search

parallelworks / interactive_session

Workflow to start an interactive session in a PW cluster
1 stars 4 forks source link

Bootstrapped Conda for Jupyter IS does not have kernel connect packages #14

Open stefangary opened 1 year ago

stefangary commented 1 year ago

When I create a bootstrapped Conda environment for a Jupyter notebook, that Conda environment does not have the packages necessary to connect to other kernels. In particular, I think the Conda environment that is installed here installs only the most basic Conda environment, and later, jupyter, but nothing else.

When I try to use the Kernel > Change Kernel functionality, no additonal kernels are listed (even if they have jinja2, ipykernel, and requests installed as specified on Slide 3 here). To make those kernels appear in the drop down list, I had to install:

conda install -c conda-forge jupyter
conda install nb_conda_kernels
conda install -c anaconda jinja2
conda install requests
pip install remote_ikernel

(Slide 3 in link above.) @avidalto, would it make sense to include a toggle (e.g. Activate notebook kernels?) to install these packages as well (or always install them when bootstrapping)?

avidalto commented 1 year ago

I think you can install the libraries below by default (in addition to jupyter):

conda install nb_conda_kernels
conda install -c anaconda jinja2

The rest are not needed anymore.