In this guide, the author demonstrates how to fine-tune Llama 2 as a dialog summarizer. They start by explaining their intention to train Llama to generate a body from a given title using their own collection of Google Keep notes. The first part of the tutorial focuses on finetuning Llama 2 on the samsum dialog summarization dataset using Huggingface libraries. The author mentions that Huggingface libraries can sometimes be complicated for the average user. The second part, which is coming at the end of the week, covers fine-tuning Llama 2 on custom data. The author provides instructions on downloading the necessary model and converting it to Hugging Face format. They also provide steps on how to run the fine-tuning notebook and perform inference on the fine-tuned model. The author concludes by teasing future content about formatting custom datasets for training Llama 2.
https://brev.dev/blog/fine-tuning-llama-2