Model accuracy calculator

Hi, @thunder, I don’t think torch.max is suitable to your case.
I think torch.argmax is a good choice.