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.- Anthology ID:
- 2021.emnlp-main.602
- Volume:
- Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing
- Month:
- November
- Year:
- 2021
- Address:
- Online and Punta Cana, Dominican Republic
- Editors:
- Marie-Francine Moens, Xuanjing Huang, Lucia Specia, Scott Wen-tau Yih
- Venue:
- EMNLP
- SIG:
- Publisher:
- Association for Computational Linguistics
- Note:
- Pages:
- 7633–7648
- Language:
- URL:
- https://preview.aclanthology.org/remove-affiliations/2021.emnlp-main.602/
- DOI:
- 10.18653/v1/2021.emnlp-main.602
- Cite (ACL):
- Tao Lei. 2021. When Attention Meets Fast Recurrence: Training Language Models with Reduced Compute. In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 7633–7648, Online and Punta Cana, Dominican Republic. Association for Computational Linguistics.
- Cite (Informal):
- When Attention Meets Fast Recurrence: Training Language Models with Reduced Compute (Lei, EMNLP 2021)
- PDF:
- https://preview.aclanthology.org/remove-affiliations/2021.emnlp-main.602.pdf
- Code
- asappresearch/sru
- Data
- Billion Word Benchmark, WikiText-103, WikiText-2