I seriously have no idea how much vram it takes to finetune or train a model in a way that makes it useful. Like training a functiongemma or similar for a certain usecase. Imagine I would want to finetune it to react to sensor readings. I know I could get any amount of vram from vast.ai, but I wonde
