Build for Intel ResNet50 Docker Container Failing

[alex-minooka]

When running:
bash build_resnet50_contanier.sh

Container fails to build causing the resulting output:

327.9 Cloning into 'mlperf_inference'...
367.0 Already on 'master'
367.0 Your branch is up to date with 'origin/master'.
367.0 error: pathspec 'third_party/pybind/' did not match any file(s) known to git
------
Dockerfile:73
--------------------
  72 |     ARG VISION_VERSION=8e078971b8aebdeb1746fea58851e3754f103053
  73 | >>> RUN source /opt/rh/gcc-toolset-11/enable && \
  74 | >>>     cd code/retinanet/pytorch-cpu && mkdir retinanet-env && cd retinanet-env && \
  75 | >>>     git clone https://github.com/pytorch/vision && \
  76 | >>>     cd vision && git checkout ${VISION_VERSION} && python setup.py install && cd .. && \
  77 | >>>     git clone https://github.com/intel/intel-extension-for-pytorch.git ipex-cpu-dev && \
  78 | >>>     cd ipex-cpu-dev && git checkout ${IPEX_VERSION} && \
  79 | >>>     git submodule sync && git submodule update --init --recursive && \
  80 | >>>     git apply /opt/workdir/code/retinanet/pytorch-cpu/runtime_ignore_dequant_check.patch && \
  81 | >>>     python setup.py install && \
  82 | >>>     export IPEX_PATH=${PWD}/build/Release/packages/intel_extension_for_pytorch && cd .. && \
  83 | >>>     git clone https://github.com/mlcommons/inference.git mlperf_inference && \
  84 | >>>     cd mlperf_inference && git checkout ${INFERENCE} && git submodule update --init --recursive third_party/pybind/ && cp mlperf.conf ../../mlperf.conf && \
  85 | >>>     cd loadgen && mkdir build && cd build && cmake .. && make && cd .. && \
  86 | >>>     CFLAGS="-std=c++14" python setup.py install && export LOADGEN_DIR=${PWD} && cd ../.. && \
  87 | >>>     git clone https://github.com/opencv/opencv.git && cd opencv && git checkout 4.x && \
  88 | >>>     mkdir build && cd build && cmake -DBUILD_TESTS=OFF -DBUILD_PERF_TESTS=OFF -DBUILD_EXAMPLES=OFF -DBUILD_opencv_apps=OFF -DBUILD_LIST=highgui,imgcodecs,imgproc .. && make -j$(nproc) && \
  89 | >>>     export OPENCV_DIR=${PWD} && cd ../.. && \
  90 | >>>     git clone https://github.com/Tencent/rapidjson.git && cd rapidjson && git checkout e4bde977 && \
  91 | >>>     export RAPIDJSON_INCLUDE_DIR=${PWD}/include && cd .. && \
  92 | >>>     git clone https://github.com/gflags/gflags.git && cd gflags && git checkout a738fdf93 && mkdir build && cd build && \
  93 | >>>     cmake .. && make && export GFLAGS_DIR=${PWD} && cd ../.. && \
  94 | >>>     export LD_LIBRARY_PATH=${OPENCV_DIR}/lib:${CONDA_PREFIX}/lib:${LD_LIBRARY_PATH} && \
  95 | >>>     export LIBRARY_PATH=${CONDA_PREFIX}/lib:${LIBRARY_PATH} && \
  96 | >>>     rm -rf /opt/conda/lib/cmake/mkl && \
  97 | >>>     export TORCH_PATH=`python -c 'import torch;print(torch.utils.cmake_prefix_path)'` && \
  98 | >>>     cd .. && \
  99 | >>>     cmake -DCMAKE_PREFIX_PATH=${TORCH_PATH} \
 100 | >>>         -DLOADGEN_DIR=${LOADGEN_DIR} \
 101 | >>>         -DOpenCV_DIR=${OPENCV_DIR} \
 102 | >>>         -DRapidJSON_INCLUDE_DIR=${RAPIDJSON_INCLUDE_DIR} \
 103 | >>>         -Dgflags_DIR=${GFLAGS_DIR} \
 104 | >>>         -DINTEL_EXTENSION_FOR_PYTORCH_PATH=${IPEX_PATH} \
 105 | >>>         -B${PWD}/build \
 106 | >>>         -H${PWD}/src && \
 107 | >>>     cmake --build ${PWD}/build --config Release
 108 |
--------------------
ERROR: failed to solve: process "/bin/sh -c source /opt/rh/gcc-toolset-11/enable &&     cd code/retinanet/pytorch-cpu && mkdir retinanet-env && cd retinanet-env &&     git clone https://github.com/pytorch/vision &&     cd vision && git checkout ${VISION_VERSION} && python setup.py install && cd .. &&     git clone https://github.com/intel/intel-extension-for-pytorch.git ipex-cpu-dev &&     cd ipex-cpu-dev && git checkout ${IPEX_VERSION} &&     git submodule sync && git submodule update --init --recursive &&     git apply /opt/workdir/code/retinanet/pytorch-cpu/runtime_ignore_dequant_check.patch &&     python setup.py install &&     export IPEX_PATH=${PWD}/build/Release/packages/intel_extension_for_pytorch && cd .. &&     git clone https://github.com/mlcommons/inference.git mlperf_inference &&     cd mlperf_inference && git checkout ${INFERENCE} && git submodule update --init --recursive third_party/pybind/ && cp mlperf.conf ../../mlperf.conf &&     cd loadgen && mkdir build && cd build && cmake .. && make && cd .. &&     CFLAGS=\"-std=c++14\" python setup.py install && export LOADGEN_DIR=${PWD} && cd ../.. &&     git clone https://github.com/opencv/opencv.git && cd opencv && git checkout 4.x &&     mkdir build && cd build && cmake -DBUILD_TESTS=OFF -DBUILD_PERF_TESTS=OFF -DBUILD_EXAMPLES=OFF -DBUILD_opencv_apps=OFF -DBUILD_LIST=highgui,imgcodecs,imgproc .. && make -j$(nproc) &&     export OPENCV_DIR=${PWD} && cd ../.. &&     git clone https://github.com/Tencent/rapidjson.git && cd rapidjson && git checkout e4bde977 &&     export RAPIDJSON_INCLUDE_DIR=${PWD}/include && cd .. &&     git clone https://github.com/gflags/gflags.git && cd gflags && git checkout a738fdf93 && mkdir build && cd build &&     cmake .. && make && export GFLAGS_DIR=${PWD} && cd ../.. &&     export LD_LIBRARY_PATH=${OPENCV_DIR}/lib:${CONDA_PREFIX}/lib:${LD_LIBRARY_PATH} &&     export LIBRARY_PATH=${CONDA_PREFIX}/lib:${LIBRARY_PATH} &&     rm -rf /opt/conda/lib/cmake/mkl &&     export TORCH_PATH=`python -c 'import torch;print(torch.utils.cmake_prefix_path)'` &&     cd .. &&     cmake -DCMAKE_PREFIX_PATH=${TORCH_PATH}         -DLOADGEN_DIR=${LOADGEN_DIR}         -DOpenCV_DIR=${OPENCV_DIR}         -DRapidJSON_INCLUDE_DIR=${RAPIDJSON_INCLUDE_DIR}         -Dgflags_DIR=${GFLAGS_DIR}         -DINTEL_EXTENSION_FOR_PYTORCH_PATH=${IPEX_PATH}         -B${PWD}/build         -H${PWD}/src &&     cmake --build ${PWD}/build --config Release" did not complete successfully: exit code: 1

Is this a known issue and is there a work around?
Device is Sapphire Rapids

[gktejus]

@alex-minooka, this is probably because they decided to remove the pybind11 submodule (see: mlcommons/inference@0f71999) but forgot to update it in the Dockerfiles (and everywhere else). My workaround for this has been to replace

git submodule update --init --recursive third_party/pybind/ && \

in the Dockerfile to

pip install pybind11 && \

[alex-minooka]

This work around worked for me. Thank you!