Puffin 7B, everythingLM, ShareGPT small trained for LoRA.
Everything is using sharegpt format