Remove finetunes from workflow

[Edwinexd]

I believe we could get better results with few-shot instead as we don't have a lot of samples being fed to the LLM, even less so with the new/upcoming sampling strategies.