LLM Fine-tune Collection

This project involves fine-tuning Llama3 8b to generate JSON formats for arithmetic questions and further post-process the output to perform calculations. This method incorporates the latest fine-tuning techniques such as Qlora, Unsloth, and PEFT. It enables faster training speeds and requires fewer computational resources.