Fast Forwarding Low-Rank Training

Adir Rahamim, Naomi Saphra, Sara Kangaslahti, Yonatan Belinkov


Abstract
Parameter efficient finetuning methods like low-rank adaptation (LoRA) aim to reduce the computational costs of finetuning pretrained Language Models (LMs). Enabled by these low-rank settings, we propose an even more efficient optimization strategy: Fast Forward, a simple and effective approach to accelerate large segments of SGD training. In a Fast Forward stage, we repeat the most recent optimizer step until the loss stops improving on a tiny validation set. By alternating between regular optimization steps and Fast Forward stages, Fast Forward provides up to an 87% reduction in FLOPs over standard SGD with Adam. We validate Fast Forward by finetuning various models on different tasks and demonstrate that it speeds up training without compromising model performance. Additionally, we analyze when and how to apply Fast Forward.
Anthology ID:
2024.emnlp-main.535
Volume:
Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing
Month:
November
Year:
2024
Address:
Miami, Florida, USA
Editors:
Yaser Al-Onaizan, Mohit Bansal, Yun-Nung Chen
Venue:
EMNLP
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
9553–9562
Language:
URL:
https://aclanthology.org/2024.emnlp-main.535
DOI:
Bibkey:
Cite (ACL):
Adir Rahamim, Naomi Saphra, Sara Kangaslahti, and Yonatan Belinkov. 2024. Fast Forwarding Low-Rank Training. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 9553–9562, Miami, Florida, USA. Association for Computational Linguistics.
Cite (Informal):
Fast Forwarding Low-Rank Training (Rahamim et al., EMNLP 2024)
Copy Citation:
PDF:
https://aclanthology.org/2024.emnlp-main.535.pdf