📘 Introduction

P2PFL is a general-purpose open-source library for the execution (simulated and in real environments) of Decentralized Federated Learning systems, specifically making use of P2P networks and the Gossip protocol. It empowers researchers and developers to build and experiment with federated learning models without relying on centralized servers, looking towards a more democratized and descentralized future for the ML community.

🤔 Why P2PFL?

Traditional federated learning often depends on a central server to orchestrate training and manage communication between participants. This approach introduces several limitations and delegates too many responsabilities to the central server. P2PFL allows nodes to communicate directly with each other, thus reducing the dependence on an external server.

✨ Key Features

P2PFL offers a range of features designed to make decentralized federated learning accessible and efficient. For detailed information, please refer to our components.

Feature

Description

🚀 Easy to Use

Get started quickly with our intuitive API.

🛡️ Reliable

Built for fault tolerance and resilience.

🌐 Scalable

Leverages the power of peer-to-peer networks.

🧪 Versatile

Experiment in simulated or real-world environments.

🔒 Private

Prioritizes data privacy with decentralized architecture.

🧩 Flexible

Designed to be easy to modify.

📈 Real-time Monitoring

Manage and track experiment through P2PFL Web Services platform.

🧠 ML Frameworks

Seamlessly integrate PyTorch, TensorFlow/Keras, and JAX models.

📡 Communication Protocol Agnostic

Choose the communication protocol that best suits your needs (e.g., gRPC).

🔌 Integrations

Enhanced capabilities through integrations: Hugging Face Datasets, ML frameworks, communication protocols, and Ray for large-scale simulations.

🫂 Who Should Use P2PFL?

In an era where decentralized AI and privacy-preserving technologies are gaining prominence, P2PFL emerges as a general-purpose tool designed for users aligned with the goals of federated learning, particularly those focusing on decentralized setups. It’s a valuable resource for researching, developing, and deploying solutions in this evolving landscape.

Specifically, P2PFL caters to:

  • 🧪 Researchers Exploring novel decentralized algorithms and protocols. A flexible platform for testing hypotheses and validating research in federated learning.

  • 💻 Developers Create machine learning based applications that are scalable, privacy-aware and allow for decentralized data management.

  • 🎓 Educators Teaching and learning about decentralized systems and federated learning. A hands-on tool for exploring concepts, conducting experiments, and showcasing real-world use cases.

🏌 Getting started

P2PFL is designed to simplify the journey to federated learning. A set of resources is provided to familiarize users with the framework.

Start by referring to the simple installation guide. Once the environment is set up, it is suggested to continue with the quickstart and the various tutorials to get an overview of the library’s functionality. For a deeper dive, review the components to understand the underlying architecture.

If you need further clarification, the implementation documentation provides detailed information. If you still have questions, feel free to open an issue or contact the development team directly. Help is at your disposal.

🤝 Contributing

We are compromised with the community to develop a more democratic AI ecosystem. so feel free to contribute! See contributing for guidelines on how to start. Please, adhere to the project’s code of conduct that you can find on code of conduct.

💬 Community

Connect with us and stay updated: