Comment by yu3zhou4

Same! I believe that you could actually train a LoRA on top of a model to get results close to that