Pretrained weights

[AWbosman]

Dear authors,

I have some networks that are pretrained and would like to retrain with your method.
In your read.me you state this would be added soon.
Do you have any tips or examples on how to use your repository?

Kindly,
Annelot

[nvedant07]

Hi @AWbosman ,

Thanks for checking out our repo! It's been a while since I worked on this, but I believe what you're looking for is the regularized loss that reduced robustness bias by adding regularization during training.

I added this file under code/regularized_loss.py. This contains a class called RegularizedLoss which you can use in your training loop just like you use nn.CrossEntropyLoss in PyTorch.

Let me know if you run into any issues!