Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

"Only Keras classifiers (v2.2.4) are supported for this defence" in poisoning_defense_neural_cleanse.ipynd #764

Closed
authwork opened this issue Nov 27, 2020 · 5 comments
Labels
question Further information is requested

Comments

@authwork
Copy link

Dear all, how to guarantee I am using v2.2.4 when I run this defense test?
I try to install keras v2.2.4, but it does not work.
Many thanks.

@beat-buesser
Copy link
Collaborator

Hi @authwork Thank you very much for using ART!

@ebubae Do you think it would be possible to support other versions of Keras than 2.2.4?

@beat-buesser beat-buesser added the question Further information is requested label Nov 27, 2020
@ebubae
Copy link
Collaborator

ebubae commented Nov 27, 2020

@authwork to install Keras 2.2.4 you may be able to use pip install keras==2.2.4

We are definitely trying to support other versions of Keras. The current implementation of the defense had some issues with the optimizer that need to be investigated. @authwork which version of Keras are you using?

@authwork
Copy link
Author

authwork commented Nov 28, 2020

Many thanks for your help @ebubae @beat-buesser
Then, when I execure to graph = tf.get_default_graph()
It has the following problem (I have tried this before, that's why I said it does not work)

global _GRAPH_UID_DICTS
---> 74     graph = tf.get_default_graph()

AttributeError: module 'tensorflow' has no attribute 'get_default_graph'

Maybe there are some other requirements for it?

update:
I found this solution is to add "tf.compat.v1.get_default_graph" manually
But it cannot work in my case.

update:
I just remove the limitation for 2.2.4, and found it seems to work...

@ebubae
Copy link
Collaborator

ebubae commented Nov 30, 2020

@authwork great that you got it to work! Which version of Tensorflow and Keras are you using for the working version?

@authwork
Copy link
Author

authwork commented Dec 2, 2020

I used the most recent version of both and remove the limitation for 2.2.4 in ART.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
question Further information is requested
Projects
None yet
Development

No branches or pull requests

3 participants