Development of an Eye Care Chatbot Based on Llama 3.2 1B Using 4-bit QLoRA Technique
Article Sidebar
Abstract:
Background: The ratio of ophthalmologists to the population in Indonesia is extremely low, leading to unequal access to eye care and making it difficult for the public to obtain early information regarding eye symptoms.
Aims: This study aims to develop and validate an efficient, domain-specific eye care chatbot using the Llama 3.2 1B Instruct model, focusing on an optimal fine-tuning pipeline for limited hardware constraints.
Methods: Utilizing an Experimental Research (Model Development) design, a QLoRA 4-bit quantization technique was applied on a Google Colab T4 GPU (15.64GB VRAM). The training dataset comprised 16,742 samples from Kaggle Eye Care and MedQuad Indonesian Translation, with a subset of 5,000 samples used for experiments.
Result: After training for 3 epochs (750 steps), the Training Loss decreased from 1.3394 to 0.7188 (46.3% improvement), while Perplexity reached 2.3620, categorized as excellent. The model maintained stability with a final gap of 0.1407 between training and validation loss.
Conclusion: The Meta-Llama 3.2 1B Instruct model, fine-tuned with 4-bit QLoRA, is highly effective for building a domain-specific healthcare chatbot, successfully operating within an 8GB VRAM limit while maintaining high generalization capabilities.
Keywords: Chatbot, Eye Care, Fine-tuning, Llama 3.2, QLoRA
Downloads
Copyright (c) 2026 Teguh Rijanandi, Eko Risdianto
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.
References
Alshahrani, M., et al. (2025). Efficient Medical Question Answering Through QLoRA Fine-Tuning. Procedia Computer Science, S1877-0509(25)02941-2. https://www.sciencedirect.com/science/article/pii/S1877050925029412
ASEAN Secretariat. (2025). Expanded ASEAN Guide on AI Governance and Ethics – Generative AI. https://asean.org/wp-content/uploads/2025/01/Expanded-ASEAN-Guide-on-AI-Governance-and-Ethics-Generative-AI.pdf
Biswas, S., Davies, L. N., Sheppard, A. L., Logan, N. S., & Wolffsohn, J. S. (2024). Utility of artificial intelligence‐based large language models in ophthalmic care. Ophthalmic and Physiological Optics, 44(3), 641–671. https://doi.org/10.1111/opo.13284
Brenndoerfer, M. (2025). QLoRA: Efficient Fine-Tuning of Quantized Language Models. MBrenndoerfer’s Blog. https://mbrenndoerfer.com/writing/qlora-efficient-finetuning-quantized-language-models
Chen, K., et al. (2025). Lightweight Clinical Decision Support System using QLoRA-Fine Tuned LLM. arXiv preprint arXiv:2505.03406. https://arxiv.org/abs/2505.03406
Debashis, D. (2025). Fine-Tuning Healthcare LLMs with QLoRA: A Complete Guide. Medium. https://medium.com/@debashis2007/fine-tuning-healthcare-llms-with-qlora-a-complete-guide-016d724018ba
Dettmers, T., Pagnoni, A., Holtzman, A., & Zettlemoyer, L. (2023). QLoRA: Efficient Finetuning of Quantized LLMs. Advances in Neural Information Processing Systems, 36, 1-20. https://arxiv.org/abs/2305.14314
Digital Divide Data. (2025). Gen AI Fine-Tuning Techniques: LoRA, QLoRA, And Adapters. https://www.digitaldividedata.com/blog/ai-fine-tuning-techniques-lora-qlora-and-adapters
Gao, Y., et al. (2024). Roles, Users, Benefits, and Limitations of Chatbots in Health Care: Scoping Review. Journal of Medical Internet Research, 26(1), e56930. https://www.jmir.org/2024/1/e56930/
HomeDOCtor Research Team. (2025). Evaluating a Nationally Localized AI Chatbot for Personalized Healthcare Assistance. Healthcare, 13(15), 1843. https://www.mdpi.com/2227-9032/13/15/1843
Hugging Face. (2024). PEFT: State-of-the-art Parameter-Efficient Fine-Tuning methods. Retrieved from https://huggingface.co/docs/peft
Kwon, H., et al. (2025). For clinical data extraction, QLoRA attains accuracy close to LoRA while requiring lower compute resources. medRxiv. https://www.medrxiv.org/content/10.1101/2025.10.21.25338506v1
Liu, Y., et al. (2023). Parameter-Efficient Fine-Tuning of LLaMA for the Clinical Domain. arXiv preprint arXiv:2307.03042. https://arxiv.org/abs/2307.03042
Meta AI. (2024). Introducing Llama 3.2: Edge-optimized, capable models for mobile and edge devices. Meta Technology Research Lab. https://ai.meta.com/blog/llama-3-2-connect-2024
Mulyana, A., et al. (2025). Resource-Efficient Fine-Tuning of LLaMA-3.2-3B for Medical Chain-of-Thought Reasoning. arXiv preprint arXiv:2510.05003. https://arxiv.org/abs/2510.05003
Pal, A., et al. (2025). Doc Bot: The Medical LLM Fine-tuned on LLaMA 3 8B Using LoRA. Research Square.https://assets-eu.researchsquare.com/files/rs-7515191/v1_covered_920c2c43-f9ee-472a-87cd-4bf968ca1e4a.pdf
Putri, A. A. D., Alberta, I. B., & Ciputra, F. (2023). Artificial Intelligence in Ophthalmology: Challenges and Readiness in Indonesia. Al-Iqra Medical Journal Jurnal Berkala Ilmiah Kedokteran, 6(2), 24–32. https://doi.org/10.26618/aimj.v6i2.10951
Qiu, P., Wu, C., Zhang, X., Lin, W., Wang, H., Zhang, Y., Wang, Y., & Xie, W. (2024). Towards building multilingual language model for medicine. Nature Communications, 15(1), 8384. https://doi.org/10.1038/s41467-024-52417-z
Rahimi, J. H., Lau, J. H., & Baldwin, T. (2025). Medical Named Entity Recognition from Indonesian Health-News Using IndoBERT. Journal of Applied Informatics and Computing, 11(574). https://jurnal.polibatam.ac.id/index.php/JAIC/article/view/11574
Rijanandi, T. (2026). MedQuad Indonesian Translation. Kaggle. https://www.kaggle.com/datasets/teguh02/medquad-indonesian-translation
Samia, S. (2025). Parameter-Efficient Fine-Tuning: The Evolution of LLM Adaptation. Medium. https://medium.com/@sahin.samia/parameter-efficient-fine-tuning-the-evolution-of-llm-adaptation-63b01d544483
Springer Nature. (2025). Impact of large language model (ChatGPT) in healthcare: an umbrella review and evidence synthesis. Journal of Translational Medicine, 23(1), 1-15. https://link.springer.com/article/10.1186/s12929-025-01131-z
Wahyudi, A., et al. (2025). A Dental Chatbot Based on IndoBERT with Next Sentence Prediction for Indonesian Healthcare. Brilliance: Research of Artificial Intelligence, 6(6620). https://jurnal.itscience.org/index.php/brilliance/article/view/6620
Wang, S., et al. (2025). Parameter-efficient fine-tuning for low-resource text classification. Frontiers in Big Data, 8, 1677332. https://www.frontiersin.org/journals/big-data/articles/10.3389/fdata.2025.1677331/full
Yang, L. W. Y., Ng, W. Y., Foo, L. L., Liu, Y., Yan, M., Lei, X., Xiao-man, Z., & Ting, D. S. W. (2021). Deep learning-based natural language processing in ophthalmology: applications, challenges and future directions. Current Opinion in Ophthalmology, 32(5), 397–405. https://doi.org/10.1097/icu.0000000000000789
