1 paper
Shreyas Subramanian, Bala Krishnamoorthy, Pranav Murthy
Despite significant advances in optimizers for training, most research works use common scheduler choices like Cosine or exponential decay. In this paper, we study \emph{GreedyLR},…