Dynamic Hot Expert Replication with Load and Topology-Aware Joint Gating for Distributed MoE Inference.
Jing Yang, Huaqing Tu, Gongming Zhao, Hongli Xu, Baoqing Wang, Long Li
Browse the full IWQoS paper archive.
Jing Yang, Huaqing Tu, Gongming Zhao, Hongli Xu, Baoqing Wang, Long Li
Browse the full IWQoS paper archive.