Reformer: The Efficient Transformer

Nikita Kitaev, Lukasz Kaiser, Anselm Levskaya

Keywords: attention, locality sensitive hashing, memory, transformer

Wednesday: Sequence Representations

Abstract: Large Transformer models routinely achieve state-of-the-art results on a number of tasks but training these models can be prohibitively costly, especially on long sequences. We introduce two techniques to improve the efficiency of Transformers. For one, we replace dot-product attention by one that uses locality-sensitive hashing, changing its complexity from O($L^2$) to O($L \log L$), where $L$ is the length of the sequence. Furthermore, we use reversible residual layers instead of the standard residuals, which allows storing activations only once in the training process instead of N times, where N is the number of layers. The resulting model, the Reformer, performs on par with Transformer models while being much more memory-efficient and much faster on long sequences.

Similar Papers

Gradientless Descent: High-Dimensional Zeroth-Order Optimization
Daniel Golovin, John Karro, Greg Kochanski, Chansoo Lee, Xingyou Song, Qiuyi Zhang,
Sign Bits Are All You Need for Black-Box Attacks
Abdullah Al-Dujaili, Una-May O'Reilly,
On the Global Convergence of Training Deep Linear ResNets
Difan Zou, Philip M. Long, Quanquan Gu,