xc-llm-ascend/Dockerfile

#
# Copyright (c) 2025 Huawei Technologies Co., Ltd. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# This file is a part of the vllm-ascend project.
#

FROM quay.io/ascend/cann:8.5.1-910b-ubuntu22.04-py3.11

ARG PIP_INDEX_URL="https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple"
ARG MOONCAKE_TAG="v0.3.8.post1"
ARG SOC_VERSION="ascend910b1"

WORKDIR /workspace

COPY . /vllm-workspace/vllm-ascend/

# Install Mooncake dependencies
RUN apt-get update -y && \
    apt-get install -y git vim wget net-tools gcc g++ cmake numactl libnuma-dev libjemalloc2 && \
    git clone --depth 1 --branch ${MOONCAKE_TAG} https://github.com/kvcache-ai/Mooncake /vllm-workspace/Mooncake && \
    cp /vllm-workspace/vllm-ascend/tools/mooncake_installer.sh /vllm-workspace/Mooncake/ && \
    cd /vllm-workspace/Mooncake && bash mooncake_installer.sh -y && \
    ARCH=$(uname -m) && \
    source /usr/local/Ascend/ascend-toolkit/set_env.sh && \
    export LD_LIBRARY_PATH=/usr/local/Ascend/ascend-toolkit/latest/${ARCH}-linux/devlib:/usr/local/Ascend/ascend-toolkit/latest/${ARCH}-linux/lib64:$LD_LIBRARY_PATH && \
    mkdir -p build && cd build && cmake .. -DUSE_ASCEND_DIRECT=ON && \
    make -j$(nproc) && make install && \
    rm -fr /vllm-workspace/Mooncake/build && \
    rm -rf /var/cache/apt/* && \
    rm -rf /var/lib/apt/lists/*

RUN pip config set global.index-url ${PIP_INDEX_URL}

# Install vLLM
ARG VLLM_REPO=https://github.com/vllm-project/vllm.git
ARG VLLM_TAG=v0.18.0
RUN git clone --depth 1 $VLLM_REPO --branch $VLLM_TAG /vllm-workspace/vllm
# 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 -v -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
ENV DEBIAN_FRONTEND=noninteractive
ENV SOC_VERSION=$SOC_VERSION \
    TASK_QUEUE_ENABLE=1 \
    OMP_NUM_THREADS=1 \
    VLLM_WORKER_MULTIPROC_METHOD=spawn \
    VLLM_ASCEND_ENABLE_VNPU=1
# Append `libascend_hal.so` path (devlib) to LD_LIBRARY_PATH
# Installing vllm-ascend on x86 can pull upstream triton back in alongside triton-ascend. Remove it immediately after this step.
RUN export PIP_EXTRA_INDEX_URL=https://mirrors.huaweicloud.com/ascend/repos/pypi && \
    source /usr/local/Ascend/ascend-toolkit/set_env.sh && \
    source /usr/local/Ascend/nnal/atb/set_env.sh && \
    export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/Ascend/ascend-toolkit/latest/$(uname -i)-linux/devlib && \
    \
    cd /vllm-workspace/vllm-ascend/csrc/vnpu_offload && \
    make install && make clean && \
    python3 -m pip install -v -e /vllm-workspace/vllm-ascend/ --extra-index https://download.pytorch.org/whl/cpu/ && \
    \
    if [ "$(uname -i)" = "x86_64" ]; then python3 -m pip uninstall -y triton; fi && \
    \
    PYTHON_TAG=$(python3 -c "import sys; print(f'cp{sys.version_info.major}{sys.version_info.minor}')") && \
    ARCH=$(python3 -c "import platform; m=platform.machine().lower(); map={'x86_64':'x86_64','amd64':'x86_64','aarch64':'aarch64','arm64':'aarch64'}; print(map.get(m,m))") && \
    \
    if [ "$PYTHON_TAG" = "cp310" ] && [ "$ARCH" = "aarch64" ]; then PTA_WHEEL="torch_npu-2.9.0.post1%2Bgit4c901a4-${PYTHON_TAG}-${PYTHON_TAG}-manylinux_2_28_${ARCH}.whl"; \
    elif [ "$PYTHON_TAG" = "cp311" ] && [ "$ARCH" = "x86_64" ]; then PTA_WHEEL="torch_npu-2.9.0.post1%2Bgitdc51c2d-${PYTHON_TAG}-${PYTHON_TAG}-manylinux_2_28_${ARCH}.whl"; \
    elif [ "$PYTHON_TAG" = "cp310" ] && [ "$ARCH" = "x86_64" ]; then PTA_WHEEL="torch_npu-2.9.0.post1%2Bgita74051c-${PYTHON_TAG}-${PYTHON_TAG}-manylinux_2_28_${ARCH}.whl"; \
    elif [ "$PYTHON_TAG" = "cp311" ] && [ "$ARCH" = "aarch64" ]; then PTA_WHEEL="torch_npu-2.9.0.post1%2Bgitee7ba04-${PYTHON_TAG}-${PYTHON_TAG}-manylinux_2_28_${ARCH}.whl"; \
    else echo "Unsupported PYTHON_TAG=$PYTHON_TAG ARCH=$ARCH"; exit 1; fi && \
    \
    TRITON_ASCEND_WHEEL="triton_ascend-3.2.0.dev20260322-${PYTHON_TAG}-${PYTHON_TAG}-manylinux_2_27_${ARCH}.manylinux_2_28_${ARCH}.whl" && \
    \
    python3 -m pip install "https://vllm-ascend.obs.cn-north-4.myhuaweicloud.com/vllm-ascend/${TRITON_ASCEND_WHEEL}" && \
    python3 -m pip install "https://vllm-ascend.obs.cn-north-4.myhuaweicloud.com/vllm-ascend/${PTA_WHEEL}" && \
    python3 -m pip cache purge

# Install clang-15 (for triton-ascend)
RUN apt-get update -y && \
    apt-get -y install clang-15 && \
    update-alternatives --install /usr/bin/clang clang /usr/bin/clang-15 20 && \
    update-alternatives --install /usr/bin/clang++ clang++ /usr/bin/clang++-15 20 && \
    rm -rf /var/cache/apt/* && \
    rm -rf /var/lib/apt/lists/*

# Install modelscope (for fast download) and ray (for multinode)
RUN python3 -m pip install modelscope 'ray>=2.47.1,<=2.48.0' 'protobuf>3.20.0' && \
    python3 -m pip cache purge

RUN echo "export LD_PRELOAD=/usr/lib/$(uname -m)-linux-gnu/libjemalloc.so.2:$LD_PRELOAD" >> ~/.bashrc
RUN echo "export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/lib" >> ~/.bashrc

CMD ["/bin/bash"]