ARG CUDA_VERSION=12.5.1

FROM nvcr.io/nvidia/tritonserver:24.04-py3-min

ARG BUILD_TYPE=all
ENV DEBIAN_FRONTEND=noninteractive

# 安装依赖（强制 IPv4）
RUN echo 'tzdata tzdata/Areas select Asia' | debconf-set-selections \
    && echo 'tzdata tzdata/Zones/Asia select Shanghai' | debconf-set-selections \
    && apt -o Acquire::ForceIPv4=true update -y \
    && apt -o Acquire::ForceIPv4=true install software-properties-common -y \
    && add-apt-repository ppa:deadsnakes/ppa -y \
    && apt -o Acquire::ForceIPv4=true update \
    && apt -o Acquire::ForceIPv4=true install python3.10 python3.10-dev -y \
    && update-alternatives --install /usr/bin/python3 python3 /usr/bin/python3.10 1 \
    && update-alternatives --set python3 /usr/bin/python3.10 \
    && apt -o Acquire::ForceIPv4=true install python3.10-distutils -y \
    && apt -o Acquire::ForceIPv4=true install curl git sudo libibverbs-dev -y \
    && apt -o Acquire::ForceIPv4=true install -y rdma-core infiniband-diags openssh-server perftest ibverbs-providers libibumad3 libibverbs1 libnl-3-200 libnl-route-3-200 librdmacm1 \
    && curl https://bootstrap.pypa.io/get-pip.py -o get-pip.py \
    && python3 get-pip.py \
    && python3 --version \
    && python3 -m pip --version \
    && rm -rf /var/lib/apt/lists/* \
    && apt clean

# 安装 datamodel_code_generator（用于 MiniCPM 模型）
RUN pip3 install datamodel_code_generator

WORKDIR /sgl-workspace

# 拷贝 sglang 源代码并构建包
COPY ./sglang /sgl-workspace/sglang

# 拷贝模型文件（修正方式）
#COPY ./Alibaba/QwQ-32B /root/.cradle/Alibaba/QwQ-32B
COPY ./DeepSeek/DeepSeek-R1-Distill-Llama-8B /root/.cradle/DeepSeek/DeepSeek-R1-Distill-Llama-8B

ARG CUDA_VERSION

# 安装依赖、安装 sglang、安装 transformers，并清理源码
RUN python3 -m pip install --upgrade pip setuptools wheel html5lib six \
    && if [ "$CUDA_VERSION" = "12.1.1" ]; then \
         CUINDEX=121; \
       elif [ "$CUDA_VERSION" = "12.4.1" ]; then \
         CUINDEX=124; \
       elif [ "$CUDA_VERSION" = "12.5.1" ]; then \
         CUINDEX=124; \
       elif [ "$CUDA_VERSION" = "11.8.0" ]; then \
         CUINDEX=118; \
         python3 -m pip install --no-cache-dir sgl-kernel -i https://docs.sglang.ai/whl/cu118; \
       else \
         echo "Unsupported CUDA version: $CUDA_VERSION" && exit 1; \
       fi \
    && python3 -m pip install --no-cache-dir torch --index-url https://download.pytorch.org/whl/cu${CUINDEX} \
    && python3 -m pip install --no-cache-dir psutil pyzmq pynvml \
    && cd /sgl-workspace/sglang/python \
    && python3 -m pip install --no-cache-dir '.[srt,openai]' --find-links https://flashinfer.ai/whl/cu${CUINDEX}/torch2.5/flashinfer-python \
    && cd / && rm -rf /sgl-workspace/sglang \
    && python3 -m pip install --no-cache-dir transformers==4.48.3 \
    && python3 -c "import sglang; print('✅ sglang module installed successfully')"

# 暴露端口
EXPOSE 30000


# 启动命令（QwQ 模型为例）
CMD ["python3", "-m", "sglang.launch_server", \
     "--host","0.0.0.0", \
     "--port","30000", \
     "--model-path", "/root/.cradle/DeepSeek/DeepSeek-R1-Distill-Llama-8B/", \
     "--tp", "1", \
     "--api-key", "token-abc123"]

# 设置回默认交互模式
ENV DEBIAN_FRONTEND=interactive