Get up and running with machine learning using Docker
Shell Scala
Latest commit e31d176 Jun 4, 2015 @arshak arshak Merge pull request #23 from robinkraft/patch-1
fix "pip install" line to use RUN command

Machine Learning Docker Images

Machine learning tools are notorious for having byzantine dependencies and academic code quality. This makes them easy to install and configure across a variety of machines and operating systems.

Currently supported tools

  • Deeplearning4j
  • GraphLab
  • H2O
  • Julia
  • MLlib
  • Theano
  • Torch7
  • Vowpal Wabbit (VW)

Getting started

  • First step is to Install Docker on Mac OS X.
  • Once Boot2Docker has been installed, launch it from Spotlight
  • In the terminal window with the title "Boot2Docker for OSX" go to the jetpack directory and start the build process (be patient, the builds can take some time)

To build an individual image, provide it as an argument to the script.

./ julia (or theano, graphlab, h2o, mllib ...)

to run the docker image

./ julia (or theano, graphlab, h2o, mllib ...)


Pulling Images

If you are having trouble with the build command, try the pre-built images

docker pull startupml/julia
docker pull startupml/theano
docker pull startupml/graphlab
docker pull startupml/h2o
docker pull startupml/mllib

Starting Over

to clean up (kill, remove container and remove image)


Linux specific

The docker daemon must run as root. To be able to run the docker client as a normal user, add that user to the docker group. To do so for the current user, run: sudo gpasswd -a ${USER} docker. See docker documentation for more detail.

No space left on device

Getting “no space left on device” errors with Boot2Docker?

Boot2Docker issues

Consider using Kitematic