SpecHub: Provable Acceleration to Multi-Draft Speculative Decoding

Ryan Sun, Tianyi Zhou, Xun Chen, Lichao Sun


Abstract
Large Language Models (LLMs) have become essential in advancing natural language processing (NLP) tasks, but their sequential token generation limits inference speed. Multi-Draft Speculative Decoding (MDSD) offers a promising solution by using a smaller draft model to generate multiple token sequences, which the target LLM verifies in parallel.However, current heuristic approaches, such as Recursive Rejection Sampling (RRS), suffer from low acceptance rates in subsequent drafts, limiting the advantages of using multiple drafts. Meanwhile, Optimal Transport with Membership Cost (OTM) can theoretically improve acceptance rates, but its computational cost is too high for real-time use.We present SpecHub, a novel, efficient sampling-verification method for MDSD that improves acceptance rates with only linear computational overhead. By simplifying the OTM problem into a compact Linear Programming model, SpecHub significantly reduces computational complexity. It further accelerates sampling by leveraging a sparse joint distribution, focusing computation on high-probability token sequences.%It integrates seamlessly into existing MDSD frameworks.In extensive experiments, Spechub consistently generates 0.05-0.27 and 0.02-0.16 more tokens per step than RRS and RRS without replacement. We attach our code at https://github.com/MasterGodzilla/Speculative_decoding_OT.
Anthology ID:
2024.emnlp-main.1148
Volume:
Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing
Month:
November
Year:
2024
Address:
Miami, Florida, USA
Editors:
Yaser Al-Onaizan, Mohit Bansal, Yun-Nung Chen
Venue:
EMNLP
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
20620–20641
Language:
URL:
https://aclanthology.org/2024.emnlp-main.1148
DOI:
Bibkey:
Cite (ACL):
Ryan Sun, Tianyi Zhou, Xun Chen, and Lichao Sun. 2024. SpecHub: Provable Acceleration to Multi-Draft Speculative Decoding. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 20620–20641, Miami, Florida, USA. Association for Computational Linguistics.
Cite (Informal):
SpecHub: Provable Acceleration to Multi-Draft Speculative Decoding (Sun et al., EMNLP 2024)
Copy Citation:
PDF:
https://aclanthology.org/2024.emnlp-main.1148.pdf