NanoFed - Federated Learning Library¶

Beta Release

NanoFed is currently in beta. APIs may change in future versions.

Overview¶

NanoFed is a lightweight, modular federated learning library. Built on PyTorch, it enables distributed model training while preserving data privacy.

Quick Installation¶

pip install nanofed
poetry install nanofed  # Using poetry

Key Features¶

🔒 Privacy First

Client data never leaves local devices
Secure model update transmission
Privacy-preserving aggregation

🚀 Easy to Use

Simple, intuitive API
PyTorch integration
Clear documentation

🔧 Flexible

Custom model support
Pluggable aggregation strategies
Extensible architecture

💻 Production Ready

Async communication
Robust error handling
Comprehensive logging

Quick Example¶

Train a model using federated learning in just a few lines of code:

Basic Training Setup¶

import asyncio
from pathlib import Path

from nanofed import HTTPClient
from nanofed import TorchTrainer, TrainingConfig

async def run_client(client_id: str, server_url: str):
   training_config = TrainingConfig(
      epochs=1,
      batch_size=256,
      learning_rate=0.1
   )
   async with HTTPClient(server_url, client_id) as client:
      model_state, _ = await client.fetch_global_model()
      await client.submit_update(model, metrics)


if __name__ == "__main__":
   asyncio.run(run_client("client1", "http://localhost:8080"))

Components¶

Client Layer

The client component handles local training and server communication:

Local model training
Secure update submission
Data privacy preservation
Training metrics collection

Server Layer

The server coordinates the federated process:

Global model distribution
Update aggregation
Training coordination
Client synchronization

Coordinator

The coordinator manages the overall training process:

Round management
Client tracking
Progress monitoring
Checkpointing

Installation¶

Requirements¶

Required Dependencies¶

Python >=3.10
PyTorch
aiohttp
pydantic

Getting Help¶

🐛 Issue Tracker

Report bugs and request features on our GitHub issue tracker.

https://github.com/camille-004/nanofed/issues

📚 Documentation

Read our comprehensive documentation and guides.

https://nanofed.readthedocs.io

📖 Source Code

Browse the source code on GitHub.

https://github.com/camille-004/nanofed

Warning

This is a beta release. While the core functionality is stable, APIs may change in future versions. Please report any issues you encounter.

License¶

NanoFed is available under the GPL-3.0 License.