eth-cscs / jupyter-utils

jupyter-utils

kernel-create

kernel_create.py creates a jupyter kernel in $HOME/.local/share/jupyter/kernels from the python environment that's active on the environment. The newly-created kernel will be automatically available on the notebooks launched from http://jupyter.cscs.ch.

For instance, let's say we want to create a virtual environment and make a jupyter kernel based on it. We first create the virtual environment

module load cray-python/3.6.5.7
python -m venv myenv

For a virtual environment to be used as jupyter kernel, it has to have the ipykernel package installed. It can be installed with pip, only first the environment has to be activated

source myenv/bin/activate
pip install ipykernel

Now everything is ready to create a jupyter kernel from myenv. With the environment still activated, we run the kernel-create command and specify the name we want to give to the environment

kernel-create -n myenv-kernel

Here we named the kernel myenv-kernel. The script should print

Kernel 'myenv-kernel' created successfully in '~/.local/share/jupyter/kernels/myenv-kernel'

Giving a look to the directory ~/.local/share/jupyter/kernels/myenv-kernel, we can see that two files have been created: kernel.json and launcher.

In case it's not possible to activate the virtual environment, the variable $TARGET_PYTHON_EXE can be used to specify the python executable to be used by the kernel:

TARGET_PYTHON_EXE=/path/to/my/bin/python kernel-create -n mykernel