Is it possible to pass a trained model to the ProjectedGradientDescentPyTorch function?

[awskkzl]

So basically I have a trained model and I need to use PGD to generate adversarial samples for it.
I notice that in the demo you build an ART classifier and fit the data in the ART classifier, then pass the classifier to PGD method.
Is it possible to pass a loaded model to PGD directly? I tried but I got "module has no 'all_framework_preprocessing'"
Please help

[beat-buesser]

Hi @awskkzl Yes, you need to provide the trained model first to PyTorchClassifer, for example similar to examples/get_started_pytorch.py.