An Updated version of Medusa 1.1. It's a little more coherent, better at following instructions than 1.1. Still experimental.

This is a merging of Airoboros and Hermes, the final model is merged with LimaRP v2 and an Experimental LoRA I'm working on, both are weighed at different gradient ratios.

Quants here: https://huggingface.co/Sao10K/Medusa-1.2-L2-7B-GGUF

Support me here :)