Cite
Notes
Only stored in your browser.
Attribution
EMS-SD: Efficient Multi-sample Speculative Decoding for Accelerating Large Language Models
arXiv 2024
Rethinking Optimization and Architecture for Tiny Language Models
Kangaroo: Lossless Self-Speculative Decoding via Double Early Exiting
from 3 papers
Kai Han
Yehui Tang
Yunhe Wang
Fangcheng Liu
Chuanjian Liu
Shangling Jui
Sichao Liu
Yi-Qi Hu
Yuchuan Tian
Zhenhua Liu