First attempt at Fine Tuning a LLM, by creating a LoRA for the LLM.
Model used was Mistral-7B-Instruct-v0.1.
Web UI used was Gradio.
Dataset was obtained from Gov.sg websites and summarised with LLMs and verified by a human.
Web UI has a long loading time and generation time due to some difficulties in making it work due to version compatability between different libraries.
Project was heavily guided and built using LLMs such as ChatGPT and Claude.
