Tutorials#

These tutorials give you hands-on experience building and running DFM federations, from first steps to real-world data pipelines. They are designed to be followed in order — each series builds on concepts from the previous one.

Tip

Recommended reading order:

  1. Zero to Thirty

  2. Weather Federation

Running Tutorials Locally#

Clone the repository and install DFM with the tutorial dependencies:

git clone https://github.com/NVIDIA/data-federation-mesh.git
cd data-federation-mesh
uv sync --all-packages --extra tutorials

Tip

See the README for detailed setup instructions, including how to install uv.

The Zero to Thirty tutorials are step-by-step written guides — all commands can be run directly in your terminal. The Weather Federation tutorials are Jupyter notebooks — launch JupyterLab and open them in order:

uv run jupyter lab

Zero to Thirty#

The fastest path to a working federation. In two short tutorials you will create a federation from scratch, write an adapter, and run a pipeline — first locally using Python multiprocessing, then over NVIDIA Flare’s POC infrastructure on the same machine. No prior DFM experience required.

Zero to Thirty

Weather Federation#

Jupyter notebooks that show DFM applied to a real-world use case: distributed weather data loading and processing. Assumes you are comfortable with the basics covered in Zero to Thirty.

Note

These tutorials are designed to be run interactively as Jupyter notebooks. The pages below are a static reference. To run them yourself, follow the Running Tutorials Locally instructions.

Weather Federation