A semi-autoregressive speculative decoding framework with coupled sequential glancing distillation and a custom decoding tree achieves 2.91x to 3.51x lossless speedup on Vicuna and LLaMA2-Chat.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Falcon: Faster and Parallel Inference of Large Language Models through Enhanced Semi-Autoregressive Drafting and Custom-Designed Decoding Tree
A semi-autoregressive speculative decoding framework with coupled sequential glancing distillation and a custom decoding tree achieves 2.91x to 3.51x lossless speedup on Vicuna and LLaMA2-Chat.