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.

Read more…