init v0.23.0

Signed-off-by: Sun Ruoxi <sunruoxi@4paradigm.com>
This commit is contained in:
2026-08-27 15:11:51 +08:00
parent b582a8e7d1
commit 7f8a1b1f7a
2849 changed files with 712887 additions and 22001 deletions

View File

@@ -15,45 +15,93 @@
# This file is a part of the vllm-ascend project.
#
FROM quay.io/ascend/cann:8.2.rc1-310p-openeuler24.03-py3.11
FROM quay.io/ascend/cann:9.1.0-310p-openeuler24.03-py3.12
ARG PIP_INDEX_URL="https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple"
ARG COMPILE_CUSTOM_KERNELS=1
ENV COMPILE_CUSTOM_KERNELS=${COMPILE_CUSTOM_KERNELS}
RUN yum update -y && \
yum install -y python3-pip git vim wget net-tools gcc gcc-c++ make cmake numactl-devel && \
rm -rf /var/cache/yum
RUN pip config set global.index-url ${PIP_INDEX_URL}
WORKDIR /workspace
COPY . /vllm-workspace/vllm-ascend/
RUN yum update -y && \
yum install -y python3-pip git vim wget net-tools gcc gcc-c++ make cmake numactl numactl-devel jemalloc patch && \
rm -rf /var/cache/yum
# Install modelscope (for fast download) and ray (for multinode)
RUN pip config set global.index-url ${PIP_INDEX_URL} && \
python3 -m pip install modelscope 'ray>=2.47.1,<=2.48.0' 'protobuf>3.20.0' && \
python3 -m pip cache purge
# Install vLLM
ARG VLLM_REPO=https://github.com/vllm-project/vllm.git
ARG VLLM_TAG=v0.11.0rc3
RUN git clone --depth 1 $VLLM_REPO --branch $VLLM_TAG /vllm-workspace/vllm
ARG VLLM_TAG=v0.23.0
ARG VLLM_COMMIT=""
RUN if [ -n "$VLLM_COMMIT" ]; then \
git init /vllm-workspace/vllm && \
git -C /vllm-workspace/vllm fetch --depth 1 $VLLM_REPO "$VLLM_COMMIT" && \
git -C /vllm-workspace/vllm checkout FETCH_HEAD; \
else \
git clone --depth 1 -b $VLLM_TAG $VLLM_REPO /vllm-workspace/vllm; \
fi
# In x86, triton will be installed by vllm. But in Ascend, triton doesn't work correctly. we need to uninstall it.
RUN VLLM_TARGET_DEVICE="empty" python3 -m pip install -e /vllm-workspace/vllm/ --extra-index https://download.pytorch.org/whl/cpu/ && \
RUN VLLM_TARGET_DEVICE="empty" python3 -m pip install -e /vllm-workspace/vllm/[audio] --extra-index https://download.pytorch.org/whl/cpu/ && \
python3 -m pip uninstall -y triton && \
python3 -m pip cache purge
# Install vllm-ascend
RUN export PIP_EXTRA_INDEX_URL=https://mirrors.huaweicloud.com/ascend/repos/pypi && \
ARG SOC_VERSION="ascend310p1"
ARG COMPILE_CUSTOM_KERNELS=1
ENV SOC_VERSION=$SOC_VERSION \
TASK_QUEUE_ENABLE=1 \
OMP_NUM_THREADS=1
COPY . /vllm-workspace/vllm-ascend/
RUN export PIP_EXTRA_INDEX_URL="https://mirrors.huaweicloud.com/ascend/repos/pypi" && \
source /usr/local/Ascend/ascend-toolkit/set_env.sh && \
source /usr/local/Ascend/nnal/atb/set_env.sh && \
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/Ascend/ascend-toolkit/latest/`uname -i`-linux/devlib && \
export CPLUS_INCLUDE_PATH=$CPLUS_INCLUDE_PATH:/usr/include/c++/12:/usr/include/c++/12/`uname -i`-openEuler-linux && \
export SOC_VERSION=ASCEND310P3 && \
python3 -m pip install -v -e /vllm-workspace/vllm-ascend/ --extra-index https://download.pytorch.org/whl/cpu/ && \
python3 -m pip install -e /vllm-workspace/vllm-ascend/ --extra-index https://download.pytorch.org/whl/cpu/ && \
python3 -m pip uninstall -y triton-ascend triton && \
python3 -m pip cache purge
# Install modelscope (for fast download) and ray (for multinode)
RUN python3 -m pip install modelscope 'ray>=2.47.1' 'protobuf>3.20.0' && \
python3 -m pip cache purge
RUN echo "export LD_PRELOAD=/usr/lib64/libjemalloc.so.2:$LD_PRELOAD" >> ~/.bashrc
RUN echo "export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/lib" >> ~/.bashrc
RUN echo 'export LD_LIBRARY_PATH=/usr/local/Ascend/driver/lib64:/usr/lib64:/root/aicpu_kernels/0/aicpu_kernels_device/sand_box:$LD_LIBRARY_PATH' >> /root/.bashrc
# Initialize driver directories, users and file groups
RUN mkdir -m 750 /var/driver -m 750 /var/dmp -m 750 /usr/slog && \
mkdir -m 755 /home/drv && \
mkdir -m 750 /home/drv/hdc_ppc -m 750 -p /var/log/npu/slog && \
ldconfig && \
ln -sf /lib /lib64 && \
groupadd -g 1000 HwHiAiUser && useradd -u 1000 -g HwHiAiUser -d /home/HwHiAiUser -m HwHiAiUser && \
groupadd -g 1101 HwDmUser && useradd -u 1101 -g HwDmUser -d /home/HwDmUser -m HwDmUser && \
groupadd -g 1102 HwBaseUser && useradd -u 1102 -g HwBaseUser -d /home/HwBaseUser -m HwBaseUser && \
groupadd -g 1100 HwSysUser && useradd -u 1100 -g HwSysUser -d /home/HwSysUser -m HwSysUser && \
usermod -a -G HwBaseUser HwHiAiUser && \
usermod -a -G HwDmUser HwHiAiUser && \
usermod -a -G HwBaseUser HwDmUser && \
usermod -a -G HwHiAiUser HwDmUser && \
usermod -a -G HwSysUser HwSysUser && \
chown HwDmUser:HwDmUser /var/dmp && \
chown HwHiAiUser:HwHiAiUser /var/driver && \
chown HwHiAiUser:HwHiAiUser /usr/slog && \
chown HwHiAiUser:HwHiAiUser /home/drv/hdc_ppc && \
chown HwHiAiUser:HwHiAiUser /var/log/npu/slog
# Dynamically generate entrypoint.sh for dual mode compatibility
RUN echo '#!/bin/bash' > /usr/local/bin/entrypoint.sh && \
echo 'if lspci 2>/dev/null | grep -qi "accelerators"; then' >> /usr/local/bin/entrypoint.sh && \
echo ' :' >> /usr/local/bin/entrypoint.sh && \
echo 'else' >> /usr/local/bin/entrypoint.sh && \
echo ' su - HwHiAiUser -c "export LD_LIBRARY_PATH=/usr/local/Ascend/driver/lib64/:/usr/lib64 && /lib/ld-linux-aarch64.so.1 /var/slogd -d >/dev/null &"' >> /usr/local/bin/entrypoint.sh && \
echo ' su - HwDmUser -c "export LD_LIBRARY_PATH=/usr/local/Ascend/driver/lib64/:/usr/lib64 && /lib/ld-linux-aarch64.so.1 /var/dmp_daemon -I -M -U 8087 &"' >> /usr/local/bin/entrypoint.sh && \
echo ' ' >> /usr/local/bin/entrypoint.sh && \
echo ' rm -rf /lib64/ld-linux-aarch64.so.1' >> /usr/local/bin/entrypoint.sh && \
echo ' ln -sf /lib/ld-linux-aarch64.so.1 /lib64/ld-linux-aarch64.so.1' >> /usr/local/bin/entrypoint.sh && \
echo 'fi' >> /usr/local/bin/entrypoint.sh && \
echo '' >> /usr/local/bin/entrypoint.sh && \
echo 'exec "$@"' >> /usr/local/bin/entrypoint.sh && \
chmod +x /usr/local/bin/entrypoint.sh
# Set entrypoint and default command
ENTRYPOINT ["/usr/local/bin/entrypoint.sh"]
CMD ["/bin/bash"]