SOTAVerified

When Attention Meets Fast Recurrence: Training Language Models with Reduced Compute

2021-02-24EMNLP 2021Code Available2· sign in to hype

Tao Lei

Code Available — Be the first to reproduce this paper.

Reproduce

Code

Abstract

Large language models have become increasingly difficult to train because of the growing computation time and cost. In this work, we present SRU++, a highly-efficient architecture that combines fast recurrence and attention for sequence modeling. SRU++ exhibits strong modeling capacity and training efficiency. On standard language modeling tasks such as Enwik8, Wiki-103 and Billion Word datasets, our model obtains better bits-per-character and perplexity while using 3x-10x less training cost compared to top-performing Transformer models. For instance, our model achieves a state-of-the-art result on the Enwik8 dataset using 1.6 days of training on an 8-GPU machine. We further demonstrate that SRU++ requires minimal attention for near state-of-the-art performance. Our results suggest jointly leveraging fast recurrence with little attention as a promising direction for accelerating model training and inference.

Tasks

Benchmark Results

DatasetModelMetricClaimedVerifiedStatus
enwik8SRU++ LargeBit per Character (BPC)0.95Unverified
enwik8SRU++ BaseBit per Character (BPC)0.97Unverified
One Billion WordSRU++ LargePPL23.5Unverified
One Billion WordSRU++PPL25.1Unverified
WikiText-103SRU++ LargeTest perplexity17.1Unverified
WikiText-103SRU++ BaseTest perplexity18.3Unverified

Reproductions