORCH_CUDA_ARCH_LIST="${CHOSEN_TORCH_CUDA_ARCH_LIST}" pip install .
# Python tools
RUN python3 -m pip install --no-cache-dir \
datamodel_code_generator \
@@ -169,16 +180,16 @@ RUN apt-get update && apt-get install -y \
RUN apt update -y \
&& apt install -y --no-install-recommends gnupg \
&& echo "deb http://developer.download.nvidia.com/devtools/repos/ubuntu2004/amd64 /" | tee /etc/apt/sources.list.d/nvidia-devtools.list \
&& apt-key adv --fetch-keys http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/7fa2af80.pub \
&& echo "deb http://developer.download.nvidia.com/devtools/repos/ubuntu2004/$(if [ "$(uname -m)" = "aarch64" ]; then echo "arm64"; else echo "amd64"; fi) /" | tee /etc/apt/sources.list.d/nvidia-devtools.list \
&& apt-key adv --fetch-keys http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/$(if [ "$(uname -m)" = "aarch64" ]; then echo "arm64"; else echo "x86_64"; fi)/7fa2af80.pub \
&& apt update -y \
&& apt install nsight-systems-cli -y
# Set up locale
RUN locale-gen en_US.UTF-8
ENV LANG en_US.UTF-8
ENV LANGUAGE en_US:en
ENV LC_ALL en_US.UTF-8
ENV LANG=en_US.UTF-8
ENV LANGUAGE=en_US:en
ENV LC_ALL=en_US.UTF-8
# Install minimal Python packages
RUN python3 -m pip install --no-cache-dir --break-system-packages \
@@ -186,7 +197,7 @@ RUN python3 -m pip install --no-cache-dir --break-system-packages \
black \
isort \
icdiff \
scikit_build_core \
scikit-build-core \
uv \
pre-commit \
pandas \
@@ -209,11 +220,14 @@ RUN curl -L https://github.com/clangd/clangd/releases/download/18.1.3/clangd-lin
&& rm -rf clangd_18.1.3 clangd.zip
# Install CMake
RUN wget https://github.com/Kitware/CMake/releases/download/v3.31.1/cmake-3.31.1-linux-x86_64.tar.gz \
&& tar -xzf cmake-3.31.1-linux-x86_64.tar.gz \
&& cp -r cmake-3.31.1-linux-x86_64/bin/* /usr/local/bin/ \
&& cp -r cmake-3.31.1-linux-x86_64/share/* /usr/local/share/ \
&& rm -rf cmake-3.31.1-linux-x86_64 cmake-3.31.1-linux-x86_64.tar.gz
RUN CMAKE_VERSION=3.31.1 \
&& ARCH=$(uname -m) \
&& CMAKE_INSTALLER="cmake-${CMAKE_VERSION}-linux-${ARCH}" \
&& wget "https://github.com/Kitware/CMake/releases/download/v${CMAKE_VERSION}/${CMAKE_INSTALLER}.tar.gz" \
&& tar -xzf "${CMAKE_INSTALLER}.tar.gz" \
&& cp -r "${CMAKE_INSTALLER}/bin/"* /usr/local/bin/ \
&& cp -r "${CMAKE_INSTALLER}/share/"* /usr/local/share/ \
&& rm -rf "${CMAKE_INSTALLER}" "${CMAKE_INSTALLER}.tar.gz"
# Install Rust toolchain for sgl-router
ENV PATH="/root/.cargo/bin:${PATH}"