Awesome

Install Docker

Install Docker for Mac PC, or Linux (for GPU).

The images

Image: base

Based on continuumio/anaconda3 and:

• Automatically save notebook as python file for easier version control (see here detail)
• Add Jupyter nbextensions
• Enable favorite jupyter notebook extensions to run at start

Image: ds

Based on base and plus any additional libraries.

Image: kaggle

Based on Kaggle/docker-python and plus kaggle-cli.

Image: ktt

Keras with Tensorflow and Theano, using gpu.

Image: fastai1

For Jeremy Howard's Practical Deep Learning For Coders, Part 1, using gpu.

Image: r

Based on rocker/tidyverse and plus any addtional libraries.

Build the images

You can run the docker build command from the terminal under the project root (docker-setup) to build the images. For example:

$ sudo docker build -f dockerfiles/ktt/Dockerfile . -t ktt
$ sudo docker build -f dockerfiles/work/Dockerfile . -t work

Or run this script build.py under the project root (docker-setup) to build the images:

$ python build.py

Run a container from an image

From the terminal, run docker-setup.sh to define shell command that runs the images:

$ source docker-setup.sh

Or Add to .bash_profile the line below so that it's automatically run when your machine starts:

source [path to docker-setup.sh]

Then you can run the above images, e.g.:

• dkrun_ktt
• dkrun_fastai1
• dkrun_kaggle
• dkrun_r

For the above images, go to http://localhost:[PORT]?token=[TOKEN] to open jupyter notebook, as printed on the terminal when you run the above images.

• dkrun_r: Go to http://localhost:8787 to open rstudio, where username and password both are rstudio.