Efficient Length-Generalizable Attention via Causal Retrieval for Long-Context Language Modeling

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Hu, Xiang, Teng, Zhihao, Zhao, Jun, Wu, Wei, Tu, Kewei
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866915338514333696
author Hu, Xiang
Teng, Zhihao
Zhao, Jun
Wu, Wei
Tu, Kewei
author_facet Hu, Xiang
Teng, Zhihao
Zhao, Jun
Wu, Wei
Tu, Kewei
contents Despite the success of Transformers, handling long contexts remains challenging due to the limited length generalization and quadratic complexity of self-attention. Thus Transformers often require post-training with a larger attention window, significantly increasing computational and memory costs. In this paper, we propose a novel attention mechanism based on dynamic context, Grouped Cross Attention (GCA), which can generalize to 1000 times the pre-training context length while maintaining the ability to access distant information with a constant attention window size. For a given input sequence, we split it into chunks and use each chunk to retrieve top-k relevant past chunks for subsequent text generation. Specifically, unlike most previous works that use an off-the-shelf retriever, our key innovation allows the retriever to learn how to retrieve past chunks that better minimize the auto-regressive loss of subsequent tokens in an end-to-end manner. Such a mechanism accommodates retrieved chunks with a fixed-size attention window to achieve long-range information access, significantly reducing computational and memory costs during training and inference. Experiments show that GCA-based models achieve near-perfect accuracy in passkey retrieval for 16M context lengths, which is 1000 times the training length.
format Preprint
id arxiv_https___arxiv_org_abs_2410_01651
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Efficient Length-Generalizable Attention via Causal Retrieval for Long-Context Language Modeling
Hu, Xiang
Teng, Zhihao
Zhao, Jun
Wu, Wei
Tu, Kewei
Computation and Language
Artificial Intelligence
Despite the success of Transformers, handling long contexts remains challenging due to the limited length generalization and quadratic complexity of self-attention. Thus Transformers often require post-training with a larger attention window, significantly increasing computational and memory costs. In this paper, we propose a novel attention mechanism based on dynamic context, Grouped Cross Attention (GCA), which can generalize to 1000 times the pre-training context length while maintaining the ability to access distant information with a constant attention window size. For a given input sequence, we split it into chunks and use each chunk to retrieve top-k relevant past chunks for subsequent text generation. Specifically, unlike most previous works that use an off-the-shelf retriever, our key innovation allows the retriever to learn how to retrieve past chunks that better minimize the auto-regressive loss of subsequent tokens in an end-to-end manner. Such a mechanism accommodates retrieved chunks with a fixed-size attention window to achieve long-range information access, significantly reducing computational and memory costs during training and inference. Experiments show that GCA-based models achieve near-perfect accuracy in passkey retrieval for 16M context lengths, which is 1000 times the training length.
title Efficient Length-Generalizable Attention via Causal Retrieval for Long-Context Language Modeling
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2410.01651