arXiv - CSCL: "Draft & Verify: Lossless Large Language Model Acc…" - Qoto Mastodon

arXiv - CSCL @arxiv_cscl@qoto.org

Draft & Verify: Lossless Large Language Model Acceleration via Self-Speculative Decoding. (arXiv:2309.08168v1 [cs.CL])

http://arxiv.org/abs/2309.08168 #arXiv #NLProc

Sep 18, 2023, 03:17 · · arxiv-cscl · · ·

Sign in to participate in the conversation