SuffixDecoding: Extreme Speculative Decoding for Emerging AI Applications

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Oliaro, Gabriele, Jia, Zhihao, Campos, Daniel, Qiao, Aurick
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866912633907576832
author Oliaro, Gabriele
Jia, Zhihao
Campos, Daniel
Qiao, Aurick
author_facet Oliaro, Gabriele
Jia, Zhihao
Campos, Daniel
Qiao, Aurick
contents Speculative decoding is widely adopted to reduce latency in large language model (LLM) inference by leveraging smaller draft models capable of handling diverse user tasks. However, emerging AI applications, such as LLM-based agents, present unique workload characteristics: instead of diverse independent requests, agentic frameworks typically submit repetitive inference requests, such as multi-agent pipelines performing similar subtasks or self-refinement loops iteratively enhancing outputs. These workloads result in long and highly predictable sequences, which current speculative decoding methods do not effectively exploit. To address this gap, we introduce \emph{SuffixDecoding}, a novel method that utilizes efficient suffix trees to cache long token sequences from prompts and previous outputs. By adaptively speculating more tokens when acceptance likelihood is high and fewer when it is low, SuffixDecoding effectively exploits opportunities for longer speculations while conserving computation when those opportunities are limited. Evaluations on agentic benchmarks, including SWE-Bench and Text-to-SQL, demonstrate that SuffixDecoding achieves speedups of up to 5.3$\times$, outperforming state-of-the-art methods -- 2.8$\times$ faster than model-based approaches like EAGLE-2/3 and 1.9$\times$ faster than model-free approaches such as Token Recycling. SuffixDecoding is open-sourced at https://github.com/snowflakedb/ArcticInference
format Preprint
id arxiv_https___arxiv_org_abs_2411_04975
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle SuffixDecoding: Extreme Speculative Decoding for Emerging AI Applications
Oliaro, Gabriele
Jia, Zhihao
Campos, Daniel
Qiao, Aurick
Computation and Language
Artificial Intelligence
Distributed, Parallel, and Cluster Computing
Machine Learning
Speculative decoding is widely adopted to reduce latency in large language model (LLM) inference by leveraging smaller draft models capable of handling diverse user tasks. However, emerging AI applications, such as LLM-based agents, present unique workload characteristics: instead of diverse independent requests, agentic frameworks typically submit repetitive inference requests, such as multi-agent pipelines performing similar subtasks or self-refinement loops iteratively enhancing outputs. These workloads result in long and highly predictable sequences, which current speculative decoding methods do not effectively exploit. To address this gap, we introduce \emph{SuffixDecoding}, a novel method that utilizes efficient suffix trees to cache long token sequences from prompts and previous outputs. By adaptively speculating more tokens when acceptance likelihood is high and fewer when it is low, SuffixDecoding effectively exploits opportunities for longer speculations while conserving computation when those opportunities are limited. Evaluations on agentic benchmarks, including SWE-Bench and Text-to-SQL, demonstrate that SuffixDecoding achieves speedups of up to 5.3$\times$, outperforming state-of-the-art methods -- 2.8$\times$ faster than model-based approaches like EAGLE-2/3 and 1.9$\times$ faster than model-free approaches such as Token Recycling. SuffixDecoding is open-sourced at https://github.com/snowflakedb/ArcticInference
title SuffixDecoding: Extreme Speculative Decoding for Emerging AI Applications
topic Computation and Language
Artificial Intelligence
Distributed, Parallel, and Cluster Computing
Machine Learning
url https://arxiv.org/abs/2411.04975