Htydjtk

I have defined the following function:

    def build(self):

      input_layer = Input(shape=(self.col_amt,))

      encoded = Dense(math.ceil(self.col_amt/2), activation='relu')(input_layer)

      decoded = Dense(self.col_amt, activation='sigmoid')(encoded)

      model = Model(input_layer, decoded)

      model.compile(optimizer='adadelta', loss='binary_crossentropy')

      return model

It's an encoder-decoder model using keras API.
self.col_amt is the number of features used.
While debugging, I get an error while i try to return the model (so on the last line).

The error is:

TypeError: __call__() missing 1 required positional argument: 'inputs'

Now here is the tricky/weird part. This exact function (i copy-pasted this code for testing purposes) is used in two child classes. When making one of the child class that contains the build() function i get the error. In the other one it runs fine. The function build() only depends on self.col_amt which is the same for the two child classes.

to conclude:

• Super class has self.col_amt and this is filled in


• Super class calls self.build() in the constructor


• When making a child class the error occurs in the last line
  of build(). 'return model'

I've been stuck on this error for quite some time as I can't seem to wrap my head around it. Any direction is helpfull.