AI Large Language Models and Supervised Fine Tuning

This intermediate-level blog post explains supervised fine-tuning of large language models, contrasting full fine-tuning and transfer learning, reviewing parameter-efficient approaches (LoRA, QLoRA), and giving practical guidance on data preparation (Q/A pairs), prompt engineering, and tooling (Unsloth, OpenAI API, Hugging Face). It includes recommendations for running SFT on single-GPU systems, using another LLM to preprocess training data, and a brief Jupyter notebook example applying these techniques to Llama3.1 with the White Rabbit dataset.