커리큘럼 러닝 기반 LoRA 파인튜닝 성능 향상 방법론
Enhancing LoRA Fine-tuning Performance Using Curriculum Learning
0
조회
0
다운로드
1
인용
0
참고문헌
키워드
초록
최근 언어모델을 활용하기 위한 연구가 활발히 이루어지며, 큰 규모의 언어모델이 다양한 과제에서혁신적인 성과를 달성하고 있다. 하지만 실제 현장은 거대 언어모델 활용에 필요한 자원과 비용이한정적이라는 한계를 접하면서, 최근에는 주어진 자원 내에서 모델을 효과적으로 활용할 수 있는 방법에 주목하고 있다. 대표적으로 학습 데이터를 난이도에 따라 구분한 뒤 순차적으로 학습하는 방법론인커리큘럼 러닝이 주목받고 있지만, 난이도를 측정하는 방법이 복잡하거나 범용적이지 않다는 한계를지닌다. 따라서, 본 연구에서는 신뢰할 수 있는 사전 정보를 통해 데이터의 학습 난이도를 측정하고, 이를 다양한 과제에 쉽게 활용할 수 있는 데이터 이질성 기반 커리큘럼 러닝 방법론을 제안한다. 제안방법론의 성능 평가를 위해 국가 R&D 과제 전문 문서 중 정보통신 분야 전문 문서 5,000건, 보건의료전문 문서 데이터 4,917건을 적용하여 실험을 수행한 결과, 제안 방법론이 LoRA 미세조정과 전체 미세조정 모두에서 전통적인 미세조정에 비해 분류 정확도 측면에서 우수한 성능을 나타냄을 확인했다.
Abstract
Recently, there has been a lot of research on utilizing Language Models, and Large Language Models have achieved innovative results in various tasks. However, the practical application faces limitations due to the constrained resources and costs required to utilize Large Language Models. Consequently, there has been recent attention towards methods to effectively utilize models within given resources. Curriculum Learning, a methodology that categorizes training data according to difficulty and learns sequentially, has been attracting attention, but it has the limitation that the method of measuring difficulty is complex or not universal. Therefore, in this study, we propose a methodology based on data heterogeneity-based Curriculum Learning that measures the difficulty of data using reliable prior information and facilitates easy utilization across various tasks. To evaluate the performance of the proposed methodology, experiments were conducted using 5,000 specialized documents in the field of information communication technology and 4,917 documents in the field of healthcare. The results confirm that the proposed methodology outperforms traditional fine-tuning in terms of classification accuracy in both LoRA fine-tuning and full fine-tuning.
참고문헌 (40)
- X. Qiu, "Pre-trained Models for Natural Language Processing : A Survey", Science China Technological Sciences, 63(10), pp.1872~1897 (2020)[DOI]
- J. Kirkpatrick, "Overcoming Catastrophic Forgetting in Neural Networks", Proceedings of the National Academy of Sciences, 114(13), pp.3521~3526 (2017)[DOI]
- Y. Bengio, "Curriculum learning", Proceedings of the 26th Annual International Conference on Machine Learning (2009), ACM
- M. Kumar, "Self-Paced Learning for Latent Variable Models", Advances in neural information processing systems (2010)
- L. Jiang, "Self-Paced Curriculum Learning", Proceedings of the AAAI Conference on Artificial Intelligence, 29(1) (2015), Association for the Advancement of Artificial Intelligence (AAAI)[DOI]
- A. Vaswani, "Attention Is All You Need", Advances in Neural Information Processing Systems, 30 (2017)
- A. Radford, "Improving Language Understanding by Generative Pre-Training", ~ (2018)
- A. Radford, "Language Models are Unsupervised Multitask Learners", OpenAI blog, 1(8), pp.9~ (2019)
- Y. Xian, "Zero-Shot Learning—A Comprehensive Evaluation of the Good, the Bad and the Ugly", IEEE Transactions on Pattern Analysis and Machine Intelligence, 41(9), pp.2251~2265 (2019), Institute of Electrical and Electronics Engineers (IEEE)[DOI]
- Y. Wang, "Generalizing from a Few Examples", ACM Computing Surveys, 53(3), pp.1~34 (2020)[DOI]
- "Alpaca: A Strong, Replicable Instruction-Following Model"[URL]
- "Vicuna: An Open-Source Chatbot Impressing GPT-4 with 90%* ChatGPT Quality"[URL]
- "KoLLaMA"[URL]
- "KoAlpaca"[URL]
- "KoGPT"[URL]
- "Mi:dm"[URL]
- X. Wang, "A Survey on Curriculum Learning", IEEE Transactions on Pattern Analysis and Machine Intelligence, pp.4555~4576 (2021), Institute of Electrical and Electronics Engineers (IEEE)[DOI]
- "KR-SBERT"[URL]
- E. Yu, "Building Specialized Language Model for National R&D through Knowledge Transfer Based on Further Pre-training", Knowledge Management Research, 22(3), pp.91~106 (2021)