Ouroboros: Generating Longer Drafts Phrase by Phrase for Faster Speculative Decoding

TL;DR

Ouroboros is a novel method that generates longer drafts phrase by phrase to accelerate speculative decoding in large language models, achieving significant speedups without additional model fine-tuning.

Contribution

It introduces a training-free approach to generate longer drafts in speculative decoding, enhancing speed without fine-tuning models.

Findings

Achieves up to 2.8x speedup over speculative decoding

Achieves up to 3.9x speedup over vanilla decoding

Does not require fine-tuning of models

Abstract

Speculative decoding is a widely used method that accelerates the generation process of large language models (LLMs) with no compromise in model performance. It achieves this goal by using an existing smaller model for drafting and then employing the target LLM to verify the draft in a low-cost parallel manner. Under such a drafting-verification framework, drafting efficiency has become a bottleneck in the final speedup of speculative decoding. Therefore, generating longer drafts at less cost can lead to better decoding speedup. To achieve this, we introduce Ouroboros, which can generate draft phrases to parallelize the drafting process and meanwhile lengthen drafts in a training-free manner. The experimental results on various typical text generation tasks show that Ouroboros can achieve speedups of up to $2.8\times$ over speculative decoding and $3.9\times$ over vanilla decoding, without fine-tuning draft and target models. The source code of Ouroboros is available at https://github.com/thunlp/Ouroboros.