Public | Automated Build

Last pushed: 3 years ago
Short Description
Reference build environment for deeplearning4j
Full Description


  • the "devkit", a reference build environment for nd4j & deeplearning4j


  • OpenJDK 8
  • Maven 3.3
  • gcc
  • blas


The devkit provides an interactive shell to build nd4j and deeplearning4j. The shell is packaged as a docker image.

Quick Start

The following command will:

  1. Start the devkit container.
  2. Share the current directory into the container (at /workspace).
  3. Clone the dl4j project to the /workspace directory.
  4. Build the project.
docker run -it -v `pwd`:/workspace -w /workspace deeplearning4j/dl4j-devkit bash
$ git clone
$ cd deeplearning4j
$ ./

Extended Use

Install prerequisites

Mac OSX:

  1. Install boot2docker:
brew install boot2docker
  1. Install docker-compose:
pip install docker-compose

Author a docker-compose file

To simplify launching the shell, author a docker-compose file into your workspace directory. Your workspace directory should contain the nd4j, deeplearning4j, and other such repositories.

Write to ./docker-compose.yml the following snippet:

  image: deeplearning4j/dl4j-devkit:latest
    - .:/workspace
    - $HOME/.m2:/home/vagrant/.m2
  working_dir: /workspace  
    - CONTAINER_NAME=shell
  command: bash

Start the shell

$ docker-compose run --rm shell

A bash shell starts up into the workspace directory that is shared into the container at /workspace.

[vagrant@abd582a75122#shell workspace]$ pwd
[vagrant@abd582a75122#shell workspace]$ ls
Canova  deeplearning4j  dl4j-spark-ml nd4j

Build projects at will:

[vagrant@abd582a75122#shell workspace]$ cd nd4j
[vagrant@abd582a75122#shell nd4j]$ mvn install -DskipTests
[vagrant@abd582a75122#shell nd4j]$

Exit the shell when finished:

[vagrant@abd582a75122#shell workspace]$ exit
Docker Pull Command
Source Repository