Train_adv still uses tensorflow v1.x

[noppelmax]

Hey,

it seems like Train_adv still uses tensorflow v1 stuff. For example

y_true = tf.placeholder("float", [None,1])

in function def get_adv_example(triple,epsilon,batch_size).

Kind regards,
Max

[noppelmax]

The error is

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-10-49b35365202a> in <module>
     14     targets_train = np.zeros([batch_size,1])
     15     for i in range(1, n_iter):
---> 16         inputs=get_two_batches(batch_size,num_targets)
     17         loss_value=model.train_on_batch(inputs,targets_train)
     18 

<ipython-input-8-d88b3fb0ff8e> in get_two_batches(batch_size, num_targets)
     86     half_batch = int(batch_size/2)
     87     triple1 = get_batch(half_batch,num_targets)
---> 88     triple2 = get_batch_adv(half_batch,num_targets)
     89 
     90     h = X_train_legit.shape[1]

<ipython-input-8-d88b3fb0ff8e> in get_batch_adv(batch_size, num_targets)
     78 
     79     epsilon = np.random.uniform(low=0.003, high=0.01)
---> 80     triple_noise,triple_adv = get_adv_example(triple,epsilon,batch_size)
     81     triple[0] = triple_adv
     82     return triple

<ipython-input-8-d88b3fb0ff8e> in get_adv_example(triple, epsilon, batch_size)
     32     anchor_noise = np.zeros_like(triple[0])
     33 
---> 34     y_true = tf.placeholder("float", [None,1])
     35     target = np.zeros([batch_size,1])
     36     target.astype(float)

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