Modern era deep learning models, especially those used in large-scale natural language processing (NLP), aim to achieve better performance by increasing the size (or parameter count) of the model, given sufficiently large training datasets. …