Optimizing Inference Performance of Transformers on CPUs

Dave Dice, Alex Kogan

The Transformer architecture revolutionized the field of natural language processing (NLP). Transformers-based models (e.g., BERT) power many important Web services, such as search…