inference/Dockerfile

FROM quay.io/ascend/cann:8.3.rc1.alpha003-910b-ubuntu22.04-py3.11
# Define environments
ENV DEBIAN_FRONTEND=noninteractive

ARG PY_BIN=/usr/local/python3.11.13/bin 
WORKDIR /vllm-workspace

# Download VLLM
ARG VLLM_REPO=https://github.com/vllm-project/vllm.git
ARG VLLM_TAG=v0.11.0rc3 
RUN git clone --depth 1  --branch $VLLM_TAG $VLLM_REPO

# Download VLLM-ASCEND
ARG VLLM_REPO=https://github.com/vllm-project/vllm-ascend.git
ARG VLLM_TAG=v0.11.0rc0
RUN git clone --depth 1  --branch $VLLM_TAG $VLLM_REPO


WORKDIR /vllm-workspace/vllm-ascend 
# patch
RUN  sed -i 's|https://mirrors.huaweicloud.com/ascend/repos/pypi|https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple|g' requirements.txt && \
sed -i 's/torch-npu==2.7.1.dev20250724/torch-npu==2.7.1/g' requirements.txt && \
sed -i 's/torch-npu==2.7.1.dev20250724/torch-npu==2.7.1/g' pyproject.toml && \
sed -i 's/torch==2.7.1/torch==2.7.1/g' pyproject.toml
# # ---------------------------
RUN git submodule update --init --recursive


#Install VLLM
WORKDIR /vllm-workspace/vllm
RUN 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  && \
    VLLM_USE_PRECOMPILED=0 && \
    VLLM_TARGET_DEVICE=empty ${PY_BIN}/pip install -v --no-deps -e .


#Install VLLM-ASCNED
WORKDIR /vllm-workspace/vllm-ascend
RUN 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  && \
VLLM_USE_PRECOMPILED=0 && \  
VLLM_TARGET_DEVICE=empty ${PY_BIN}/pip install -v --no-deps -e .