·
254 commits
to master
since this release
FedScale 0.5 is the first major release of FedScale after years of development.
Major Features
- Distributed/standalone fast-forward FL evaluations
- 21 realistic FL datasets
- 70+ lightweight FL models
- PyTorch and TensorFlow support
- GPU, x86, and ARM hardware backend support
- Real-world client system traces
- Synchronous and asynchronous training with straggler mitigation support
- Homepage, API documentation
Credits
FedScale 0.5 was the work of a large set of new contributors from Michigan and outside. Thanks also to all the FedScale users who have suggested new features or reported bugs.