Jupyter

Introduction

Jupyter is a web-based interactive environment for creating and running language-agnostic HTML notebook applications. The three core programming languages Julia, python, and R are installed via so-called kernels. Each cell in a notebook can contain code, text (using GitHub Flavored Markdown), mathematics, plots and even images or videos. They are mainly used for creating and sharing computational documents. Jupyter offers a simple, streamlined, document-centric experience.

Installation and Kernels

While Jupyter runs code in many programming languages, Python is a requirement for installing the Jupyter Notebook library. After updating the integrated pip3 package manager to the latest version with

pip3 install --upgrade pip

install the Jupyter Notebook package.

pip3 install jupyter

The IPython kernel comes pre-installed so you can start using Python in notebooks straight away. To run notebooks in other languages such as Julia or R you have to install additional kernels. A full list of Jupyter kernels is available here https://github.com/jupyter/jupyter/wiki/Jupyter-kernels. For instructions on how to install a new kernel we refer to the documentation of each individual programming language.

Basic Steps

Start the notebook server from the command line using

jupyter notebook

This will print some information in your terminal including the URL of the web application. By default this is http://localhost:8888. In the dashboard you can start, stop, and create new notebooks or list files and directories.

Tutorial

Based on the PyTorch tutorial https://pytorch.org/tutorials/beginner/basics/.

Disadvantages

• Version control

• Reproducability

• Running cells in different order

• Missing interactive elements