Compiling NuPIC on Ubuntu 14

vitaly-krugl edited this page Aug 20, 2016 · 10 revisions
Clone this wiki locally

Quick Install Script

If you don't care what's actually happening in the installation described below, you can just run this script (with sudo rights!) on a freshly installed Ubuntu 14 LTS, and it should do everything you need. However, I suggest you install NuPIC by following the step-by-step directions below so you'll know exactly what the installation process does in your environment.


Install some basic depedencies

sudo apt-get update -y
sudo apt-get install git g++ cmake python-dev -y

Clone nupic.core and nupic

git clone
git clone

Set some environment vars

These point to our checkout directories. We'll need them later for running tests and scripts. (You'll probably want this in your .bashrc or .bash_profile.)

export NUPIC=$HOME/nupic
export NUPIC_CORE=$HOME/nupic.core

Get the latest pip

Notice we are not using apt-get because it provides a very old version of pip that is hard to update properly.

curl | sudo python

NuPIC Core

Install nupic.core python dependencies

This takes awhile because numpy!

pip install -r bindings/py/requirements.txt --user

Install pycapnp

This will install the python Capnproto bindings and the C++ Capnproto program.

pip install pycapnp==0.5.8 --user

Configure and generate C++ build files

See the README for details about the following commands.

mkdir -p $NUPIC_CORE/build/scripts
cd $NUPIC_CORE/build/scripts
cmake $NUPIC_CORE -DCMAKE_INSTALL_PREFIX=../release -DPY_EXTENSIONS_DIR=$NUPIC_CORE/bindings/py/nupic/bindings


Note: The -j3 option specifies '3' as the maximum number of parallel jobs/threads that Make will use during the build in order to gain speed. However, you can increase this number depending your CPU.

make -j3


Stay in the $NUPIC_CORE/build/scripts directory.

make install

Run Tests

cd $NUPIC_CORE/build/release/bin

Install the Python Bindings

This will expose the C++ nupic.core program to python by installing the nupic.bindings python module.

python install --user



python install --user

Run Tests

First, make sure that py.test is on your path. It got installed into ~/.local/bin in a previous step.

export PATH=~/.local/bin:$PATH

Now run unit tests.

python $NUPIC/scripts/ -u

Note: If executing that last line generates an error such as "ImportError: No module named _markerlib", it may be caused by Ubuntu's pre-installed pip version which seems to be outdated. Running these 3 lines should fix the problem.

sudo apt-get install python-pip python-dev build-essential
sudo pip install --upgrade pip
sudo pip install --upgrade virtualenv

Running swarms

To run swarms, a MySQL database is necessary. If MySQL is already installed and has already been setup with a username and password, then you will need to modify the user and password values accordingly in the file nupic-default.xml located at $NUPIC/config/default/nupic-default.xml. However, if it is not installed, then simply executing

sudo apt-get install mysql-server

will install MySQL on Ubuntu. During installation, you will be prompted to enter a password for the "root" user. Simply hit the Enter button to complete installation. By default, this should create a MySQL database located at "localhost" with the username "root" and an empty password. Once done, you can verify that everything was setup correctly by executing

python $NUPIC/examples/swarm/

which should generate some text, with the very last line confirming that the connection test was successful.

Note: To run swarms, the script must be executed. For this to work, you must either add the script to the python path or simply invoke the entire path everytime as such

$NUPIC/scripts/ $NUPIC/your_folder/yoursearch_def.json --maxWorkers=4

More details and in-depth explanation about swarms and how to run them can be found here