English

Pit One Against Many: Leveraging Attention-head Embeddings for Parameter-efficient Multi-head Attention

Computation and Language 2023-10-13 v1

Abstract

Scaling pre-trained language models has resulted in large performance gains in various natural language processing tasks but comes with a large cost in memory requirements. Inspired by the position embeddings in transformers, we aim to simplify and reduce the memory footprint of the multi-head attention (MHA) mechanism. We propose an alternative module that uses only a single shared projection matrix and multiple head embeddings (MHE), i.e. one per head. We empirically demonstrate that our MHE attention is substantially more memory efficient compared to alternative attention mechanisms while achieving high predictive performance retention ratio to vanilla MHA on several downstream tasks. MHE attention only requires a negligible fraction of additional parameters (3nd3nd, where nn is the number of attention heads and dd the size of the head embeddings) compared to a single-head attention, while MHA requires (3n23n)d23nd(3n^2-3n)d^2-3nd additional parameters.

Keywords

Cite

@article{arxiv.2310.07911,
  title  = {Pit One Against Many: Leveraging Attention-head Embeddings for Parameter-efficient Multi-head Attention},
  author = {Huiyin Xue and Nikolaos Aletras},
  journal= {arXiv preprint arXiv:2310.07911},
  year   = {2023}
}

Comments

Accepted at EMNLP 2023 Findings

R2 v1 2026-06-28T12:47:59.878Z