beiyuouo / fedhf

🔨 A Flexible Federated Learning Simulator for Heterogeneous and Asynchronous.

FedHF

FedHF is a loosely coupled, Heterogeneous resource supported, and Flexible federated learning framework.

Accelerate your research

Features

• Losely coupled
• Heterogeneous resource supported
• Flexible federated learning framework
• Support for asynchronous aggregation
• Support for multiple federated learning algorithms

Algorithms Supported

Synchronous Aggregation

• [FedAvg] Communication-Efficient Learning of Deep Networks from Decentralized Data(AISTAT) [paper]

Asynchronous Aggregation

• [FedAsync] Asynchronous Federated Optimization(OPT2020) [paper]

Tiered Aggregation

• [TiFL] TiFL: A Tier-based Federated Learning System (HPDC 2020) [paper]

Getting Start

pip install fedhf

# If you want to use wandb to view log, please login first
wandb login

You can see the Document for more details.

Contributing

For more information, please see the Contributing page.

Citation

In progress

Licence

This work is provided under Apache License Version 2.0.

Acknowledgement

Many thanks to FedLab and FedML for their great work.