Back to Gap Map Model components / Federated learning

PFLlib

open source / Overall score: 2.0

Personalized federated learning library and benchmark, built to be beginner-friendly. It implements traditional and personalized federated algorithms over heterogeneous-data scenarios, simulates hundreds of clients on a single GPU, includes privacy-evaluation attacks, and maintains a public leaderboard.

Real-device deployment lives in a separate repository (HtFL-OnDevice) and is not part of this record. `TsingZ0` is an individual's account.

Openness

5 high confidence

5.0

license: Apache-2.0(OSI
source: public(the whole library)
core features withheld: no — a research library and benchmark with no paid edition

PFLlib is an Apache-2.0 research library and benchmark with no paid tier.

https://raw.githubusercontent.com/TsingZ0/PFLlib/HEAD/LICENSE recorded 2026-09-27
The LICENSE body: the verbatim Apache License, Version 2.0 text with no added clause.
https://raw.githubusercontent.com/TsingZ0/PFLlib/HEAD/README.md recorded 2026-09-27
README: "PFLlib: Personalized Federated Learning Library and Benchmark", a "beginner-friendly federated learning (FL) library and benchmark" with an official website and leaderboard. No paid tier is named.
https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/TsingZ0%2FPFLlib recorded 2026-09-27
Repository record for TsingZ0/PFLlib: license apache-2.0, not archived, not a fork, pushed 2025-11-25, 2,006 stars.
https://ungh.cc/repos/TsingZ0/PFLlib/files/master recorded 2026-09-27
The 196-file tree of TsingZ0/PFLlib at `master`: one root LICENSE, and no ee/, enterprise/, commercial/, proprietary/ or premium/ path anywhere in it. Path absence only, paired with the statement cited beside it.

Adoption

2 low confidence

2.0

PFLlib is run from its repository and publishes no package, so stars are the fallback.

https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/TsingZ0%2FPFLlib recorded 2026-09-27
stargazers_count = 2,006 for TsingZ0/PFLlib.

Capability

2 high confidence

2.0

PFLlib simulates federations on one PC, and the real-device deployment it points to lives in another repository, level with FedJAX.

https://raw.githubusercontent.com/TsingZ0/PFLlib/HEAD/README.md recorded 2026-09-27
README: "master FL in 2 hours—run it on your PC!"; the platform can simulate "500 clients on one NVIDIA GeForce RTX 3090 GPU card"; "Real-machine deployment: HtFL-OnDevice"; DLG attack evaluation.

Verified 2026-09-27