Adaptive Layer Splitting for Wireless LLM Inference in Edge Computing: A Model-Based Reinforcement Learning Approach

TL;DR

This paper presents a model-based reinforcement learning framework to optimize the splitting point of large language models in edge computing, improving inference efficiency and balancing computational load in wireless environments.

Contribution

It introduces a novel MBRL-based approach for adaptive LLM splitting point selection, reducing evaluation costs and enhancing deployment efficiency in edge scenarios.

Findings

Effective balancing of inference performance and computational load

Significant reduction in performance evaluation costs

Robust performance under varying network conditions

Abstract

Optimizing the deployment of large language models (LLMs) in edge computing environments is critical for enhancing privacy and computational efficiency. Toward efficient wireless LLM inference in edge computing, this study comprehensively analyzes the impact of different splitting points in mainstream open-source LLMs. On this basis, this study introduces a framework taking inspiration from model-based reinforcement learning (MBRL) to determine the optimal splitting point across the edge and user equipment (UE). By incorporating a reward surrogate model, our approach significantly reduces the computational cost of frequent performance evaluations. Extensive simulations demonstrate that this method effectively balances inference performance and computational load under varying network conditions, providing a robust solution for LLM deployment in decentralized settings.