Gated Slot Attention for Efficient Linear-Time Sequence Modeling
Soochow University, China · Massachusetts Institute of Technology · University of California, Santa Cruz · Suzhou University · Tencent AI Lab · Luxi Technology · LuxiTech Co., Ltd · University of Waterloo · Soochow University, China,
PDF 由论文原始站点提供，PaperCompass 不保存论文文件。
摘要
Linear attention Transformers and their gated variants, celebrated for enabling parallel training and efficient recurrent inference, still fall short in recall-intensive tasks compared to traditional Transformers and demand significant resources for training from scratch.This paper introduces Gated Slot Attention (GSA), which enhances Attention with Bounded-memory-Control (ABC) by incorporating a gating mechanism inspired by Gated Linear Attention (GLA).Essentially, GSA comprises a two-layer GLA linked via $\operatorname{softmax}$, utilizing context-aware memory reading and adaptive forgetting to improve memory capacity while maintaining compact recurrent state size.This design greatly enhances both training and inference efficiency through GLA's hardware-efficient training algorithm and reduced state size.Additionally, retaining the $\operatorname{softmax}$ operation is particularly beneficial in ``finetuning pretrained Transformers to RNNs'' (T2R) settings, reducing the need for extensive training from scratch.Extensive experiments confirm GSA's superior performance in scenarios requiring in-context recall and in T2R settings.