The Transformer architecture is crucial for numerous AI models, but it still faces challenges in long-range language modeling. Though several specific transformer architectures have been designed to tackle issues of long-range dependencies, existing methods like Transformer-XL are plagued by a high percentage of ineffective memories. In this study, we present a plug-and-play strategy, known as TRAining-free Memory Selection (TRAMS), that selects tokens participating in attention calculation based on one simple metric. This strategy allows us to keep tokens that are likely to have a high attention score with the current queries and ignore the other ones. We have tested our approach on the word-level benchmark (WikiText-103) and the character-level benchmark (enwik8), and the results indicate an improvement without having additional training or adding additional parameters.
@article{arxiv.2310.15494,
title = {TRAMS: Training-free Memory Selection for Long-range Language Modeling},
author = {Haofei Yu and Cunxiang Wang and Yue Zhang and Wei Bi},
journal= {arXiv preprint arXiv:2310.15494},
year = {2023}
}