place of softmax function

[sdjsngs]

Thanks for your work， but i have noticed that in main.py you use the softmax function after the loss.backward(), this is a unusual
operation， in a classification task , the softmax function often between the fc layer ( in model class ) and loss.bacward() ,
Sincerely， hope you to replay
thank you

[doge-ac-cn]

I think it's right.
metric(prediction, label.long()) is torch.nn.CrossEntropyLoss() , it contains softmax.
And the second softmax is just to calculate the average accuracy of all the slices .