Is resnet56 getting overfit to achieve 93% accuracy? Could you please provide accuracy and loss graph for resnet56?

akamaster / pytorch_resnet_cifar10

Proper implementation of ResNet-s for CIFAR10/100 in pytorch that matches description of the original paper.

BSD 2-Clause "Simplified" License

1.22k stars 335 forks source link

Is resnet56 getting overfit to achieve 93% accuracy? Could you please provide accuracy and loss graph for resnet56? #26

Closed tomshalini closed 3 years ago

tomshalini commented 3 years ago

When I trained resnet 56 on cifar10 and plotted the graphs. It's getting overfit as shown below.

akamaster commented 3 years ago

I have about the same plots. The phenomenon you are observing is common in deep nets, which actually is a topic of interest. See for example https://openreview.net/pdf?id=Sy8gdB9xx