Run a Set of Jupyter Notebooks from the Command Line


You’ve got a load of Jupyter notebooks in a directory. You’re going to put them on Github to share with the students in your class, users of your library, readers of your textbook or whoever. Do they all actually run through without errors? Are you sure? Even though you just ran conda update on that dependency?

You could go into each notebook, hit ‘Restart Kernel and Run All Cells…’ and scroll down to make sure there are no exceptions. It’d be nicer to batch run a set of notebooks from the command line. Here’s a script to do that.

# ! python
# coding: utf-8

import os
import argparse
import glob

import nbformat
from nbconvert.preprocessors import ExecutePreprocessor
from nbconvert.preprocessors.execute import CellExecutionError

# Parse args
parser = argparse.ArgumentParser(description="Runs a set of Jupyter \
file_text = """ Notebook file(s) to be run, e.g. '*.ipynb' (default),
'my_nb1.ipynb', 'my_nb1.ipynb my_nb2.ipynb', 'my_dir/*.ipynb'
parser.add_argument('file_list', metavar='F', type=str, nargs='*', 
parser.add_argument('-t', '--timeout', help='Length of time (in secs) a cell \
    can run before raising TimeoutError (default 600).', default=600, 
parser.add_argument('-p', '--run-path', help='The path the notebook will be \
    run from (default pwd).', default='.', required=False)
args = parser.parse_args()
print('Args:', args)
if not args.file_list: # Default file_list
    args.file_list = glob.glob('*.ipynb')

# Check list of notebooks
notebooks = []
print('Notebooks to run:')
for f in args.file_list:
    # Find notebooks but not notebooks previously output from this script
    if f.endswith('.ipynb') and not f.endswith('_out.ipynb'):
        notebooks.append(f[:-6]) # Want the filename without '.ipynb'

# Execute notebooks and output
num_notebooks = len(notebooks)
for i, n in enumerate(notebooks):
    n_out = n + '_out'
    with open(n + '.ipynb') as f:
        nb =, as_version=4)
        ep = ExecutePreprocessor(timeout=int(args.timeout), kernel_name='python3')
            print('Running', n, ':', i, '/', num_notebooks)
            out = ep.preprocess(nb, {'metadata': {'path': args.run_path}})
        except CellExecutionError:
            out = None
            msg = 'Error executing the notebook "%s".\n' % n
            msg += 'See notebook "%s" for the traceback.' % n_out
        except TimeoutError:
            msg = 'Timeout executing the notebook "%s".\n' % n
            # Write output file
            with open(n_out + '.ipynb', mode='wt') as f:
                nbformat.write(nb, f)

You can use it to run all cells in a single notebook from the command line with

python my_nb1.ipynb

You’ll get a new notebook my_nb1_out.ipynb for you to check the output. I’ve chosen not to overwrite the existing notebook because this can introduce git diffs you didn’t want on notebooks that don’t need fixing.

Run a set of notebooks with

python my_nb1.ipynb my_nb2.ipynb my_nb3.ipynb

Again you’ll get notebooks my_nb[1,2,3]_out.ipynb to check.

Run all the notebooks in a directory with

python notebooks/*.ipynb

The default is to run all notebooks in the working directory so


is the same as

python ./*.ipynb


  • --help gives help.

  • --timeout sets a limit in seconds for cell execution. The default is 600. The script will give up and skip to the next notebook after printing a message.

  • --run-path sets the path the notebook will be executed from. If you have relative paths in your notebook, you’ll probably want the run path to be set to the notebook path, so call e.g.

    python notebooks/*.ipynb --run-path=notebooks/