Jupyter Notebook

A Jupyter Notebook is a browser-based document that packages written text and executable code together in a single file that others can import, run, edit, and interact with. Widely used across the sciences and in data science, it lets a user write code, run it, and see the results, including tables, charts, and model output, inline, next to the prose that explains what the code does. This makes notebooks a standard tool for reproducible analysis and for developing AI training and inference workflows.

What is a Jupyter Notebook?

A notebook is organized into cells. Code cells execute and display their output directly beneath them; text cells hold formatted notes, equations, and documentation. A user interacts with the notebook through a web browser, while the code actually runs on a Jupyter server that can be on the same machine or on a remote compute host with more powerful resources. The notebook file itself is portable, so an analysis can be shared and re-run by others.

Why Jupyter Notebooks matter

Notebooks combine narrative and computation, which makes results easy to follow and to reproduce. A researcher can present the reasoning, the exact code, and the generated output in one artifact, and a colleague can rerun it to verify the findings. This tight loop of write, run, and inspect suits exploratory data analysis and iterative model development, where seeing intermediate results quickly is essential.

How Jupyter Notebooks work with containers

Notebook environments are commonly paired with containers so that all the software the code depends on travels with it. Because the container carries the exact dependencies, a recipient does not have to reconstruct a matching environment by hand, they start the notebook server from the container and it runs the same way everywhere.

For example, a user might have a container with Python, TensorFlow, and Jupyter installed. They run a Jupyter server from that container on a machine with a GPU and open the interface in a browser at an address such as localhost:8080. In the notebook they write TensorFlow code to train a model, along with instructions for using it, run and test that code directly in the browser, and then distribute the notebook together with the container definition so others can replicate the results on their own hardware.

Jupyter Notebooks in HPC

On clusters, users often launch a notebook server on a compute node with the CPUs, GPUs, or memory their work requires, then connect from their laptop's browser. This gives an interactive, familiar interface backed by the power of shared high-performance computing resources, bridging convenient local-style development with large-scale compute.