AttributeError: 'NoneType' object has no attribute 'original_name_scope'

So, I'm trying to apply some flipout layers to a deep-q network. However, Keras seems to throw the above error when I run the DQN. How can I resolve this?

Full Traceback:

Traceback (most recent call last):
  File "thoughtmouse.py", line 57, in <module>
    combined()
  File "thoughtmouse.py", line 18, in combined
    state_action_reward_loop(agent)
  File "thoughtmouse.py", line 30, in state_action_reward_loop
    rewards = agent.replay(1)
  File "/home/ai/Downloads/ScreenMouse/bdqn.py", line 88, in replay
    target = (reward + self.gamma * np.amax(self.model.predict(next_state)[0]))
  File "/home/ai/anaconda3/envs/drl/lib/python3.6/site-packages/tensorflow/python/keras/engine/training.py", line 1096, in predict
    x, check_steps=True, steps_name='steps', steps=steps)
  File "/home/ai/anaconda3/envs/drl/lib/python3.6/site-packages/tensorflow/python/keras/engine/training.py", line 2289, in _standardize_user_data
    self._set_inputs(cast_inputs)
  File "/home/ai/anaconda3/envs/drl/lib/python3.6/site-packages/tensorflow/python/training/checkpointable/base.py", line 442, in _method_wrapper
    method(self, *args, **kwargs)
  File "/home/ai/anaconda3/envs/drl/lib/python3.6/site-packages/tensorflow/python/keras/engine/training.py", line 2529, in _set_inputs
    outputs = self.call(inputs, training=training)
  File "/home/ai/anaconda3/envs/drl/lib/python3.6/site-packages/tensorflow/python/keras/engine/sequential.py", line 233, in call
    inputs, training=training, mask=mask)
  File "/home/ai/anaconda3/envs/drl/lib/python3.6/site-packages/tensorflow/python/keras/engine/sequential.py", line 253, in _call_and_compute_mask
    with ops.name_scope(layer._name_scope()):
  File "/home/ai/anaconda3/envs/drl/lib/python3.6/site-packages/tensorflow/python/layers/base.py", line 284, in _name_scope
    return self._current_scope.original_name_scope
AttributeError: 'NoneType' object has no attribute 'original_name_scope'

Code of DQN:

class DQN:
    def __init__(self, state_size, action_size):
        self.state_size = state_size
        self.action_size = action_size
        self.memory = deque(maxlen=2000)
        self.gamma = 0.95    # discount rate
        self.epsilon = 0.81  # exploration rate
        self.epsilon_min = 0.2
        self.epsilon_decay = 0.965
        self.lr = 0.001
        self.model = self._build_model()
        self.optimizer = Adam(lr=self.lr)
    def _build_model(self):
        model = Sequential()
##        model.add(tfp.layers.Convolution2DFlipout(2, kernel_size=5,
##                            activation='relu'
##                        ))
        model.add(tf.layers.Dropout(rate=0.3))
        model.add(Flatten())
        #model.add(tfp.layers.DenseFlipout(self.action_size * 4, activation='relu'))
        #model.add(tfp.layers.DenseFlipout(self.action_size * 3, activation='relu'))
        model.add(tfp.layers.DenseFlipout(self.action_size * 2, activation='relu'))
        model.add(tfp.layers.DenseFlipout(self.action_size))

        return model
 
    def remember(self, state, action, reward, next_state, done):
        self.memory.append((state, action, reward, next_state, done))

    def act(self, state):
       if np.random.rand() <= self.epsilon:
            return random.randrange(self.action_size)
       else:
            act_values = self.model.predict(state)
            return np.argmax(act_values[0])  # returns action

    def replay(self, batch_size):
        minibatch = random.sample(self.memory, batch_size)
        for state, action, reward, next_state, done in minibatch:
            target = reward
            if not done:
                target = (reward + self.gamma * np.amax(self.model.predict(next_state)[0]))
            target_f = self.model.predict(state)
            target_f[0][action] = target
            self.history = self.model.fit(state, target_f, epochs=1, verbose=0)
        if self.epsilon > self.epsilon_min:
            self.epsilon *= self.epsilon_decay
        loss = self.history.history['loss'][0]
        return loss

    def load(self, name):
        self.model.load_weights(name)

    def save(self, name):
        self.model.save_weights(name)

    def initialize(self):
        self.model.compile(loss='mean_squared_logarithmic_error', optimizer = self.optimizer)

Start with DQN._build_model in bdqn.py and the model.predict calls in replay, then reproduce the traceback through thoughtmouse.py. Inspect how the Sequential model combines tf.layers.Dropout and TensorFlow Probability Flipout layers. Done means the DQN can run predict and replay without the original_name_scope AttributeError.