Jupyter Notebooks

Jupyter Notebooks and JupyterLab are best run in Katana OnDemand.

Katana OnDemand comes with some built in environment and kernels that are available for use out of the box.

Python virtual environments¶

If you need to use the Jupyter Notebook or Jupyter Lab with your own Python Virtual Environments you will need to create your own Python Jupyter kernel. Here is an example:

Create and load the virtual environment¶

$ module load python/3.8.3
$ python3 -m venv --system-site-packages /home/z1234567/.venvs/jupyter-kernel
$ source /home/z1234567/.venvs/jupyter-kernel/bin/activate

Create the Jupyter Kernel

Using the helper scriptInstalling the kernel manually

This script can automatically setup Jupyter kernels for use in Katana OnDemand.

(jupyter-kernel) $ install_jupyter_kernels

(jupyter-kernel) $ python3 -m ipykernel install --prefix=$HOME/.local --name=jlab-kernel

Warning

The --name=XXXX isn't strictly necessary, but if you don't use it, the kernel will be called "Python 3". This will not be distinguishable from the Katana supplied Python 3 and could cause confusion.

Now when you load a JupyterLab session, you should see your personal kernel in the list of available kernels. This kernel will have access to your virtual environment.

Conda environments¶

If you need to use the Jupyter Notebook or Jupyter Lab with your own Conda environment you will need to create your own Python or R Jupyter kernel. Here is an example:

Create and activate the environment¶

$ conda create -n my_conda_env -c conda-forge ipykernel
$ conda activate my_conda_env

Create the Jupyter Kernel¶

You can alternatively install the kernel manually: Installing the kernel manually_

(my_conda_env) $ install_jupyter_kernels

Kernels other than ipykernel, need jupyter to be installed in addition to the kernel.

For example, to use the R kernel r-irkernel, you must also install the jupyter package before running the helper script.

(my_conda_env) $ conda install jupyter