New AI technique speeds up language models on edge devices

In a research paper, scientists propose a technique -- Hardware-Aware Transformers (HAT) -- that finds Transformer-based models optimized for edge devices.