← Publications

FedML Parrot: A Scalable Federated Learning Simulation System via Heterogeneity-aware Scheduling on Hierarchical Sequential Training

Zhenheng Tang, Xiaowen Chu, Ryan Yide Ran, Sunwoo Lee, Shaohuai Shi, Yonggang Zhang, Yuxin Wang, Alex Qiaozhong Liang, Salman Avestimehr, Chaoyang He

IEEE Transactions on Computers, 2026

• DOI

This work presents FedML Parrot, a federated learning simulation system that improves training efficiency and reduces hardware requirements through sequential client training, decomposed local/global aggregation, straggler-aware scheduling, and a distributed client state manager, enabling simulation of over 1000 stateful or stateless clients with 1.2 to 4 times faster training than FedScale and 10 to 100 times less memory usage than FedML while allowing seamless transition from simulation to real-world deployment.

BibTeX

@article{tang2026fedml,
  title={FedML Parrot: A Scalable Federated Learning Simulation System via Heterogeneity-aware Scheduling on Hierarchical Sequential Training},
  author={Tang, Zhenheng and Chu, Xiaowen and Ran, Ryan Yide and Lee, Sunwoo and Zhang, Yonggang and Wang, Yuxin and Liang, Alex Qiaozhong and Avestimehr, Salman and He, Chaoyang},
  journal={IEEE Transactions on Computers},
  year={2026},
  publisher={IEEE}
}