Hi! Harper Carroll from Brev.dev here. In this tutorial video, I walk you through how to fine-tune Mistral 7B, which outperforms Llama 2 13B on all tested benchmarks, on your own data.... like how I do with my journal entries from over the years, teen angst and all.

Notebook: https://github.com/brevdev/notebooks/...

Notebook to fine-tune on a HF dataset: https://github.com/brevdev/notebooks/...

My explanation on how QLoRA works: https://brev.dev/blog/how-qlora-works

Join the Discord:   / discord  

Find me on 𝕏: https://x.com/HarperSCarroll