CelverHans PGD model was not running

TensorFlow
,
1

Dear Fellow,

I have tried cleverhans FGSM that was ran successfully after that I tried PGD but it gives these king of error, I tried s many things but nothing happened. Does anyone have any solution

epsilon = 0.1

adv_example_untargeted_label = fast_gradient_method(logits_model, original_image, epsilon,40,8,5,np.inf, targeted=False)

adv_example_untargeted_label_pred = model.predict(adv_example_untargeted_label)

AssertionError Traceback (most recent call last)
in
1 epsilon = 0.1
2
----> 3 adv_example_untargeted_label = fast_gradient_method(logits_model, original_image, epsilon,40,8,5,np.inf, targeted=False)
4
5 adv_example_untargeted_label_pred = model.predict(adv_example_untargeted_label)

/usr/local/lib/python3.7/dist-packages/cleverhans/tf2/attacks/projected_gradient_descent.py in projected_gradient_descent(model_fn, x, eps, eps_iter, nb_iter, norm, loss_fn, clip_min, clip_max, y, targeted, rand_init, rand_minmax, sanity_checks)
58 “”"
59
—> 60 assert eps_iter <= eps, (eps_iter, eps)
61 if norm == 1:
62 raise NotImplementedError(

AssertionError: (40, 0.1)

3

Hi @Talha_Naeem

Could you please provide some more details on the issue along with the reproducible code to replicate the error?

Please refer the code example available in TensorFlow tutorial named ‘Adversarial example using FGSM’ which might be helpful to you in this issue. Thank you