dockerfile晋级操作

╰+哭是因爲堅強的太久メ 2023-02-19 08:29 14阅读 0赞

dockerfile常规操作中介绍了安装源更换、软件更新、模块安装等命令，参考文件移步[这里][Link 1]。常规操作适用于容器内单个环境的情况，对于多个模型不同的运行环境时，一个个安装会很麻烦，遇到模块版本不同的情况就更尴尬了。今天介绍使用虚拟环境解决容器内多环境同时存在的情况。

*  dockerfile文件

FROM nvidia/cuda:10.0-cudnn7-devel-ubuntu16.04
    
    
    RUN sed -i "s#archive.ubuntu.com#mirrors.aliyun.com#g" /etc/apt/sources.list \
        && sed -i "s#security.ubuntu.com#mirrors.aliyun.com#g" /etc/apt/sources.list #修改下载源
     
    RUN apt-get upgrade -y \
       && apt-get update 
    
    ENV PATH="/root/anaconda3/bin:${PATH}"  #conda运行环境加入系统环境
    ARG PATH="/root/anaconda3/bin:${PATH}"
    
    RUN apt-get install unzip
     
    COPY Anaconda3-2020.02-Linux-x86_64.sh ./ 
    COPY yolov5.zip ./ 
    COPY run.sh ./
    COPY requests_py2.7.txt ./  #默认环境的依赖模块
    COPY env_py36.yml ./        #虚拟环境1
    COPY env_yolov5-py37.yml ./  #虚拟环境2
    
    RUN unzip yolov5.zip \
        && bash Anaconda3-2020.02-Linux-x86_64.sh -b 
    
    RUN conda env create -f env_py36.yml \     #创建虚拟环境
        && conda env create -f env_yolov5-py37.yml
    
    RUN pip2 install -U -r requests_py2.7.txt \  #安装默认环境的依赖模块
        && rm yolov5.zip \
        && rm Anaconda3-2020.02-Linux-x86_64.sh \
        && rm requests_py2.7.txt \
        && rm *.yml
    	
    RUN apt-get clean all   #删除安装包
        && rm -rf /var/lib/apt/lists/* #安装包状态信息的存储区
    	
    EXPOSE 7001  #开放镜像端口
    ENTRYPOINT ["/bin/bash", "run.sh"] #启动镜像后执行的文件,强烈建议使用bash作为运行环境

*  启动脚本

cd carinner  #使用默认环境
    nohup python2 service_carinner.py --port 7001 > /dev/null 2>&1 &
    source activate py3.6  #启动环境py3.6
    cd ../openposeKps
    nohup python service_kps.py --port 7002 > /dev/null 2>&1 &
    cd ../scene
    nohup python service_scene.py --port 7004 > /dev/null 2>&1 &
    cd ../imgcrop
    nohup python service_ipl.py --port 8001 > /dev/null 2>&1 &
    conda deactivate     #退出环境py3.6
    source activate yolov5_py3.7  #启动环境yolov5_py3.7
    cd ../yolov5
    python3 service_yolo.py --port 7003
    conda deactivate    #退出环境yolov5_py3.7

*  环境配置文件

1.  生成环境配置文件命令

root@ffeffbc15c98:/workspace# conda activate py3.6
    (py3.6) root@ffeffbc15c98:/workspace# conda env export >py3.6.yml

2. 配置环境内容

name: py3.6
    channels:
      - defaults
    dependencies:
      - _libgcc_mutex=0.1=main
      - ca-certificates=2020.1.1=0
      - certifi=2020.4.5.2=py36_0
      …………
      - wheel=0.34.2=py36_0
      - xz=5.2.5=h7b6447c_0
      - zlib=1.2.11=h7b6447c_3
      - pip:
        - cffi==1.14.0
        - numpy==1.18.5
        …………
        - torchvision==0.2.1
        - tornado==6.0.4
    prefix: /root/anaconda3/envs/py3.6

### 参考文献： ###

1.  [How to install packages with miniconda in Dockerfile?][How to install packages with miniconda in Dockerfile]
2.  [Activating a Conda environment in your Dockerfile][]
3.  [Dockerfile reference][]

[Link 1]: https://blog.csdn.net/hzhj2007/article/details/81514710
[How to install packages with miniconda in Dockerfile]: https://stackoverflow.com/questions/58269375/how-to-install-packages-with-miniconda-in-dockerfile
[Activating a Conda environment in your Dockerfile]: https://pythonspeed.com/articles/activate-conda-dockerfile/
[Dockerfile reference]: https://docs.docker.com/engine/reference/builder/#arg