Temperature-Centric Investigation of Speculative Decoding with Knowledge Distillation

TL;DR

This paper investigates how decoding temperature influences the efficiency of speculative decoding in large language models, highlighting the importance of temperature settings and knowledge distillation for improved inference speed.

Contribution

It provides a comprehensive analysis of temperature effects on speculative decoding and proposes methods to enhance speedup, especially at higher temperatures.

Findings

Knowledge distillation helps mitigate high-temperature decoding challenges.

Generation configurations significantly impact speculative decoding performance.

High-temperature settings require tailored strategies for effective inference.

Abstract

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.