AmirHossein Sojoodi

AmirHossein Sojoodi

PhD student in Computer Engineering at Queen's University

Building DL4J

Published:
3 minute read

Environment Setup

This instructions has been tested on Ubuntu 16.04

  1. Get and install Ubuntu
  2. Install Nvidia driver (if you have Nvidia cuda-capable graphic card)

The easy option is this:

sudo add-apt-repository ppa:xorg-edgers/ppa
sudo apt-get update
sudo apt-get install nvidia-_YourPrefferedVersion_
  1. Install Java Development Kit (8)

You can install oracle or openjdk. Oracle has been tested, too.

  1. Install Cuda Sdk

You can obtain its docs here

  1. Install Nexus Repository Manager (preferred to speedup your build phase)

You can view its documentation here Now your environment is good and ready, Let’s head to the Building Steps

Requirements

  • git
  • cmake (3.6 or higher)
  • OpenMP
  • gcc (4.9 or higher)
  • maven (3.3 or higher)

Check the cmake and maven version. They have to be at lease the mentioned version. If default cmake in Ubuntu was not higher than 3.6, you have to install it manually:

sudo apt-get remove cmake
sudo apt-get purge --auto-remove cmake

Then download cmake from here Extract it by tar xzf nameOfThefile.tar.gz and cd into it. like:

cd cmake-3.13.0-rc1/
./bootstrap
make 
sudo make install

check the version again.

Install other Requirements

sudo apt-get purge maven maven2 maven3
sudo add-apt-repository ppa:natecarlson/maven3
sudo apt-get update
sudo apt-get install maven build-essential libgomp1 git

Install Prerequisite Architecture

OpenBLAS

sudo apt-get install libopenblas-dev

You will also need to ensure that /opt/OpenBLAS/lib (or any other home directory for OpenBLAS) is on your PATH. In order to get OpenBLAS to work with Apache Spark, you will also need to make sure that libraries liblapack.so.3 and libblas.so.3 are present in /usr/lib/openblas-base. If they don’t exist you can do the following:

sudo cp libopenblas.so liblapack.so.3
sudo cp libopenblas.so libblas.so.3

ATLAS

sudo apt-get install libatlas-base-dev libatlas-dev

Installing the DL4J Stack

Get the source code

git clone https://github.com/deeplearning4j/deeplearning4j.git

It will take some time according to your network speed. Then you have to checkout to a stable version of the code. Then create a simple branch.

git checkout latest_release
git checkout -b newBranch

Before running the DL4J stack build script, you must ensure certain environment variables are defined before running your build.

You will need to know the exact path of the directory where you are running the DL4J build script (you are encouraged to use a clean empty directory). Otherwise, your build will fail. Once you determine this path, add /libnd4j to the end of that path and export it to your local environment. This will look like:

export LIBND4J_HOME="/path/to/deeplearning4j/libnd4j"
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$LIBND4J_HOME/include

You can add those lines at the end of your bashrc file.

Build Commands

When you are in the base folder run these commands. assume that:

  • cuda version is 8.0
  • scala version is 2.11
  • spark version is 2
./change-cuda-versions.sh 8.0
./change-scala-versions.sh 2.11
./change-spark-versions.sh 2

if you don’t want to use Cuda, build with this command:

mvn clean install -Dmaven.test.skip -DskipTests -Dmaven.javadoc.skip=true -pl '!./nd4j/nd4j-backends/nd4j-backend-impls/nd4j-cuda,!./nd4j/nd4j-backends/nd4j-backend-impls/nd4j-cuda-platform,!./nd4j/nd4j-backends/nd4j-tests,!./deeplearning4j/deeplearning4j-cuda/'

And run this command if you have intend to use Cuda:

replace libnd4j.cuda=x.x with your cuda version, like -Dlibnd4j.cuda=8.0 replace libnd4j.compute=xx with your gpu capabality, like -Dlibnd4j.compute=30 for GTX 680. Checkout your device compute capability here

mvn clean install -Dlibnd4j.cuda=x.x -Dlibnd4j.compute=xx -Dmaven.test.skip -DskipTests -Dmaven.javadoc.skip=true -pl '!./nd4j/nd4j-backends/nd4j-tests'

Sometimes you have to use -U for maven to force update the packages through Nexus.

It’s better to add this maven repository(proxy) to Nexus

Resources

You can check DL4J tutorials and docs pages for more. here and there

You May Also Enjoy

Utilize PTX Just-In-Time (JIT) Compilation in CUDA

4 minute read

Published:

In this post I’ve written about how to utilize PTX Just-In-Time (JIT) compilation in CUDA. PTX is a low-level assembly-like language that is used to represent the GPU code. The PTX code is then compiled to the machine code by the NVIDIA driver at runtime. This process is called Just-In-Time (JIT) compilation. But before I write about how to use PTX JIT compilation, I’ll provide some background on why you might want to use it.

Collection of my dotfiles (public version)

1 minute read

Published:

This is a collection of my dotfiles that I use on my Linux system. I have written about them in various posts, but I thought I can put them all in one place. You can checkout the github repository here, too.

Tmux Tips and Tricks

10 minute read

Published:

(If you are already familiar with tmux, and you are looking for a simple .tmux.conf, go to the end of this post.)

Setup LAMMPS

4 minute read

Published:

Large-scale Atomic/Molecular Massively Parallel Simulator (LAMMPS) is a classical molecular dynamics code that can be used to model atoms or, more generally, as a parallel particle simulator at various scales. The complete documentation of LAMMPS can be found here. In this post, I will provide a guide on how to setup LAMMPS on a Linux machine. My setup is on a cluster with NVIDIA GPUs, UCX, and OpenMPI. Also, we have a built-in module system to load the necessary modules.