add_loss error with latest TF version

[Will-Wildmon]

Hi,

Many thanks for your work on this project, it's greatly appreciated.
Just to let you know, I am getting the following error when using this package with the latest version of TensorFlow:

---

ValueError Traceback (most recent call last)
in <cell line: 1>()

3 frames
701
702 # Adds the loss automatically
--> 703 self.add_loss(self.sparsity_coefficient * entropy_loss)
704
705 # Visualization of the aggregated feature importances

ValueError: Exception encountered when calling TabNet.call().

add_loss() can only be called from inside build() or call(), on a tensor input. Received invalid value: 0.0

Arguments received by TabNet.call():
• inputs=tf.Tensor(shape=(None, 14), dtype=float32)
• training=True