PACER: Blockwise Pre-verification for Speculative Decoding with Adaptive Length

TL;DR

Pacer introduces a dynamic, blockwise pre-verification method for speculative decoding in large language models, significantly improving inference speed by adaptively controlling draft length and reducing unnecessary computations.

Contribution

It proposes a novel trainable pre-verification layer that adaptively determines draft length, enhancing the efficiency of speculative decoding for large language models.

Findings

Achieves up to 2.66x speedup over standard decoding.

Outperforms fixed-length speculative decoding methods.

Attains up to 3.09x speedup when combined with Ouroboros.

Abstract

Speculative decoding (SD) is a powerful technique for accelerating the inference process of large language models (LLMs) without sacrificing accuracy. Typically, SD employs a small draft model to generate a fixed number of draft tokens, which are then verified in parallel by the target model. However, our experiments reveal that the optimal draft length varies significantly across different decoding steps. This variation suggests that using a fixed draft length limits the potential for further improvements in decoding speed. To address this challenge, we propose Pacer, a novel approach that dynamically controls draft length using a lightweight, trainable pre-verification layer. This layer pre-verifies draft tokens blockwise before they are sent to the target model, allowing the draft model to stop token generation if the blockwise pre-verification fails. We implement Pacer on multiple SD model pairs and evaluate its performance across various benchmarks. Our results demonstrate that Pacer achieves up to 2.66x Speedup over autoregressive decoding and consistently outperforms standard speculative decoding. Furthermore, when integrated with Ouroboros, Pacer attains up to 3.09x Speedup.