/docker-ML-tutorial

Docker tutorial for Machine Learning engineer

Primary LanguageDockerfile

Docker tutorial for ML engineers

Summarized Demystifying Docker for Data Scientists – A Docker Tutorial for Your Deep Learning Projects

You could learn

  1. how to make container from image
  2. how to make image from container
  3. how to use dockerfile to build image

Build docker container manually

  • docker run -it --name mycntkdemo microsoft/cntk:2.2-cpu-python3.5 /bin/bash
  • pip install lightgbm: install python lightgbm package in docker.
  • exit: exit from the container.
  • docker start -ia mycntkdemo: start mycntkdemo named container.
  • Ctrl-p + Ctrl-q: move container to the background.
  • docker attach mycntkdemo: move container from background to foreground.
  • docker run -it --name mycntkdemo -v /home/test:/root/test: map local dir(/home/test) to docker dir(/root/test)
  • docker run -it --name mycntkdemo -v /home/test:/root/test microsoft/cntk:2.2-cpu-python3.5 /bin/bash: map local dir to docker dir when initialize a container.
  • docker commit mycntkdemo mycntkwlgbm:version1: tag container and save as new image.

Build docker container using Dockerfile

  • Dockerfile.
cat Dockerfile
FROM microsoft/cntk:2.2-cpu-python3.5
RUN apt-get update \
	&& apt-get install -y git
COPY . /root/mylightgbmex
RUN /root/anaconda3/envs/cntk-py35/bin/pip install -r /root/mylightgbmex/requirements.txt
CMD ["echo 'hello'"]
  • Requirements for package installation.
cat requirements.txt
wheel
lightgbm
  • docker build -t mycntkwlgbmimage: creates an image named mycntkwlgbmimage by reading Dockerfile.

References