I am trying to implement a custom variational autoencoder. Following is the code to reproduce.
epsilon_std = 1.0 vx = tf.keras.layers.Input(batch_shape=(None, ma