Should i do loss.backward() or loss.mean().backward()

[shirui-japina (Shirui Zhang)]

Your question is exactly the solution to my question.