Saved in:
Bibliographic Details
Main Authors: Christopher, Jacob K, Bartoldson, Brian R, Ben-Nun, Tal, Cardei, Michael, Kailkhura, Bhavya, Fioretto, Ferdinando
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2408.05636
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909487054454784
author Christopher, Jacob K
Bartoldson, Brian R
Ben-Nun, Tal
Cardei, Michael
Kailkhura, Bhavya
Fioretto, Ferdinando
author_facet Christopher, Jacob K
Bartoldson, Brian R
Ben-Nun, Tal
Cardei, Michael
Kailkhura, Bhavya
Fioretto, Ferdinando
contents Speculative decoding has emerged as a widely adopted method to accelerate large language model inference without sacrificing the quality of the model outputs. While this technique has facilitated notable speed improvements by enabling parallel sequence verification, its efficiency remains inherently limited by the reliance on incremental token generation in existing draft models. To overcome this limitation, this paper proposes an adaptation of speculative decoding which uses discrete diffusion models to generate draft sequences. This allows parallelization of both the drafting and verification steps, providing significant speedups to the inference process. Our proposed approach, $\textit{Speculative Diffusion Decoding (SpecDiff)}$, is validated on standard language generation benchmarks and empirically demonstrated to provide up to 7.2x speedups over standard generation processes and up to 1.75x speedups over existing speculative decoding approaches.
format Preprint
id arxiv_https___arxiv_org_abs_2408_05636
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Speculative Diffusion Decoding: Accelerating Language Generation through Diffusion
Christopher, Jacob K
Bartoldson, Brian R
Ben-Nun, Tal
Cardei, Michael
Kailkhura, Bhavya
Fioretto, Ferdinando
Computation and Language
Machine Learning
Speculative decoding has emerged as a widely adopted method to accelerate large language model inference without sacrificing the quality of the model outputs. While this technique has facilitated notable speed improvements by enabling parallel sequence verification, its efficiency remains inherently limited by the reliance on incremental token generation in existing draft models. To overcome this limitation, this paper proposes an adaptation of speculative decoding which uses discrete diffusion models to generate draft sequences. This allows parallelization of both the drafting and verification steps, providing significant speedups to the inference process. Our proposed approach, $\textit{Speculative Diffusion Decoding (SpecDiff)}$, is validated on standard language generation benchmarks and empirically demonstrated to provide up to 7.2x speedups over standard generation processes and up to 1.75x speedups over existing speculative decoding approaches.
title Speculative Diffusion Decoding: Accelerating Language Generation through Diffusion
topic Computation and Language
Machine Learning
url https://arxiv.org/abs/2408.05636