DeepLearning 모형 변환 도구가 모두 쌓여 커다란 Docker 환경 구축

36341 단어 TensorFlow PyTorch onnx openvino tensorrt tech

1. Introduction

귀찮아.이 세상의 모든 환경 구축은 매우 번거롭다.의존관계 파괴제는 고통스럽다.따라서 Giithub Actions에서 모델 변환에 관련된 환경을 모두 추출docker build하여 거대한 전체 DL 모델 변환 환경을 구축했다.부족한 주변 공구는 각각 추가 설치.컨테이너 내에서 GUI, iGPU/dGPU 및 호스트에 연결된 USB 장치에 액세스할 수 있으므로 실행 환경으로도 직접 사용할 수 있습니다.하지만 어쨌든 Image는 크고 안전성 GABA입니다.GiitHub Actions의 컨테이너 건설 용량 제한에 걸려 넘어지는 술책을 피하기 위해서다.

2. Environment

Python 3.6+

TensorFlow v2.6.0+

PyTorch v1.10.0+

TorchVision

TorchAudio

OpenVINO 2021.4.582+

TensorRT 8.2+

pycuda 2021.1

tensorflowjs

coremltools

onnx

onnxruntime

onnx_graphsurgeon

onnx-simplifier

onnxconverter-common

onnx-tensorrt

onnx2json

json2onnx

tf2onnx

torch2trt

onnx-tf

tensorflow-datasets

tf_slim

edgetpu_compiler

tflite2tensorflow

openvino2tensorflow

gdown

pandas

matplotlib

Intel-Media-SDK

Intel iHD GPU (iGPU) support

OpenCL

Docker

CUDA 11.4

3. Procedure

3-1. Doke rile 만들기

Dockerfile

FROM nvidia/cuda:11.4.2-cudnn8-devel-ubuntu20.04

ENV DEBIAN_FRONTEND=noninteractive
ARG OSVER=ubuntu2004
ARG TENSORFLOWVER=2.6.0
ARG CPVER=cp38
ARG OPENVINOVER=2021.4.582
ARG OPENVINOROOTDIR=/opt/intel/openvino_2021
ARG TENSORRTVER=cuda11.4-trt8.2.0.6-ea-20210922
ARG APPVER
ARG WKDIR=/home/user

# dash -> bash
RUN echo "dash dash/sh boolean false" | debconf-set-selections \
    && dpkg-reconfigure -p low dash
COPY bashrc ${WKDIR}/.bashrc
WORKDIR ${WKDIR}

# Install dependencies (1)
RUN apt-get update && apt-get install -y \
        automake autoconf libpng-dev nano python3-pip \
        curl zip unzip libtool swig zlib1g-dev pkg-config \
        python3-mock libpython3-dev libpython3-all-dev \
        g++ gcc cmake make pciutils cpio gosu wget \
        libgtk-3-dev libxtst-dev sudo apt-transport-https \
        build-essential gnupg git xz-utils vim \
        libva-drm2 libva-x11-2 vainfo libva-wayland2 libva-glx2 \
        libva-dev libdrm-dev xorg xorg-dev protobuf-compiler \
        openbox libx11-dev libgl1-mesa-glx libgl1-mesa-dev \
        libtbb2 libtbb-dev libopenblas-dev libopenmpi-dev \
    && sed -i 's/# set linenumbers/set linenumbers/g' /etc/nanorc \
    && apt clean \
    && rm -rf /var/lib/apt/lists/*

# python3 -> python
RUN ln -s /usr/bin/python3 /usr/bin/python

# Install dependencies (2)
RUN pip3 install --upgrade pip \
    && pip install --upgrade numpy==1.19.5 \
    && pip install --upgrade tensorflowjs \
    && pip install --upgrade coremltools \
    && pip install --upgrade onnx \
    && pip install --upgrade onnxruntime \
    && pip install --upgrade onnx-simplifier \
    && pip install --upgrade onnxconverter-common \
    && pip install --upgrade tf2onnx \
    && pip install --upgrade onnx-tf \
    && pip install --upgrade tensorflow-datasets \
    && pip install --upgrade openvino2tensorflow \
    && pip install --upgrade tflite2tensorflow \
    && pip install --upgrade gdown \
    && pip install --upgrade PyYAML \
    && pip install --upgrade matplotlib \
    && pip install --upgrade tf_slim \
    && pip install --upgrade pandas \
    && pip install --upgrade numexpr \
    && pip install --upgrade onnx2json \
    && pip install --upgrade json2onnx \
    && python3 -m pip install onnx_graphsurgeon \
        --index-url https://pypi.ngc.nvidia.com \
    && pip install torch==1.10.0+cu113 torchvision==0.11.1+cu113 torchaudio==0.10.0+cu113 \
        -f https://download.pytorch.org/whl/cu113/torch_stable.html \
    && pip install pycuda==2021.1 \
    && ldconfig \
    && pip cache purge \
    && apt clean \
    && rm -rf /var/lib/apt/lists/*

# Install sclblonnx non-version check custom .ver
RUN wget https://github.com/PINTO0309/openvino2tensorflow/releases/download/${APPVER}/sclblonnx-0.1.9_nvc-py3-none-any.whl \
    && pip3 install sclblonnx-0.1.9_nvc-py3-none-any.whl \
    && rm sclblonnx-0.1.9_nvc-py3-none-any.whl \
    && apt clean \
    && rm -rf /var/lib/apt/lists/*

# Install custom tflite_runtime, flatc, edgetpu-compiler
RUN wget https://github.com/PINTO0309/openvino2tensorflow/releases/download/${APPVER}/tflite_runtime-${TENSORFLOWVER}-${CPVER}-none-linux_x86_64.whl \
    && chmod +x tflite_runtime-${TENSORFLOWVER}-${CPVER}-none-linux_x86_64.whl \
    && pip3 install --force-reinstall tflite_runtime-${TENSORFLOWVER}-${CPVER}-none-linux_x86_64.whl \
    && rm tflite_runtime-${TENSORFLOWVER}-${CPVER}-none-linux_x86_64.whl \
    && wget https://github.com/PINTO0309/openvino2tensorflow/releases/download/${APPVER}/flatc.tar.gz \
    && tar -zxvf flatc.tar.gz \
    && chmod +x flatc \
    && rm flatc.tar.gz \
    && wget https://github.com/PINTO0309/tflite2tensorflow/raw/main/schema/schema.fbs \
    && curl https://packages.cloud.google.com/apt/doc/apt-key.gpg | apt-key add - \
    && echo "deb https://packages.cloud.google.com/apt coral-edgetpu-stable main" | tee /etc/apt/sources.list.d/coral-edgetpu.list \
    && apt-get update \
    && apt-get install edgetpu-compiler \
    && pip cache purge \
    && apt clean \
    && rm -rf /var/lib/apt/lists/*

# Install OpenVINO
RUN wget https://github.com/PINTO0309/openvino2tensorflow/releases/download/${APPVER}/l_openvino_toolkit_p_${OPENVINOVER}.tgz \
    && tar xf l_openvino_toolkit_p_${OPENVINOVER}.tgz \
    && rm l_openvino_toolkit_p_${OPENVINOVER}.tgz \
    && l_openvino_toolkit_p_${OPENVINOVER}/install_openvino_dependencies.sh -y \
    && sed -i 's/decline/accept/g' l_openvino_toolkit_p_${OPENVINOVER}/silent.cfg \
    && l_openvino_toolkit_p_${OPENVINOVER}/install.sh --silent l_openvino_toolkit_p_${OPENVINOVER}/silent.cfg \
    && source ${OPENVINOROOTDIR}/bin/setupvars.sh \
    && ${INTEL_OPENVINO_DIR}/install_dependencies/install_openvino_dependencies.sh \
    && sed -i 's/sudo -E //g' ${INTEL_OPENVINO_DIR}/deployment_tools/model_optimizer/install_prerequisites/install_prerequisites.sh \
    && sed -i 's/tensorflow/#tensorflow/g' ${INTEL_OPENVINO_DIR}/deployment_tools/model_optimizer/requirements.txt \
    && sed -i 's/numpy/#numpy/g' ${INTEL_OPENVINO_DIR}/deployment_tools/model_optimizer/requirements.txt \
    && sed -i 's/onnx/#onnx/g' ${INTEL_OPENVINO_DIR}/deployment_tools/model_optimizer/requirements.txt \
    && ${INTEL_OPENVINO_DIR}/deployment_tools/model_optimizer/install_prerequisites/install_prerequisites.sh \
    && rm -rf l_openvino_toolkit_p_${OPENVINOVER} \
    && echo "source ${OPENVINOROOTDIR}/bin/setupvars.sh" >> .bashrc \
    && echo "${OPENVINOROOTDIR}/deployment_tools/ngraph/lib/" >> /etc/ld.so.conf \
    && echo "${OPENVINOROOTDIR}/deployment_tools/inference_engine/lib/intel64/" >> /etc/ld.so.conf \
    && pip cache purge \
    && apt clean \
    && rm -rf /var/lib/apt/lists/*

# Install TensorRT additional package
RUN wget https://github.com/PINTO0309/openvino2tensorflow/releases/download/${APPVER}/nv-tensorrt-repo-${OSVER}-${TENSORRTVER}_1-1_amd64.deb \
    && dpkg -i nv-tensorrt-repo-${OSVER}-${TENSORRTVER}_1-1_amd64.deb \
    && apt-key add /var/nv-tensorrt-repo-${OSVER}-${TENSORRTVER}/7fa2af80.pub \
    && apt-get update \
    && apt-get install -y \
        tensorrt uff-converter-tf graphsurgeon-tf \
        python3-libnvinfer-dev onnx-graphsurgeon \
    && rm nv-tensorrt-repo-${OSVER}-${TENSORRTVER}_1-1_amd64.deb \
    && cd /usr/src/tensorrt/samples/trtexec \
    && make \
    && apt clean \
    && rm -rf /var/lib/apt/lists/*

# Install Custom TensorFlow (MediaPipe Custom OP, FlexDelegate, XNNPACK enabled)
RUN wget https://github.com/PINTO0309/openvino2tensorflow/releases/download/${APPVER}/tensorflow-${TENSORFLOWVER}-${CPVER}-none-linux_x86_64.whl \
    && pip3 install --force-reinstall tensorflow-${TENSORFLOWVER}-${CPVER}-none-linux_x86_64.whl \
    && rm tensorflow-${TENSORFLOWVER}-${CPVER}-none-linux_x86_64.whl \
    && pip cache purge \
    && apt clean \
    && rm -rf /var/lib/apt/lists/*

# Install onnx-tensorrt
RUN git clone --recursive https://github.com/onnx/onnx-tensorrt \
    && cd onnx-tensorrt \
    && git checkout 1f041ce6d7b30e9bce0aacb2243309edffc8fb3c \
    && mkdir build && cd build \
    && cmake .. -DTENSORRT_ROOT=/usr/src/tensorrt \
    && make -j$(nproc) && make install

# Install torch2trt
RUN git clone https://github.com/NVIDIA-AI-IOT/torch2trt \
    && cd torch2trt \
    && git checkout 0400b38123d01cc845364870bdf0a0044ea2b3b2 \
    # https://github.com/NVIDIA-AI-IOT/torch2trt/issues/619
    && wget https://github.com/NVIDIA-AI-IOT/torch2trt/commit/8b9fb46ddbe99c2ddf3f1ed148c97435cbeb8fd3.patch \
    && git apply 8b9fb46ddbe99c2ddf3f1ed148c97435cbeb8fd3.patch \
    && python3 setup.py install

# Download the ultra-small sample data set for INT8 calibration
RUN mkdir sample_npy \
    && wget -O sample_npy/calibration_data_img_sample.npy https://github.com/PINTO0309/openvino2tensorflow/releases/download/${APPVER}/calibration_data_img_sample.npy

# Clear caches
RUN apt clean \
    && rm -rf /var/lib/apt/lists/*

# Create a user who can sudo in the Docker container
ENV USERNAME=user
RUN echo "root:root" | chpasswd \
    && adduser --disabled-password --gecos "" "${USERNAME}" \
    && echo "${USERNAME}:${USERNAME}" | chpasswd \
    && echo "%${USERNAME}    ALL=(ALL)   NOPASSWD:    ALL" >> /etc/sudoers.d/${USERNAME} \
    && chmod 0440 /etc/sudoers.d/${USERNAME}
USER ${USERNAME}
RUN sudo chown ${USERNAME}:${USERNAME} ${WKDIR}\
    && sudo chmod 777 ${WKDIR}/.bashrc

# OpenCL settings - https://github.com/intel/compute-runtime/releases
RUN cd ${OPENVINOROOTDIR}/install_dependencies/ \
    && yes | sudo -E ./install_NEO_OCL_driver.sh \
    && cd ${WKDIR} \
    && wget https://github.com/intel/compute-runtime/releases/download/21.29.20389/intel-gmmlib_21.2.1_amd64.deb \
    && wget https://github.com/intel/intel-graphics-compiler/releases/download/igc-1.0.7862/intel-igc-core_1.0.7862_amd64.deb \
    && wget https://github.com/intel/intel-graphics-compiler/releases/download/igc-1.0.7862/intel-igc-opencl_1.0.7862_amd64.deb \
    && wget https://github.com/intel/compute-runtime/releases/download/21.29.20389/intel-opencl_21.29.20389_amd64.deb \
    && wget https://github.com/intel/compute-runtime/releases/download/21.29.20389/intel-ocloc_21.29.20389_amd64.deb \
    && wget https://github.com/intel/compute-runtime/releases/download/21.29.20389/intel-level-zero-gpu_1.1.20389_amd64.deb \
    && sudo dpkg -i *.deb \
    && rm *.deb \
    && sudo apt clean \
    && sudo rm -rf /var/lib/apt/lists/*

# Final processing of onnx-tensorrt install
RUN echo "export PATH=${PATH}:/usr/src/tensorrt/bin:/onnx-tensorrt/build" >> ${HOME}/.bashrc \
    && echo "cd ${HOME}/onnx-tensorrt" >> ${HOME}/.bashrc \
    && echo "sudo python3 setup.py install" >> ${HOME}/.bashrc \
    && echo "cd ${WKDIR}" >> ${HOME}/.bashrc \
    && echo "cd ${HOME}/workdir" >> ${HOME}/.bashrc

3-2. GitHub workflow YAML 제작

GiitHub에서 사용하는 패키지를 게시할 때 PyPI의 패키지와 전체 장착 컨테이너를 GiitHub Actions에 자동으로 구축하여 실행합니다.GiitHub Action에 등록하여 사용하십시오.GiitHub Actions의 백엔드에 할당된 저장소의 크기가 고정되어 있기 때문에 구축된 이미지의 저장 용량을 일시적으로 회피할 수 없는 경우No space left on device at System.IO.FileStream.WriteNative에 Abort를 실행합니다.따라서 Check space before cleanup의 부분에서 용량을 절약하기 위해 기본적으로 도입된 스팸 이미지와 불필요한 모듈을 제거한다.이것을 실행하지 않으면 용량이 부족하면 Abort를 진행합니다.docker-image.yml

name: Docker Image CI
on:
  release:
    types:
      - published
jobs:
  build:
    runs-on: ubuntu-latest
    steps:
      - name: Git checkout
        uses: actions/checkout@v2
      - name: downcase REPO
        run: echo "REPO=${GITHUB_REPOSITORY,,}" >> ${GITHUB_ENV}
      - name: Get Tag
        run: echo "TAG=${GITHUB_REF##*/}" >> ${GITHUB_ENV}

      - name: Check space before cleanup
        run: df -h
      - name: Clean space as per https://github.com/actions/virtual-environments/issues/709
        run: |
          sudo rm -rf "/usr/local/share/boost"
          sudo rm -rf "$AGENT_TOOLSDIRECTORY"
          docker rmi $(docker image ls -aq)
          df -h
      - name: Setup Python
        uses: actions/setup-python@v2
        with:
          python-version: '3.x'
      - name: Install dependencies
        run: |
          python -m pip install --upgrade pip
          pip install setuptools wheel twine
      - name: Package build and publish
        env:
          TWINE_USERNAME: ${{ secrets.PYPI_USERNAME }}
          TWINE_PASSWORD: ${{ secrets.PYPI_PASSWORD }}
        run: |
          python setup.py sdist bdist_wheel
          twine upload --repository pypi dist/*
      - name: Enable buildx
        uses: docker/setup-buildx-action@v1
      - name: Login
        uses: docker/login-action@v1
        with:
          registry: ghcr.io
          username: ${{ github.actor }}
          password: ${{ secrets.GITHUB_TOKEN }}
      - name: Docker build/push
        uses: docker/build-push-action@v2
        with:
          context: .
          push: true
          tags: ghcr.io/${{ env.REPO }}:latest
          build-args: APPVER=${{ env.TAG }}

4. 구축된 컨테이너 사용 방법

$ docker pull ghcr.io/pinto0309/openvino2tensorflow:latest

# If you don't need to access the GUI of 
# the HostPC and the USB camera.
$ docker run -it --rm \
  -v `pwd`:/home/user/workdir \
  ghcr.io/pinto0309/openvino2tensorflow:latest

# If conversion to TF-TRT is not required.
# And if you need to access the HostPC GUI and USB camera.
$ xhost +local: && \
  docker run -it --rm \
  -v `pwd`:/home/user/workdir \
  -v /tmp/.X11-unix/:/tmp/.X11-unix:rw \
  --device /dev/video0:/dev/video0:mwr \
  --net=host \
  -e XDG_RUNTIME_DIR=$XDG_RUNTIME_DIR \
  -e DISPLAY=$DISPLAY \
  --privileged \
  ghcr.io/pinto0309/openvino2tensorflow:latest

# If you need to convert to TF-TRT.
# And if you need to access the HostPC GUI and USB camera.
$ xhost +local: && \
  docker run --gpus all -it --rm \
  -v `pwd`:/home/user/workdir \
  -v /tmp/.X11-unix/:/tmp/.X11-unix:rw \
  --device /dev/video0:/dev/video0:mwr \
  --net=host \
  -e XDG_RUNTIME_DIR=$XDG_RUNTIME_DIR \
  -e DISPLAY=$DISPLAY \
  --privileged \
  ghcr.io/pinto0309/openvino2tensorflow:latest

# If you are using iGPU (OpenCL).
# And if you need to access the HostPC GUI and USB camera.
$ xhost +local: && \
  docker run -it --rm \
  -v `pwd`:/home/user/workdir \
  -v /tmp/.X11-unix/:/tmp/.X11-unix:rw \
  --device /dev/video0:/dev/video0:mwr \
  --net=host \
  -e LIBVA_DRIVER_NAME=iHD \
  -e XDG_RUNTIME_DIR=$XDG_RUNTIME_DIR \
  -e DISPLAY=$DISPLAY \
  --privileged \
  ghcr.io/pinto0309/openvino2tensorflow:latest

Reference

이 문제에 관하여(DeepLearning 모형 변환 도구가 모두 쌓여 커다란 Docker 환경 구축), 우리는 이곳에서 더 많은 자료를 발견하고 링크를 클릭하여 보았다 https://zenn.dev/pinto0309/articles/7a0b8501a5b200

텍스트를 자유롭게 공유하거나 복사할 수 있습니다.하지만 이 문서의 URL은 참조 URL로 남겨 두십시오.

우수한 개발자 콘텐츠 발견에 전념 (Collection and Share based on the CC Protocol.)

자바 프로그래밍 입문-연습문제 #2

React Router v6 - 새로운 기능은 무엇입니까?

좋은 웹페이지 즐겨찾기

개발자 우수 사이트 수집

개발자가 알아야 할 필수 사이트 100선 추천 우리는 당신을 위해 100개의 자주 사용하는 개발자 학습 사이트를 정리했습니다