Fetching the paper…

Stochastic Normalized Gradient Descent with Momentum for Large-Batch Training · Around