KNN-SSD: Enabling Dynamic Self-Speculative Decoding via Nearest Neighbor Layer Set Optimization

TL;DR

KNN-SSD improves the domain robustness of Self-Speculative Decoding for large language models by using KNN search to optimize layer skipping, resulting in significant speedups during inference.

Contribution

This work introduces KNN-SSD, a novel method that enhances domain generalizability of self-speculative decoding through KNN-based layer set optimization.

Findings

Achieved 1.3x-1.6x speedup in LLM inference.

Improved robustness of speculative decoding across diverse domains.

Validated on various models and multiple tasks.

Abstract

Speculative Decoding (SD) has emerged as a widely used paradigm to accelerate the inference of large language models (LLMs) without compromising generation quality. It works by efficiently drafting multiple tokens using a compact model and then verifying them in parallel using the target LLM. Notably, Self-Speculative Decoding proposes skipping certain layers to construct the draft model, which eliminates the need for additional parameters or training. Despite its strengths, we observe in this work that drafting with layer skipping exhibits significant sensitivity to domain shifts, leading to a substantial drop in acceleration performance. To enhance the domain generalizability of this paradigm, we introduce KNN-SSD, an algorithm that leverages K-Nearest Neighbor (KNN) search to match different skipped layers with various domain inputs. We evaluated our algorithm in various models and multiple tasks, observing that its application leads to 1.3x-1.6x speedup in LLM inference.