Jupyter notebooks through ssh tunneling

One can launch a test server to check whether everything is working correctly, using the following command:

python -m http.server 8080

Main steps

To start ssh tunneling, run the following on your (local) computer:

ssh -L 8080:localhost:8080 thoellinger@genologin.toulouse.inra.fr

Then launch Jupyter on the server (Genotoul here):

module load system/Anaconda3-5.2.0 && cd /work2/project/regenet/workspace/thoellinger/shared/notebooks && jupyter notebook --no-browser --port 8080

(One could also try with another port - for instance 8081).

Then, on your local computer, open a browser and go to localhost:8080 (resp. localhost:8081). Enter the required token when prompted to access the notebook.

Note that one cannot run the notebook from a node (srun --pty bash then jupyter notebook does not work, because of denied permission). Instead, one can launch the notebook from standard user, then specify in the notebooks the cells that shall be launched from a node or on slurm. For instance, this is what is done here.

Convert a notebook into html from shell

One can simply use:

jupyter nbconvert --to html test_notebooks/first_notebook.ipynb

Or create a dedicated script (e.g. convert_notebooks_to_html.sh) with a content like:

#!/bin/sh                                                                          
path_to_html="/home/thoellinger/public_html/autoupdate/"

module load system/Anaconda3-5.2.0
jupyter nbconvert --to html /work2/project/regenet/workspace/thoellinger/notebooks\
/first_notebook/first_notebook.ipynb --output "${path_to_html}first_notebook"

Use a custom theme with jt

I personally use thfollowing theme:

jt -t monokai -f fira -fs 10 -nf ptsans -nfs 11 -N -kl -cursw 2 -cursc r -cellw 95% -T

The default theme can be restored with:

jt -r

Note that these commands can be directly executed inside a notebook, using the ! prefix, but the page needs to be reloaded to see the changes.

More info available here.