What is the FedScale Tool Assembly?
FedScale is an open-source federated learning benchmark and runtime platform. It offers realistic FL tasks, scalable runtime, and diverse datasets, focusing on benchmarking model and system performance at scale. FedScale is designed to address client heterogeneity and large-scale FL deployment challenges, facilitating reproducible FL research across domains such as image classification and speech recognition.
What can you use the FedScale Tool Assembly for?
- Benchmarking federated learning models and systems
- Heterogeneity-aware FL research
- Large-scale FL deployment simulation
Limitations/Remarks
- Focused on benchmarking and research
- Ongoing development and feature expansion
Specific Use Cases/Applications
- Image classification, speech recognition, and other FL domains
- Reproducible FL research and benchmarking
How to access the FedScale Tool Assembly?
- Official Documentation
- GitHub Repository
- Install via pip or from source
References
Original Framework Contributors
- FedScale Community
For a full list, see the FedScale GitHub contributors.
More information
Training
Affiliations
Editor
KU Leuven / University of UHasselt