/deeplearning4j

Deep Learning for Java, Scala & Clojure on Hadoop & Spark With GPUs - From Skymind

Primary LanguageJavaOtherNOASSERTION

Deeplearning4J: Neural Net Platform

Join the chat at https://gitter.im/deeplearning4j/deeplearning4j Maven Central Javadoc

Deeplearning4J is an Apache 2.0-licensed, open-source, distributed neural net library written in Java and Scala.

Deeplearning4J integrates with Hadoop and Spark and runs on several backends that enable use of CPUs and GPUs. The aim is to create a plug-and-play solution that is more convention than configuration, and which allows for fast prototyping.

The most recent stable release in Maven Central is 0.5.0, and the current master is 0.5.1-SNAPSHOT. For more on working with snapshots, see this page.


Using Deeplearning4j

To get started using Deeplearning4j, please go to our Quickstart. You'll need to be familiar with a Java automated build tool such as Maven and an IDE such as IntelliJ.

Main Features

  • Versatile n-dimensional array class
  • GPU integration(Supports devices starting from Kepler,cc3.0. You can check your device's compute compatibility here.)
  • Scalable on Hadoop, Spark and Akka + AWS et al

Modules

  • cli = command line interface for deeplearning4j
  • core = core neural net structures and supporting components such as datasets, iterators, clustering algorithms, optimization methods, evaluation tools and plots.
  • scaleout = integrations
    • aws = loading data to and from aws resources EC2 and S3
    • nlp = natural language processing components including vectorizers, models, sample datasets and renderers
    • akka = setup concurrent and distributed applications on the JVM
    • api = core components like workers and multi-threading
    • zookeeper = maintain configuration for distributed systems
    • hadoop-yarn = common map-reduce distributed system
    • spark = integration with spark
      • dl4j-spark = spark 1.2-compatible
      • dl4j-spark-ml = spark 1.4-compatible, based on ML pipeline
  • ui = provides visual interfaces with models like nearest neighbors
  • test-resources = datasets and supporting components for tests

Documentation

Documentation is available at deeplearning4j.org and JavaDocs.

Support

We are not supporting Stackoverflow right now. Github issues should be limited to bug reports. Please join the community on Gitter, where we field questions about how to install the software and work with neural nets. For support from Skymind, please see our contact page.

Installation

To install Deeplearning4J, there are a couple approaches (briefly described on our Quickstart and below). More information can be found on the ND4J website and here.

Use Maven Central Repository

Search for deeplearning4j to get a list of jars you can use.

Add the dependency information into your pom.xml. We highly recommend downloading via Maven unless you plan to help us develop DL4J.

Yum Install / Load RPM (Fedora or CentOS)

Create a yum repo and run yum install to load the Red Hat Package Management (RPM) files. First create the repo file to setup the configuration locally.

$ sudo vi /etc/yum.repos.d/dl4j.repo 

Add the following to the dl4j.repo file:

[dl4j.repo]

name=dl4j-repo
baseurl=http://ec2-52-5-255-24.compute-1.amazonaws.com/repo/RPMS
enabled=1
gpgcheck=0

Then run the following command on the dl4j repo packages to install them on your machine:

$ sudo yum install [package name] -y
$ sudo yum install DL4J-Distro -y 

Note, be sure to install the ND4J modules you need first, especially the backend and then install DataVec and DL4J.


Contribute

  1. Check for open issues or open a fresh one to start a discussion around a feature idea or a bug.
  2. If you feel uncomfortable or uncertain about an issue or your changes, don't hesitate to contact us on Gitter using the link above.
  3. Fork the repository on GitHub to start making your changes to the master branch (or branch off of it).
  4. Write a test which shows that the bug was fixed or that the feature works as expected.
  5. Send a pull request and bug us on Gitter until it gets merged and published. :)