Multiple Classes fails in Eager Mode ("tf.keras.Model")

[sml0820]

System information

• Have I written custom code (as opposed to using a stock example script provided in TensorFlow):
  No
• Bazel version:
  N/A
• CUDA/cuDNN version:
  N/A
• GPU model and memory:
  N/A
• OS Platform and Distribution (e.g., Linux Ubuntu 16.04):
  Tried on MacOS using tensorflow as well as Linux Ubuntu 16.04 using tensorflow-gpu
• TensorFlow installed from (source or binary):
  Installed utilizing pip
• TensorFlow version (use command below):
  1.7
• Python version:
  3.6
• Exact command to reproduce:

import tensorflow as tf  
import tensorflow.contrib.eager as tfe  

tfe.enable_eager_execution()

class CustomLayer(tf.keras.Model):
    def __init__(self):
        super(CustomLayer, self).__init__()
        print("blah")

class CustomNetwork(tf.keras.Model):
    def __init__(self):
        super(CustomNetwork, self).__init__()
        self.custom_layers = CustomLayer()

    def forward(self, x, y=None):
        x = self.custom_layers(x)

CustomNetwork().forward(tf.convert_to_tensor([1]))

Describe the problem

Trying to utilize multiple classes fails in tensorflow eager mode utilizing "tf.keras.Model". If I change "tf.keras.Model" to "tfe.Network" it works - keep in mind I am utilizing tensorflow 1.7. The error I get running the above code results in the error below:

Source code / logs

blah
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-12-9afa9b91ddef> in <module>()
----> 1 CustomNetwork().forward(tf.convert_to_tensor([1]))

<ipython-input-11-484119102aec> in forward(self, x, y)
      5 
      6     def forward(self, x, y=None):
----> 7         x = self.custom_layers(x)

~/anaconda3/lib/python3.6/site-packages/tensorflow/python/keras/_impl/keras/engine/base_layer.py in __call__(self, inputs, **kwargs)
    237     """
    238     # Actually call the layer (optionally building it).
--> 239     output = super(Layer, self).__call__(inputs, **kwargs)
    240     if context.executing_eagerly():
    241       return output

~/anaconda3/lib/python3.6/site-packages/tensorflow/python/layers/base.py in __call__(self, inputs, *args, **kwargs)
    712 
    713         if not in_deferred_mode:
--> 714           outputs = self.call(inputs, *args, **kwargs)
    715           if outputs is None:
    716             raise ValueError('A layer\'s `call` method should return a Tensor '

~/anaconda3/lib/python3.6/site-packages/tensorflow/python/keras/_impl/keras/engine/network.py in call(self, inputs, training, mask)
    635     outputs, _ = self._run_internal_graph(inputs,
    636                                           training=training,
--> 637                                           mask=masks)
    638     return outputs
    639 

~/anaconda3/lib/python3.6/site-packages/tensorflow/python/keras/_impl/keras/engine/network.py in _run_internal_graph(self, inputs, training, mask)
    770     # does not return a list the same size as `call`
    771     tensor_map = {}
--> 772     for x, y, mask in zip(self.inputs, inputs, masks):
    773       tensor_map[str(id(x))] = (y, mask)
    774 

TypeError: zip argument #1 must support iteration

[tensorflowbutler]

Thank you for your post. We noticed you have not filled out the following field in the issue template. Could you update them if they are relevant in your case, or leave them as N/A? Thanks.
Bazel version
CUDA/cuDNN version
GPU model and memory

[sml0820]

Believe I need to use .call instead of .forward