projected_gradient_descent() gives an error

[ananthceh19]

Describe the bug projected_gradient_descent() gives an error: "TypeError: Cannot convert 0.3 to EagerTensor of dtype uint8" when run on Google Colab.

To Reproduce Steps to reproduce the behavior: While running the following code (present in cleverhans/cleverhans/tutorials/tf2/mnist_tutorial.py: for epoch in range(FLAGS.nb_epochs):

keras like display of progress

    progress_bar_train = tf.keras.utils.Progbar(60000)
    for (x, y) in data.train:
        if FLAGS.adv_train:
            # Replace clean example with adversarial example for adversarial training
            print('x is ',x.shape)
            print('y is ',y.shape)
            print('Calling PGD')
           #print(eps)
           #import pdb; pdb.set_trace()
            x = projected_gradient_descent(model, x, 0.3, 0.01, 40, np.inf)
            #print(x)
        train_step(x, y)
        progress_bar_train.add(x.shape[0], values=[("loss", train_loss.result())])

........ projected_gradient_descent() gives the following error:

TypeError Traceback (most recent call last)

in () 36 #print(eps) 37 #import pdb; pdb.set_trace() ---> 38 x = projected_gradient_descent(model, x, 0.3, 0.01, 40, np.inf) 39 train_step(x, y) 40 progress_bar_train.add(x.shape[0], values=[("loss", train_loss.result())]) 4 frames /usr/local/lib/python3.7/dist-packages/tensorflow/python/ops/gen_math_ops.py in minimum(x, y, name) 5927 try: 5928 _result = pywrap_tfe.TFE_Py_FastPathExecute( -> 5929 _ctx, "Minimum", name, x, y) 5930 return _result 5931 except _core._NotOkStatusException as e: TypeError: Cannot convert 0.3 to EagerTensor of dtype uint8 **Expected behavior** A clear and concise description of what you expected to happen. eps cannot be of type uint8 and projected_gradient_descent method is not expected to give this error. **Screenshots** If applicable, add screenshots to help explain your problem. **System configuration** - OS - Linux e2205817a30a 4.19.112+ #1 SMP Thu Jul 23 08:00:38 PDT 2020 x86_64 x86_64 x86_64 GNU/Linux (Google colab notebook) - Python version - Python 3.7.10 - TensorFlow version - 2.4.1

[tejuafonja]

Hi, @ananthceh19 thanks for reporting this. Which version of cleverhans are you using, please?