What recipe did you use to train the wide_resnet50_2 and wide_resnet101_2

[mjack3]

Hello, thank you for your help.

I would like to know if the pretrained weights for wide_resnet50_2 and wide_resnet101_2 of this repository are using the recipe that you describe in your paper "Resnet Strikes Back".

If not, could you refer to me where could I find the information?

I am involved in a project where I get better performance using your wide_resnet50_2 despite the fact that torchvision=0.12.0 has a wide_resnet50_2 with better performance (imageNet). I am trying to understand why :)

Thanks