Dockerfile

# 使用官方光源基础镜像
FROM image.sourcefind.cn:5000/dcu/admin/base/custom:vllm0.8.5-ubuntu22.04-dtk25.04-rc7-das1.5-py3.10-20250521-fixpy-rocblas0521-beta2

# 创建目标目录
RUN mkdir -p /workspace/test/results


# 设置工作目录/workspace/test/results方便挂载结果
WORKDIR /workspace/test/results

# 将主机上的 test.sh 复制到容器中的 /workspace/test
COPY ./run.sh /workspace/test
COPY ./benchmark_throughput_0.7.2.py /workspace/test
COPY ./models-to-test.cfg /workspace/test 

# 确保 test.sh 有可执行权限
RUN chmod +x /workspace/test/run.sh

# 设置容器启动时运行的命令
# 使用单个 CMD 执行所有命令
CMD bash -c "\
  rocm-bandwidth-test > rocm-bandwidth-test.log && \
  hy-smi > hy-smi.log && \
  hy-smi -c > hy-smi-c.log && \
  pip list > pip-list.log && \
  lscpu > lscpu.log && \
  git clone https://www.ghproxy.cn/github.com/ROCm/rccl-tests.git --depth 1 -b master && \
  cd /workspace/test/results/rccl-tests && \
  source /opt/dtk/env.sh && \
  make MPI=1 MPI_HOME=/opt/mpi ROCM_HOME=/opt/dtk/ NCCL_HOME=/opt/dtk/rccl/ CUSTOM_RCCL_LIB=/opt/dtk/rccl/lib/librccl.so -j32 && \
  ./build/all_reduce_perf -b 8 -e 1G -f 2 -g 8 > /workspace/test/results/all_reduce_perf_8.log && \
  ./build/all_reduce_perf -b 4 -e 1G -f 2 -g 4 > /workspace/test/results/all_reduce_perf_4.log && \
  cd /workspace/test/results && \
  git clone http://developer.sourcefind.cn/codes/OpenDAS/dcu_env_check.git && \
  apt-get update && \
  apt-get install -y iproute2 dmidecode ipmitool && \
  cd /workspace/test/results/dcu_env_check  && \
  bash system_check.sh > /workspace/test/results/system_check.log  && \
  cd /workspace/test/results && \
  bash /workspace/test/run.sh > test.log 2>&1"