[WIP]

This repo contains scripts and notebooks to showcase how to perform Supervised fine-tuning of LLMs with the Hugging Face ecosystem and AWS Sagemaker.

It is mostly inspired by the work presented in The Alignment Handbook and the Sagemaker Section of the Hugging Face notebooks repository.

• mistral-7B/: Examples notebooks to fine-tune Mistral-7B on an English conversational dataset on a machine with a single consumer GPU and on AWS Sagemaker instances

• german-model/: Scripts to fine-tune LeoLM-13b on a German conversational dataset

It's recommended to follow Installation instructions of the The Alignment Handbook repository