Change Depth of Torchvision Wide Resnet 50 Model

vision
1

I am using the wide_resnet50_2 model from torchvision.models. I want to change the depth of the model to 28, as the paper mentions various depths and performance levels per depth.

How would I do this using the wide_resnet50_2 model? Is there a simple configuration of the base _resnet class that is equivalent to wide resnet 50, of depth 28 from the original paper.

2

So I have seen this implementation of wideresnet in pytorch that has depth 28, but the architecture seems quite different to the one provided by torchvision.