A Large Batch Optimizer Reality Check: Traditional, Generic Optimizers Suffice Across Batch Sizes

Benchmark Model Rank Results
image-classification-on-imagenetResNet-50 MLPerf v0.7 - 2512 steps–Top 1 Accuracy: 75.92%
question-answering-on-squad11BERT-Large 32k batch size with AdamW–F1: 91.58