Paraméterhatékony finomhangolási technikák alkalmazása nagy nyelvi modellekre
| Témavezető: | Csanády Bálint Zsombor |
| ELTE TTK Számítógéptudományi tanszék, MI kutatócsoport | |
| email: | csbalint@protonmail.ch |
Projekt leírás
LLMs such as the proprietary GPT-4 and the open-source Llama 2 present themselves as compelling solutions for large-scale data annotation in NLP. Indeed, minimal prompt-tuning enables them to be highly proficient in handling a wide variety of NLP tasks. However, running such LLMs on millions of prompts demands large and expensive computational resources. Previously, we focused on leveraging LLM’s language modeling capabilities on classification tasks involving millions of items, while utilizing relatively modest resources. This method, we called LamBERT, involved annotating a small subset of the corpus using Lama 2, and fine tuning a BERT model based on this annotation. The aim of the project is to assess PEFT techniques such as LoRA, prefix tuning, and P-tuning to potentially further increase the quality of data initially provided by the Lama 2 annotation.
Előfeltételek
Szükséges nyelvtudás: angol
Programozási ismeretek: Python
Hivatkozások
Touvron, H., Martin, L., Stone, K., Albert, P., Almahairi, A., Babaei, Y., Bashlykov, N., Batra, S., Bhargava, P., Bhosale, S., et al. Llama 2: Open foundation and fine-tuned chat models. arXiv preprint arXiv:2307.09288, 2023.
Pu, G., Jain, A., Yin, J., and Kaplan, R. Empirical analysis of the strengths and weaknesses of PEFT techniques for LLMs. arXiv preprint arXiv:2304.14999, 2023.
Hu, E. J., Shen, Y., Wallis, P., Allen-Zhu, Z., Li, Y., Wang, S., Wang, L., and Chen, W. LoRa: Low-rank adaptation of large language models. arXiv preprint arXiv:2106.09685, 2021.
Li, X. L. and Liang, P. Prefix-tuning: Optimizing continuous prompts for generation. arXiv preprint arXiv:2101.00190, 2021.
Liu, X., Ji, K., Fu, Y., Tam, W., Du, Z., Yang, Z., and Tang, J. P-tuning: Prompt tuning can be comparable to fine-tuning across scales and tasks. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), pages 61–68. 2022.