FedML Parrot

FedML Parrot is a scalable federated learning system designed to address the inefficiencies and limitations of existing simulation platforms. Current research focuses on improving training efficiency, particularly for large-scale experiments with heterogeneous devices and stateful clients, employing techniques like sequential training and heterogeneity-aware scheduling. This work is significant because it facilitates more realistic and efficient simulation of federated learning, enabling researchers to develop and test new algorithms and security mechanisms more effectively, ultimately accelerating the practical adoption of federated learning across diverse applications.