On this page

Jupyter Notebook

Manipulate Notebooks from the Command Line

See StackOverflow.

  • Execute notebook and save results

    jupyter nbconvert --execute --to notebook --inplace notebook.ipynb

  • Export notebook to .py

    jupyter nbconvert --to python notebook.ipynb

  • Export notebook to .html

    (Convert without execution)

    jupyter nbconvert --to html notebook.ipynb

    (Execute before conversion, with no execution time limit)

    jupyter nbconvert --execute --to html --ExecutePreprocessor.timeout=-1 notebook.ipynb

    (Execute before conversion and set output dir to ./html)

    jupyter nbconvert --execute --to html --ExecutePreprocessor.timeout=-1 --output-dir='./html' notebook.ipynb

Programmatically Create Jupyter Notebooks (with different parameters)

Specify parameters in the first cell of the source jupyter notebook (source.ipynb):

You can then use nbparameterise to programmatically replace input values in a notebook before running it.

set-notebook-params.py

import os
import nbformat
from nbparameterise import extract_parameters, parameter_values, replace_definitions


# Extract parameters in source notebook
with open("source.ipynb") as f:
	nb = nbformat.read(f, as_version=4)
orig_parameters = extract_parameters(nb)


# Parameters to replace
model_names = [
	'model3-small-search-space4', 'model3-small-search-space5'
]

# Create new notebooks with replaced parameters
for name in model_names:
    print("Running for model:", name)

    # Update the parameters
    params = parameter_values(orig_parameters, model_name=name)
    new_nb = replace_definitions(nb, params, execute=False)

    # Save
    new_nb_fp = f"new_notebook_with_param_{name}.ipynb"
    with open(new_nb_fp, 'w') as f:
        nbformat.write(new_nb, f)
    
    # Execute new notebook
    os.system(f"jupyter nbconvert --execute --ExecutePreprocessor.timeout=-1 --to notebook --inplace {new_nb_fp}")
    # Convert to html
    os.system(f"jupyter nbconvert --output-dir='./html' --to html {new_nb_fp}")
Pandas Django