Decentralized Training — Fellowship Code Reference

A growing collection of small, well-documented machine-learning examples built during the Decentralized Training fellowship. The aim is to give fellows a readable reference source — runnable code paired with plain-language explanations — that they can learn from and adapt.

---

Structure

The repo is organized one directory per week. Each week is a self-contained project with its own README.md, dependencies, and run instructions.

.
└── week-2/   # Federated learning on MNIST with Flower (FedAvg)

So far only week-2 exists. Start there: week-2/README.md walks through federated learning end-to-end, from the model to the server's aggregation step.

More weeks will be added here as the fellowship progresses.

---

Getting started

Pick a week, cd into it, and follow its README:

cd week-2
python -m venv .venv && source .venv/bin/activate   # Windows: .venv\Scripts\activate
pip install -r requirements.txt
python main.py

---

Contributing

Contributions from fellows are very welcome — whether that's a new week, a fix, a clearer explanation, or an experiment that helped something click.

A few guidelines to keep this useful as a reference:

• Keep each week self-contained. Add a new top-level week-N/ directory with its own README.md, requirements.txt, and run instructions.
• Optimize for reading. Favor clarity over cleverness; comment the why, not just the what. Each file should map cleanly to one concept where possible.
• Make it runnable. Pin dependencies and verify a clean venv runs your code from scratch.
• Document the gotchas. If you hit a bug or a confusing pitfall, write it down — that's often the most valuable part for the next fellow.

To contribute: fork the repo, add your work on a branch, and open a pull request describing what you built and what a reader should learn from it.

Happy training! 🌼