Reducing Underflow in Mixed Precision Training by Gradient Scaling.

Ruizhe Zhao, Brian Vogel, Tanvir Ahmed, Wayne Luk

Browse the full IJCAI paper archive.