TypeError: The added layer must be an instance of class Layer

[tolandwehr]

Wanted to add SN layers as described, added SpectralNormalizationKeras.py to the respective dir. However, the layer could not be integrated. Here the respective parts of code

from SpectralNormalizationKeras import DenseSN, ConvSN2D

(...)

 def build_critic(self, spectral_normalization=True):

        model = Sequential()

        model.add(ConvSN2D(16, kernel_size=3, strides=2,kernel_initializer='glorot_uniform', input_shape=self.img_shape, padding="same"))
        model.add(LeakyReLU(alpha=0.2))
        model.add(Dropout(0.25))
        model.add(ConvSN2D(32, kernel_size=3, strides=2,kernel_initializer='glorot_uniform', padding="same"))
        model.add(ZeroPadding2D(padding=((0,1),(0,1))))
        model.add(BatchNormalization(momentum=0.8))
        model.add(LeakyReLU(alpha=0.2))
        model.add(Dropout(0.25))
        model.add(ConvSN2D(64, kernel_size=3, strides=2,kernel_initializer='glorot_uniform', padding="same"))
        model.add(BatchNormalization(momentum=0.8))
        model.add(LeakyReLU(alpha=0.2))
        model.add(Dropout(0.25))
        model.add(ConvSN2D(128, kernel_size=3, strides=1,kernel_initializer='glorot_uniform',padding="same"))
        model.add(BatchNormalization(momentum=0.8))
        model.add(LeakyReLU(alpha=0.2))
        model.add(Dropout(0.25))
        model.add(Flatten())
        model.add(DenseSN(1,kernel_initializer='glorot_uniform'))

        model.summary()

        img = Input(shape=self.img_shape)
        validity = model(img)

        return Model(img, validity)

Here the error call:

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-24-da7bad03b7a7> in <module>
      1 if __name__ == '__main__':
----> 2     wgan = WGANGP()
      3     wgan.train(epochs=30001, batch_size=256, sample_interval=1500)

<ipython-input-23-9f58d066c64d> in __init__(self)
     27         # Build the generator and critic
     28         self.generator = self.build_generator()
---> 29         self.critic = self.build_critic()
     30 
     31         #-------------------------------

<ipython-input-23-9f58d066c64d> in build_critic(self, spectral_normalization)
    141         model = Sequential()
    142 
--> 143         model.add(ConvSN2D(16, kernel_size=3, strides=2,kernel_initializer='glorot_uniform', input_shape=self.img_shape, padding="same"))
    144         model.add(LeakyReLU(alpha=0.2))
    145         model.add(Dropout(0.25))

~\Anaconda3\envs\Tensorflow\lib\site-packages\tensorflow\python\keras\engine\sequential.py in add(self, layer)
    126       raise TypeError('The added layer must be '
    127                       'an instance of class Layer. '
--> 128                       'Found: ' + str(layer))
    129     self.built = False
    130     if not self._layers:

TypeError: The added layer must be an instance of class Layer. Found: <SpectralNormalizationKeras.ConvSN2D object at 0x000001BF340526D8>

Lots of thanks in advance for any suggestion of how to overcome this.