Researchers at MIT have created an algorithm-based architecture called SpAtten that reduces attention computation and memory access in natural language processing (NLP) systems.
source https://www.allaboutcircuits.com/news/hardware-key-advancing-NLP/
No comments:
Post a Comment