AdversarialAttack icon indicating copy to clipboard operation
AdversarialAttack copied to clipboard

Some questions

Open Fansgithub2019 opened this issue 5 years ago • 0 comments

First of all, thank you for sharing. I tested the C & W attack and compared it with the C & W attack using tensorflow 1.x on cleverhans. It seems to be much slower on tensorflow2.0. Have you compared them? In addition, @ tf.function is not used when calculating gradients on attacker.py, why is this? Looking forward to your answer.

Fansgithub2019 avatar Mar 05 '20 12:03 Fansgithub2019