227

I have some code in a .ipynb file and got it to the point where I don't really need the "interactive" feature of IPython Notebook. I would like to just run it straight from a Mac Terminal Command Line.

Basically, if this were just a .py file, I believe I could just do python filename.py from the command line. Is there something similar for a .ipynb file?

Borealis
  • 7,514
  • 14
  • 61
  • 110
Vincent
  • 6,572
  • 12
  • 40
  • 56
  • Something like this: https://github.com/paulgb/runipy ? – idjaw Feb 22 '16 at 03:37
  • if you go to the runipy page referenced above you see that it is unmaintained and they point to Jupyter's "execute api" at https://nbconvert.readthedocs.io/en/latest/execute_api.html – markgalassi May 28 '22 at 16:32

12 Answers12

209

nbconvert allows you to run notebooks with the --execute flag:

jupyter nbconvert --execute <notebook>

If you want to run a notebook and produce a new notebook, you can add --to notebook:

jupyter nbconvert --execute --to notebook <notebook>

Or if you want to replace the existing notebook with the new output:

jupyter nbconvert --execute --to notebook --inplace <notebook>

Since that's a really long command, you can use an alias:

alias nbx="jupyter nbconvert --execute --to notebook"
nbx [--inplace] <notebook>
minrk
  • 35,947
  • 9
  • 90
  • 87
  • 12
    This turns the notebook into a static HTML page – Jim Daniël Teunis Sep 29 '18 at 15:25
  • 5
    Documentation or this and programmatic ways of running notebooks can be found at https://nbconvert.readthedocs.io/en/latest/execute_api.html#module-nbconvert.preprocessors – TomDotTom Mar 05 '19 at 19:27
  • Is there a way to specify which cells to execute? For example cell 10 and all below, or a range cell 10 to cell 20. – Al Conrad Jul 25 '19 at 16:19
  • 2
    Also, if you have long running cells, [you can set](https://nbconvert.readthedocs.io/en/latest/execute_api.html) `--ExecutePreprocessor.timeout=[timeout_in_seconds]` (use `-1` for no restriction). – Zaccharie Ramzi Aug 07 '19 at 08:07
  • The latest updates makes it work like this, you get a HTML out the notebook file `jupyter-nbconvert --execute 03-Print\ Formatting\ with\ Strings.ipynb` – b-ak Sep 16 '19 at 13:44
145

From the command line you can convert a notebook to python with this command:

jupyter nbconvert --to python nb.ipynb

https://github.com/jupyter/nbconvert

You may have to install the python mistune package:

sudo pip install -U mistune
evandrix
  • 5,889
  • 4
  • 26
  • 35
ditkin
  • 6,401
  • 1
  • 32
  • 37
  • 14
    Subcommand `ipython nbconvert` is deprecated in favor of jupyter nbconvert – PenguinEngineer Nov 06 '17 at 17:33
  • 18
    I don't think this answer is correct. II realise it works but the question is how to run .ipynb from terminal, not how to convert it into a .py and then run it. – Giacomo Nov 23 '18 at 12:48
  • 3
    Nbconvert will fail if any cell takes longer than 30s to run, you may want to add `--ExecutePreprocessor.timeout=600`. – bckygldstn Jan 24 '19 at 17:15
  • 1
    The latest updates makes it work like this, you get a HTML out the notebook file `jupyter-nbconvert --execute 03-Print\ Formatting\ with\ Strings.ipynb` – b-ak Sep 16 '19 at 13:45
43

In your Terminal run ipython: 

ipython

then locate your script and put there:

%run your_script.ipynb
Martin Ptacek
  • 439
  • 4
  • 3
40

You can export all your code from .ipynb and save it as a .py script. Then you can run the script in your terminal.

code export sample

Hope it helps.

eric
  • 2,447
  • 18
  • 24
17

Using ipython:

ipython --TerminalIPythonApp.file_to_run=<notebook>.ipynb

Normally, I would prefer this option as it is really self-describing. If you prefer to use less characters, use:

ipython -c "%run <notebook>.ipynb"

which is basically what Keto already suggested (unfortunately a little bit hidden) as a comment.

Murmel
  • 4,805
  • 41
  • 50
  • 3
    Guys, admit it, you would not find the comment without this answer, so why not upvote both?:) – mirekphd Jun 09 '21 at 19:04
  • I'd also like to add that it is prudent to specify an absolute path to the notebook (it can save you a lot of searching if `ipython` is not the only part of the "run chain" (in my case `ipython` was inside a bash script run from a Kubernetes CronJob and it would fail without the full absolute path to the notebook). – mirekphd Jun 10 '21 at 14:41
10

For new version instead of:

ipython nbconvert --to python <YourNotebook>.ipynb

You can use jupyter instend of ipython:

jupyter nbconvert --to python <YourNotebook>.ipynb
Vijay Panchal
  • 317
  • 3
  • 8
9

In my case, the command that best suited me was:

jupyter nbconvert --execute --clear-output <notebook>.ipynb

Why? This command does not create extra files (just like a .py file) and the output of the cells is overwritten everytime the notebook is executed.

If you run:

jupyter nbconvert --help

--clear-output Clear output of current file and save in place, overwriting the existing notebook.

alejandro
  • 634
  • 7
  • 8
5

Update with quoted comment by author for better visibility:

Author's note "This project started before Jupyter's execute API, which is now the recommended way to run notebooks from the command-line. Consider runipy deprecated and unmaintained." – Sebastian Palma

Install runipy library that allows running your code on terminal

pip install runipy

After just compiler your code:

runipy <YourNotebookName>.ipynb

You can try cronjob as well. All information is here

Oliver Bestwalter
  • 4,806
  • 29
  • 46
Axis
  • 1,956
  • 1
  • 20
  • 38
  • 13
    Author's note "This project started before Jupyter's execute API, which is now the recommended way to run notebooks from the command-line. Consider runipy deprecated and unmaintained." – de-russification Mar 23 '18 at 15:15
3

I had the same problem and I found papermill. The advantages against the others solutions is that you can see the results while the notebook is running. I find this feature interesting when the notebook takes very long. It is very easy to use:

pip install papermill
papermill notebook.ipynb output.ipynb

It has also, other handy options as saving the output file to Amazon S3, Google Cloud, etc. See the page for more information.

pglez82
  • 409
  • 3
  • 10
2

You can also use the boar package to run your notebook within a python code.

from boar.running import run_notebook

outputs = run_notebook("nb.ipynb")

If you update your notebook, you won't have to convert it again to a python file.

More information at:

https://github.com/alexandreCameron/boar/blob/master/USAGE.md

Alex
  • 150
  • 9
1

From the terminal run

jupyter nbconvert --execute --to notebook --inplace --allow-errors --ExecutePreprocessor.timeout=-1 my_nb.ipynb

The default timeout is 30 seconds. -1 removes the restriction.

If you wish to save the output notebook to a new notebook you can use the flag --output my_new_nb.ipynb

Mohammad S
  • 61
  • 1
  • 3
0

You can also use jupytext https://jupytext.readthedocs.io/en/latest/index.html.

This allows you to pair your notebook. So for each ipynb file you have a .py file as well with some comments. The .py file can be executed as usual. You can enjoy benefits of two worlds with the cost of one extra file though.

Oh, and by the way if you are using version control you can only commit .py files with a nice diff instead of the ugly .ipynb diffs.

(The syntax in the .py files is similar to Databricks notebooks iy you are familiar with them...)

Ken Jiiii
  • 148
  • 4
  • 17