Virtual environments in Jupyter Notebook

Virtual environments are great for compartmentalizing dependencies for each project. Jupyter notebooks are great for rapid prototyping. How to combine the two?

One of the very first things when starting a new Python project should be creating a new virtual environment. They help maintain consistency, keep dependencies isolated and mitigate version conflicts among other things.

Let’s start by creating a new virtual environment called venv for Python and activate the same as follows,

virtualenv -p python3 venv
source venv/bin/activate

Note: We added an additional flag -p to ensure that Python 3 is the default interpreter in our environment since Python 2 has reached its end of life (goodbye ol’ friend).

Once we are inside the virtualenv we can now install all our dependencies via pip install <package> or pip install -r requirements.txt if we happen to have a requirements.txt file for our project.

Now for the interesting bit: To make this venv available as a kernel in our Jupyter notebooks, we simply need to install ipykernel within the virtual environment and then register the same as follows,

pip install ipykernel
ipython kernel install --user --name=<foo-kernel>

Now whenever we launch a new jupyter notebook instance, we would be able to select foo-kernel beside the main global python interpreter (usually located in /usr/bin/python3) by navigating to “Kernel” and then clicking on “Change kernel” in our Jupyter notebook.

An alternative to the same, abeit less graceful, is to install Jupyter notebook in each environment seperately and launch it from within the environment for the packages to be made available.