How do I install Python packages in Jupyter notebooks on OpenShift?

Sometimes you want to install a new package that isn’t in your notebook image, usually while you’re prototyping new techniques and aren’t sure if a new package will be useful. Here’s how to install a new package from within the Jupyter notebook itself. Make sure you’re able to launch a Jupyter notebook on OpenShift: follow the quickstart instructions on the Get Started page and then follow the instructions in the How Do I? recipe for launching a Jupyter notebook.

For this example, we’ll install scikit-learn. We’ll start by pasting the following code in to a notebook cell and then executing it by pressing Shift-Enter:

!pip install --user scikit-learn

This will execute the pip install command as the notebook user. Next up, we need to make sure that the user-local library directory is in Python’s search path — paste this code into a notebook cell and execute it:

import sys
sys.path.append("/home/nbuser/.local/lib/python2.7/site-packages")

Once we have the package installed and the path amended, we can use the newly-installed library from our notebook. Try it out with this handwriting recognition example from the scikit-learn tutorial:

from sklearn import datasets
from sklearn import svm

digits = datasets.load_digits()

clf = svm.SVC(gamma=0.001, C=100.)

clf.fit(digits.data[:-1], digits.target[:-1])
clf.predict(digits.data[-1:])

You can also run a notebook containing these cells on OpenShift with a single command:

oc new-app --template radanalytics-jupyter-notebook \
  -p NAME=scikit-learn-notebook \
  -p JUPYTER_NOTEBOOK_PASSWORD=secret \
  -p JUPYTER_NOTEBOOK_X_INCLUDE=https://radanalytics.io/assets/scikit-learn-notebook/scikit-learn.ipynb