CIFAR-100 training on Resnet-18 downloaded from torchvision.models

vision
3

Sadly no!

However, the difference arises from the reduced number of parameters in the locally defined models by these repositories.