Scrooge: Accelerating Attention Inference in LLMs via Early Termination Mechanism.
Gwangeun Byeon, Seongwook Kim, Taein Kim, Jungmin Lee, Seokin Hong
Browse the full DATE paper archive.
Gwangeun Byeon, Seongwook Kim, Taein Kim, Jungmin Lee, Seokin Hong
Browse the full DATE paper archive.