python example_gan.py --> ModuleNotFoundError: No module named 'keras_adversarial.legacy'

[PredM]

Hello,

I have no clue why I get the following error when I am executing: python example_gan.py

Using TensorFlow backend.
Traceback (most recent call last):
  File "example_gan.py", line 16, in <module>
    from keras_adversarial.legacy import l1l2, Dense, fit
ModuleNotFoundError: No module named 'keras_adversarial.legacy'

I cloned your repo and executed setup.py. The file legacy.py is also in the right folder. Has anybody a suggestion what I can do to fix it?

Thanks in advance!

[mdbloice]

I got the same error as this.

The only way I was able to get this to work was to first make sure I didn't have the package installed with pip (I installed via pip rather than cloning the repository as I didn't read the README properly...), so the first thing I did was to uninstall the pip version:

pip uninstall keras_adversarial

Then, I installed it by cloning the repo and using setup.py as per the README.

However, when running python gan_example.py I then got the error:

Traceback (most recent call last):
  File "example_gan.py", line 18, in <module>
    from mnist_utils import mnist_data
ImportError: No module named mnist_utils

And I found that mnist_utils was actually just a small module with two functions in it:

def mnist_process(x):
    x = x.astype(np.float32) / 255.0
    return x

def mnist_data():
    (xtrain, ytrain), (xtest, ytest) = mnist.load_data()
    return mnist_process(xtrain), mnist_process(xtest)

See https://github.com/bstriner/keras-adversarial/blob/master/examples/mnist_utils.py.

So I commented out line 18 of example_gan.py and added an import for mnist from the keras.datasets module:

#from mnist_utils import mnist_data
from keras.datasets import mnist

and finally pasted in the two functions, mnist_process() and mnist_data() that I showed above, just below the imports.

That got it working. Maybe this helps someone.