Jupyter notebooks tips and tricks

Shortcuts

• Shift + Enter to run a cell (code or markdown).
• A to insert a new cell above the current cell.
• B to insert a new cell below the current cell.
• M to change the current cell to Markdown
• Y to change the current cell to code.
• D + D (twice) to delete the selected cells.

Magic commands

• Time the execution of a single line of code or the whole cell

# Run the code multiple times and find mean runtime
%timeit CODE_LINE
%%timeit CODE_CELL
# Run once and report
%time CODE_LINE
%%time CODE_CELL

• Use ! prefix to run a single bash command line
• %%bash to change the current code cell to run to bash mode, basically writing bash commands in there

%%bash
echo "hello from $BASH"

• %%js, %% html, %%latex, %%python2, %%python3, ... run and render code cells in that specified language or format.
• autoreload IPython extension is exceptionally helpful when you do not want to worry about reloading modules before executing new code. In other words when you change something in a certain module the current notebook uses, changes will take place when you run new code cells without having to worry about anything.

%load_ext autoreload
%autoreload 2

• Embedd tensorboard in jupyter notebook

%load_ext tensorboard
%tensorboard --logdir logs/model_training_logs

• Finally you can list all available magics by running %lsmagic, this will show both line and cell magics currently defined.

Other

• Sometimes you will have this memory hungry variable, you can reclaim memory by setting it to NONE and then forcing gc to run

some_var = None
gc.collect()

• Use sudo service jupyter restart to restart jupyter because every once in a while jupyter would throw a fit and restarting kernels will not be enough to get it back to a responsive state.
• Add ? before almost any function, variable, ... and run the code cell to access its documentation.
• tqdm which means "progress" in Arabic (taqadum, تقدّم) is not really related to jupyter notebooks but it can be used to show a smart progress meter. Just by wrapping any iterable with tqdm(iterable)

from tqdm import tqdm
for i in tqdm(range(10000)):
    pass

• When you want to count the number of files in a directory you can run the following command

!ls DIR_NAME | wc -l

Classic notebooks extensions

There are many great extensions in jupyter_contrib_nbextensions. You should be using Jupyter lab though instead.

First you need to instead jupyter_contrib_nbextensions and then you can install various useful extensions.

pip install jupyter_contrib_nbextensions
jupyter contrib nbextension install --user

These are the ones I love:

• code_prettify backed by autopep8 is great for reformatting code in notebook code cells base on PEP 8 style guide

pip install autopep8
jupyter nbextension enable code_prettify/autopep8

• spellchecker highlights incorrectly spelled words in Markdown cells which saved me from a few embarrassing typos.

jupyter nbextension enable spellchecker/main

• toggle_all_line_numbers as its names suggests it adds a toolbar button to toggle between showing line numbers or not

jupyter nbextension enable toggle_all_line_numbers/main

• varInspector is great for debugging python and R kernels. It displays the values of all defined variables in a floating window

jupyter nbextension enable varInspector/main

Themes

• dunovank/jupyter-themes has one of the best themes I came across. I tried it and then I stopped using it because I switch environments all the time so it was best for me to get used to the stock theme.

pip install jupyterthemes
# dark
jt -t onedork -fs 95 -altp -tfs 11 -nfs 115 -cellw 88% -T
# light
jt -t grade3 -fs 95 -altp -tfs 11 -nfs 115 -cellw 88% -T
# Restore default theme
jt -r

Jupyter lab extensions

There are two extensions that I only use at the moment

• krassowski/jupyterlab-go-to-definition which allow me to use Alt + click to jump to a definition using your mouse, or Ctrl + Alt + B keyboard-only alternative.

jupyter labextension install @krassowski/jupyterlab_go_to_definition

• krassowski/jupyterlab-lsp adds support for code navigation + hover suggestions + linters + autocompletion. Check out their full list of supported language servers

pip install --pre jupyter-lsp
jupyter labextension install @krassowski/jupyterlab-lsp
conda install -c conda-forge python-language-server

Finally you need to rebuild the jupyter lab app

jupyter lab build

Themes

There are many themes out there, the first customization plugin in my list is not a theme though. It is a topbar extension to quickly switch between light and dark themes

jupyter labextension install jupyterlab-topbar-extension jupyterlab-theme-toggle

Here is a list a few themes I used recently

jupyter labextension install @telamonian/theme-darcula
jupyter labextension install @rahlir/theme-gruvbox
jupyter labextension install @kenshohara/theme-nord-extension

Share it and get it in touch, my Twitter DMs are open!