MoH: Multi-Head Attention as Mixture-of-Head Attention

TL;DR

This paper introduces Mixture-of-Head attention (MoH), an improved attention mechanism that enhances efficiency and performance by allowing tokens to select relevant heads, outperforming traditional multi-head attention with fewer heads.

Contribution

The paper proposes MoH, a novel attention architecture that treats heads as experts, enabling selective head usage and weighted summation, leading to efficiency gains and improved accuracy.

Findings

MoH outperforms traditional multi-head attention with fewer heads.

Pre-trained models like LLaMA3-8B can be fine-tuned into MoH models.

MoH achieves higher accuracy across multiple benchmarks.

Abstract

In this work, we upgrade the multi-head attention mechanism, the core of the Transformer model, to improve efficiency while maintaining or surpassing the previous accuracy level. We show that multi-head attention can be expressed in the summation form. Drawing on the insight that not all attention heads hold equal significance, we propose Mixture-of-Head attention (MoH), a new architecture that treats attention heads as experts in the Mixture-of-Experts (MoE) mechanism. MoH has two significant advantages: First, MoH enables each token to select the appropriate attention heads, enhancing inference efficiency without compromising accuracy or increasing the number of parameters. Second, MoH replaces the standard summation in multi-head attention with a weighted summation, introducing flexibility to the attention mechanism and unlocking extra performance potential. Extensive experiments on ViT, DiT, and LLMs demonstrate that MoH outperforms multi-head attention by using only 50%-90% of the attention heads. Moreover, we demonstrate that pre-trained multi-head attention models, such as LLaMA3-8B, can be further continue-tuned into our MoH models. Notably, MoH-LLaMA3-8B achieves an average accuracy of 64.0% across 14 benchmarks, outperforming LLaMA3-8B by 2.4% by utilizing only 75% of the attention heads. We believe the proposed MoH is a promising alternative to multi-head attention and provides a strong foundation for developing advanced and efficient attention-based models.