Block-State Transformers
Jonathan Pilault, Mahan Fathi, Orhan Firat, Christopher Pal, Pierre-Luc Bacon, Ross Goroshin
Abstract
State space models (SSMs) have shown impressive results on tasks that require modeling long-range dependencies and efficiently scale to long sequences owing to their subquadratic runtime complexity. Originally designed for continuous signals, SSMs have shown superior performance on a plethora of tasks, in vision and audio; however, SSMs still lag Transformer performance in Language Modeling tasks. In this work, we propose a hybrid layer named Block-State Transformer (*BST*), that internally combines an SSM sublayer for long-range contextualization, and a Block Transformer sublayer for short-term representation of sequences. We study three different, and completely *parallelizable*, variants that integrate SSMs and block-wise attention. We show that our model outperforms similar Transformer-based architectures on language modeling perplexity and generalizes to longer sequences. In addition, the Block-State Transformer demonstrates a more than *tenfold* increase in speed at the layer level compared to the Block-Recurrent Transformer when model parallelization is employed.
BibTeX
@inproceedings{
pilault2023blockstate,
title={Block-State Transformers},
author={Jonathan Pilault and Mahan Fathi and Orhan Firat and Christopher Pal and Pierre-Luc Bacon and Ross Goroshin},
booktitle={Thirty-seventh Conference on Neural Information Processing Systems},
year={2023},
url={https://openreview.net/forum?id=XRTxIBs2eu}
}