search

cleverhans-lab / cleverhans

An adversarial example library for constructing attacks, building defenses, and benchmarking both
MIT License
6.15k stars 1.39k forks source link

JAX attacks depend on experimental features that does not exist #1232

Open rekino opened 1 year ago

rekino commented 1 year ago

The issue tracker should only be used to report bugs or feature requests. If you are looking for support from other library users, please ask a question on StackOverflow.

Describe the bug When I try to import the PGD attack from JAX module, I get the following error: ModuleNotFoundError: No module named 'jax.experimental.stax' The reason is that the FGSM implementation imports the logsoftmax function from the experimental package in jax/attacks/fast_gradient_method.py

To Reproduce Steps to reproduce the behavior:

  1. Open a google colab IPython notebook
  2. add the following code !pip install git+https://github.com/cleverhans-lab/cleverhans.git#egg=cleverhans to install cleverhans
  3. then try to import from cleverhans.jax.attacks.projected_gradient_descent import projected_gradient_descent
  4. execute and see the error

Expected behavior the logsoftmax function should be imported from jax.nn package. Change the import to from jax.nn import log_softmax as logsoftmax and the error would be gone.

Screenshots If applicable, add screenshots to help explain your problem.

System configuration