Poplar: Efficient Scaling of Distributed DNN Training on Heterogeneous GPU Clusters.

WenZheng Zhang, Yang Hu, Jing Shi, Xiaoying Bai

Browse the full AAAI paper archive.