SWIFT: On-the-Fly Self-Speculative Decoding for LLM Inference Acceleration

TL;DR

SWIFT introduces an adaptive, self-speculative decoding method that accelerates large language model inference by skipping layers on-the-fly without extra training or models, achieving significant speedups.

Contribution

It presents a novel, plug-and-play, layer-skipping decoding algorithm that adaptively accelerates LLM inference without auxiliary models or additional training.

Findings

Achieves 1.3x-1.6x speedup in inference.

Preserves the original output distribution.

Works across diverse models and tasks.

Abstract

Speculative decoding (SD) has emerged as a widely used paradigm to accelerate LLM inference without compromising quality. It works by first employing a compact model to draft multiple tokens efficiently and then using the target LLM to verify them in parallel. While this technique has achieved notable speedups, most existing approaches necessitate either additional parameters or extensive training to construct effective draft models, thereby restricting their applicability across different LLMs and tasks. To address this limitation, we explore a novel plug-and-play SD solution with layer-skipping, which skips intermediate layers of the target LLM as the compact draft model. Our analysis reveals that LLMs exhibit great potential for self-acceleration through layer sparsity and the task-specific nature of this sparsity. Building on these insights, we introduce SWIFT, an on-the-fly self-speculative decoding algorithm that adaptively selects intermediate layers of LLMs to skip during inference. SWIFT does not require auxiliary models or additional training, making it a plug-and-play solution for accelerating LLM inference across diverse input data streams. Our extensive experiments across a wide range of models and downstream tasks demonstrate that SWIFT can achieve over a 1.3x-1.6x speedup while preserving the original distribution of the generated text. We release our code in https://github.com/hemingkx/SWIFT.