Exploiting Inter-Layer Expert Affinity for Accelerating Mixture-of-Experts Model Inference

TL;DR

This paper introduces ExFlow, a lightweight optimization that exploits inter-layer expert affinity in pre-trained Mixture-of-Experts models to significantly reduce communication overhead and accelerate inference on distributed systems.

Contribution

It presents a novel context-coherent expert parallelism method that requires only one Alltoall communication, improving efficiency without fine-tuning or accuracy loss.

Findings

Reduces cross-GPU routing latency by up to 67%

Achieves up to 2.2x inference throughput improvement

Demonstrates implicit expert affinity in pre-trained GPT MoE models

Abstract

In large language models like the Generative Pre-trained Transformer, the Mixture of Experts paradigm has emerged as a powerful technique for enhancing model expressiveness and accuracy. However, deploying GPT MoE models for parallel inference on distributed systems presents significant challenges, primarily due to the extensive Alltoall communication required for expert routing and aggregation. This communication bottleneck exacerbates the already complex computational landscape, hindering the efficient utilization of high-performance computing resources. In this paper, we propose a lightweight optimization technique called ExFlow, to largely accelerate the inference of these MoE models. We take a new perspective on alleviating the communication overhead by exploiting the inter-layer expert affinity. Unlike previous methods, our solution can be directly applied to pre-trained MoE models without any fine-tuning or accuracy degradation. By proposing a context-coherent expert parallelism on distributed systems, our design only uses one Alltoall communication to deliver the same functionality while previous methods all require two Alltoalls. By carefully examining the conditional probability in tokens' routing across multiple layers, we proved that pre-trained GPT MoE models implicitly exhibit a strong inter-layer expert affinity. We then design an efficient integer programming model to capture such features and show that by properly placing the experts on corresponding GPUs, we can reduce up to 67% cross-GPU routing latency. Our solution beats the cutting-edge MoE implementations with experts from 8 to 64, with up to 2.2x improvement in inference throughput. We further provide a detailed study of how the model implicitly acquires this expert affinity at the very early training stage and how this affinity evolves and stabilizes during training.