Mixture of Attentions For Speculative Decoding

TL;DR

This paper introduces a Mixture of Attentions architecture for speculative decoding in large language models, achieving faster decoding speeds and robustness in both single-device and client-server scenarios.

Contribution

It proposes a novel Mixture of Attentions architecture for small models in speculative decoding, improving speed, accuracy, and deployment flexibility over existing methods.

Findings

Achieves 9.5% speedup on EAGLE-2

Demonstrates state-of-the-art latency with minimal server calls

Maintains higher accuracy during disconnections

Abstract

The growth in the number of parameters of Large Language Models (LLMs) has led to a significant surge in computational requirements, making them challenging and costly to deploy. Speculative decoding (SD) leverages smaller models to efficiently propose future tokens, which are then verified by the LLM in parallel. Small models that utilise activations from the LLM currently achieve the fastest decoding speeds. However, we identify several limitations of SD models including the lack of on-policyness during training and partial observability. To address these shortcomings, we propose a more grounded architecture for small models by introducing a Mixture of Attentions for SD. Our novel architecture can be applied in two scenarios: a conventional single device deployment and a novel client-server deployment where the small model is hosted on a consumer device and the LLM on a server. In a single-device scenario, we demonstrate state-of-the-art speedups improving EAGLE-2 by 9.5% and its acceptance length by 25%. In a client-server setting, our experiments demonstrate: 1) state-of-the-art latencies with minimal calls to the server for different network conditions, and 2) in the event of a complete disconnection, our approach can maintain higher accuracy compared to other SD methods and demonstrates advantages over API calls to LLMs, which would otherwise be unable to continue the generation process.