"GliDe with a CaPE: A Low-Hassle Method to Accelerate Speculative Decoding."

Cunxiao Du et al. (2024)

Details and statistics

DOI: 10.48550/ARXIV.2402.02082

access: open

type: Informal or Other Publication

metadata version: 2024-03-08

a service of  Schloss Dagstuhl - Leibniz Center for Informatics