Fetching the paper…

Large Batch Optimization for Deep Learning: Training BERT in 76 minutes · Around