Scalable Efficient Training of Large Language Models with Low-dimensional Projected Attention

TL;DR

This paper introduces Low-dimensional Projected Attention (LPA), a scalable method that enhances large language models by applying low-rank projections to attention layers, improving efficiency and effectiveness at various scales.

Contribution

The paper proposes LPA, a novel low-rank attention module that improves large language model training efficiency and performance without sacrificing scalability.

Findings

LPA saves up to 12.4% training time.

LPA achieves about 5% better test perplexity.

LPA improves downstream task performance.

Abstract

Improving the effectiveness and efficiency of large language models (LLMs) simultaneously is a critical yet challenging research goal. In this paper, we find that low-rank pre-training, normally considered as efficient methods that will compromise performance, can be scalably effective when reduced parameters are precisely targeted. Specifically, applying the low-dimensional module only to the attention layer -- resolves this issue and enhances both effectiveness and efficiency. We refer to this structure as Low-dimensional Projected Attention (LPA) and provide an explanatory analysis. Through extensive experimentation at parameter scales of 130M, 370M, and scaling up to 3B, we have validated the effectiveness and scalability of LPA. Our results show that LPA model can save up to 12.4% in time while achieving an approximate 5% improvement in test perplexity (ppl) and on downstream tasks compared with the vanilla Transformer.