arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2002.07028cs.LGstat.ML
收藏

多头注意力模型中的低秩瓶颈

Low-Rank Bottleneck in Multi-head Attention Models

Google Research(谷歌研究院);MIT(麻省理工学院)

Srinadh Bhojanapalli, Chulhee Yun, Ankit Singh Rawat, Sashank J. Reddi, Sanjiv Kumar

摘要:

基于注意力的Transformer架构使得自然语言处理领域取得了重大进展。除了新的预训练技术之外,最近的改进关键在于使用相对较大的令牌嵌入维度。不幸的是,这会导致在下游任务中采用的模型过大。在本文中,我们确定的重要因素之一,有助于大的嵌入大小的要求。特别是,我们的分析强调,在当前的架构中,头部的数量和每个头部的大小之间的缩放引起了注意头部的低秩瓶颈,导致这种限制。我们在实验中进一步验证了这一点。作为一个解决方案,我们建议设置的头部大小的注意力单元的输入序列长度,并独立于头部的数量,从而在多头注意力层具有可证明的更强的表达能力。我们的经验表明,这使我们能够训练模型具有相对较小的嵌入维数和更好的性能扩展。

英文摘要:

Attention based Transformer architecture has enabled significant advances in the field of natural language processing. In addition to new pre-training techniques, recent improvements crucially rely on working with a relatively larger embedding dimension for tokens. Unfortunately, this leads to models that are prohibitively large to be employed in the downstream tasks. In this paper we identify one of the important factors contributing to the large embedding size requirement. In particular, our analysis highlights that the scaling between the number of heads and the size of each head in the current architecture gives rise to a low-rank bottleneck in attention heads, causing this limitation. We further validate this in our experiments. As a solution we propose to set the head size of an attention unit to input sequence length, and independent of the number of heads, resulting in multi-head attention layers with provably more expressive power. We empirically show that this allows us to train models with a relatively smaller embedding dimension and with better performance scaling.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑