Changing KERNEL_LENGTHSCALE from tf.cast to tf.constant results in the documentation build failing during serialisation (hitting an assert id(obj) not in self.memo in pickle.py while using dill). This does not happen in the quickdocs run and was not easy to reproduce in the integration tests.
Low priority but would be good to get to the bottom of this.
Changing KERNEL_LENGTHSCALE from tf.cast to tf.constant results in the documentation build failing during serialisation (hitting an
assert id(obj) not in self.memo
in pickle.py while using dill). This does not happen in the quickdocs run and was not easy to reproduce in the integration tests.Low priority but would be good to get to the bottom of this.