GPU only uses 150/350 watts while training a model. Increasing the batch size only increases the use of VRAM. Is there a way to use full potential of my GPU?
