can we train more LLMs on this and also the codebase of handmade hero?