You could even argue it's not so much the models themselves (LLM or diffusion), but rather the fact that fine tuning is remarkably accessible. Research is hot o

You could even argue it's not so much the models themselves (LLM or diffusion), but rather the fact that fine tuning is remarkably accessible. Research is hot on ways to quantize parameters, swap portions for low memory, and the like. It's amazing to watch and use them.