In the network architecture of discriminator:
tf.keras.layers.Dense(units=1, activation="sigmoid"),
You don't need any activation function for discriminator, since you are using Wasserstein Loss. Using sigmoid here would greatly limit your convergeance speed.
In the network architecture of discriminator:
tf.keras.layers.Dense(units=1, activation="sigmoid"),You don't need any activation function for discriminator, since you are using Wasserstein Loss. Using sigmoid here would greatly limit your convergeance speed.