Temperature-Centric Investigation of Speculative Decoding with Knowledge Distillation

Speculative decoding stands as a pivotal technique to expedite inference in autoregressive (large) language models. This method employs a smaller draft model to speculate a block of tokens, which the target model then evaluates for acceptance. Despite a wealth of studies aimed at increasing the effi...

Full description

Saved in:
Bibliographic Details
Published in:arXiv.org 2024-10
Main Authors: Ouyang, Siru, Wang, Shuohang, Jiang, Minhao, Zhong, Ming, Yu, Donghan, Han, Jiawei, Shen, Yelong
Format: Article
Language:English
Subjects:
Configuration management
Decoding
High temperature
Temperature
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Speculative decoding stands as a pivotal technique to expedite inference in autoregressive (large) language models. This method employs a smaller draft model to speculate a block of tokens, which the target model then evaluates for acceptance. Despite a wealth of studies aimed at increasing the efficiency of speculative decoding, the influence of generation configurations on the decoding process remains poorly understood, especially concerning decoding temperatures. This paper delves into the effects of decoding temperatures on speculative decoding's efficacy. Beginning with knowledge distillation (KD), we first highlight the challenge of decoding at higher temperatures, and demonstrate KD in a consistent temperature setting could be a remedy. We also investigate the effects of out-of-domain testing sets with out-of-range temperatures. Building upon these findings, we take an initial step to further the speedup for speculative decoding, particularly in a high-temperature generation setting. Our work offers new insights into how generation configurations drastically affect the performance of speculative decoding, and underscores the need for developing methods that focus on diverse decoding configurations. Code is publically available at https://github.com/ozyyshr/TempSpec.
ISSN:2331-8422