LoRA: Low-Rank Adaptation of Large Language Models

dc.contributor.authorHu, Edward J.
dc.contributor.authorShen, Yelong
dc.contributor.authorWallis, Phillip
dc.contributor.authorAllen‑Zhu, Zeyuan
dc.contributor.authorLi, Yuanzhi
dc.contributor.authorWang, Shean
dc.contributor.authorWang, Lu
dc.contributor.authorChen, Weizhu
dc.date.accessioned2025-06-02T13:28:10Z
dc.date.available2025-06-02T13:28:10Z
dc.date.issued2021-06-17
dc.description모델 파라미터를 고정한 상태에서 LoRA라는 저차원 학습 구조를 주입하여, 대규모 언어모델의 파인튜닝 효율성과 메모리 사용을 크게 개선한 방법을 제안합니다. 10,000배 적은 학습 파라미터와 3배 낮은 메모리 사용량을 보입니다 ©2021 Microsoft Research
dc.description.abstractAn important paradigm of natural language processing consists of large-scale pre-training on general domain data and adaptation to particular tasks or domains. As we pre-train larger models, full fine-tuning, which retrains all model parameters, becomes less feasible. Using GPT-3 175B as an example -- deploying independent instances of fine-tuned models, each with 175B parameters, is prohibitively expensive. We propose Low-Rank Adaptation, or LoRA, which freezes the pre-trained model weights and injects trainable rank decomposition matrices into each layer of the Transformer architecture, greatly reducing the number of trainable parameters for downstream tasks. Compared to GPT-3 175B fine-tuned with Adam, LoRA can reduce the number of trainable parameters by 10,000 times and the GPU memory requirement by 3 times. LoRA performs on-par or better than fine-tuning in model quality on RoBERTa, DeBERTa, GPT-2, and GPT-3, despite having fewer trainable parameters, a higher training throughput, and, unlike adapters, no additional inference latency. We also provide an empirical investigation into rank-deficiency in language model adaptation, which sheds light on the efficacy of LoRA. We release a package that facilitates the integration of LoRA with PyTorch models and provide our implementations and model checkpoints for RoBERTa, DeBERTa, and GPT-2 at this https URL.
dc.description.sponsorshipMicrosoft Research
dc.identifier.urihttps://arxiv.org/abs/2106.09685
dc.identifier.urihttp://data.inu.ac.kr/handle/123456789/1959
dc.language.isoen_US
dc.publisherarXiv
dc.subjectLoRA
dc.subjectLow-Rank Adaptation
dc.subjectEfficient Fine-Tuning
dc.subjectTransformer
dc.subjectNLP
dc.titleLoRA: Low-Rank Adaptation of Large Language Models
dc.typeArticle

Files

Original bundle
Now showing 1 - 1 of 1
No Thumbnail Available
Name:
2106.09685v2.pdf
Size:
1.53 MB
Format:
Adobe Portable Document Format
License bundle
Now showing 1 - 1 of 1
No Thumbnail Available
Name:
license.txt
Size:
97 B
Format:
Item-specific license agreed to upon submission
Description: