@inproceedings{ouyang-etal-2024-temperature,
title = "Temperature-Centric Investigation of Speculative Decoding with Knowledge Distillation",
author = "Ouyang, Siru and
Wang, Shuohang and
Jiang, Minhao and
Zhong, Ming and
Yu, Donghan and
Han, Jiawei and
Shen, Yelong",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2024",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-emnlp.767",
pages = "13125--13137",
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.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="ouyang-etal-2024-temperature">
<titleInfo>
<title>Temperature-Centric Investigation of Speculative Decoding with Knowledge Distillation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Siru</namePart>
<namePart type="family">Ouyang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shuohang</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Minhao</namePart>
<namePart type="family">Jiang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ming</namePart>
<namePart type="family">Zhong</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Donghan</namePart>
<namePart type="family">Yu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiawei</namePart>
<namePart type="family">Han</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yelong</namePart>
<namePart type="family">Shen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2024-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: EMNLP 2024</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yaser</namePart>
<namePart type="family">Al-Onaizan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mohit</namePart>
<namePart type="family">Bansal</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yun-Nung</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Miami, Florida, USA</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<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.</abstract>
<identifier type="citekey">ouyang-etal-2024-temperature</identifier>
<location>
<url>https://aclanthology.org/2024.findings-emnlp.767</url>
</location>
<part>
<date>2024-11</date>
<extent unit="page">
<start>13125</start>
<end>13137</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Temperature-Centric Investigation of Speculative Decoding with Knowledge Distillation
%A Ouyang, Siru
%A Wang, Shuohang
%A Jiang, Minhao
%A Zhong, Ming
%A Yu, Donghan
%A Han, Jiawei
%A Shen, Yelong
%Y Al-Onaizan, Yaser
%Y Bansal, Mohit
%Y Chen, Yun-Nung
%S Findings of the Association for Computational Linguistics: EMNLP 2024
%D 2024
%8 November
%I Association for Computational Linguistics
%C Miami, Florida, USA
%F ouyang-etal-2024-temperature
%X 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.
%U https://aclanthology.org/2024.findings-emnlp.767
%P 13125-13137
Markdown (Informal)
[Temperature-Centric Investigation of Speculative Decoding with Knowledge Distillation](https://aclanthology.org/2024.findings-emnlp.767) (Ouyang et al., Findings 2024)
ACL
- Siru Ouyang, Shuohang Wang, Minhao Jiang, Ming Zhong, Donghan Yu, Jiawei Han, and Yelong Shen. 2024. Temperature-Centric Investigation of Speculative Decoding with Knowledge Distillation. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 13125–13137, Miami, Florida, USA. Association for Computational Linguistics.