TsingZ0/PFLlib

Health v2 · maintenance only

Adoption not part of the score

2155 stars · 416 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

PFLlib is a beginner-friendly Python library and benchmark for traditional and personalized federated learning, offering 39 algorithms, 24 datasets, and privacy evaluation tools built on PyTorch. It is designed to simulate federated learning scenarios with hundreds of clients on a single consumer GPU.

Use cases

• simulate federated learning experiments on my PC
• compare personalized federated learning algorithms on non-IID data
• evaluate privacy attacks like DLG against federated models
• benchmark federated learning with 500 clients on one GPU
• learn federated learning quickly as a beginner
• test differential privacy in federated training

When to choose

• you want a comprehensive, easy-to-extend FL library with many algorithms and datasets
• you need to benchmark personalized FL under data heterogeneity
• you have limited GPU resources but need large-scale client simulations

When to avoid

• you need production federated learning deployment across real devices rather than simulation
• you require model heterogeneity support, for which HtFLlib is the extended project

Facets

library · maturity active

machine-learning benchmarking privacy data-science machine-learning artificial-intelligence privacy microservices python cross-platform federated-learning personalized-federated-learning non-iid-data differential-privacy pytorch benchmark data-heterogeneity gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="TsingZ0/PFLlib")