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 more than tenfold increase in speed at the layer level compared to the Block-Recurrent Transformer when model parallelization is employed.
@article{arxiv.2306.09539,
title = {Block-State Transformers},
author = {Mahan Fathi and Jonathan Pilault and Orhan Firat and Christopher Pal and Pierre-Luc Bacon and Ross Goroshin},
journal= {arXiv preprint arXiv:2306.09539},
year = {2023}
}
Comments
NeurIPS'23 - Thirty-seventh Conference on Neural Information Processing Systems