Anybody here fine-tuning LLMs? I’m really curious about data ingestion for fine-tuning small LLMs -> phi-2, qlora-mistral Where are you looking at when it comes to dataset preparation?

maybe this helps? https://github.com/OpenAccess-AI-Collective/axolotl

What are you trying to achieve from the fine-tuning @Mert Bozkır?

I’m gonna setup autogen and Mixture of Experts to get neat results with small models! How does it sound to you?

Interesting. But I think models should work out of the box for this. cc/ @Mert Bozkır