fully_correlated_conditional_repeat fvar broadcast_to issue

[johnamcleod]

Bug

Some parameter combinations for fully_correlated_conditional_repeat throw errors:
q_sqrt = None
full_cov and full_output_cov not both False.
throw an InvalidArgumentError.

To reproduce

Minimal, reproducible example

This is a slightly modified version of the unit test for fully_correlated_conditional_repeat.

import gpflow
import tensorflow as tf

Kmm = tf.ones((6, 6)) + gpflow.default_jitter() * tf.eye(6)
Kmn = tf.ones((6, 20, 5))
Knn = tf.ones((20, 5, 5))
f = tf.ones((6, 7))
q_sqrt = None
white = True

m, v = gpflow.conditionals.util.fully_correlated_conditional_repeat(
    Kmn, Kmm, Knn, f, full_cov=False, full_output_cov=True, q_sqrt=q_sqrt, white=white,
)

Stack trace, or error message

/usr/local/lib/python3.7/dist-packages/gpflow/conditionals/util.py in fully_correlated_conditional_repeat(Kmn, Kmm, Knn, f, full_cov, full_output_cov, q_sqrt, white)
    460             fvar = fvar[None, ...] + addvar  # [R, N, P]
    461     else:
--> 462         fvar = tf.broadcast_to(fvar[None], tf.shape(fmean))
    463 
    464     shape_constraints.extend(

/usr/local/lib/python3.7/dist-packages/tensorflow/python/ops/gen_array_ops.py in broadcast_to(input, shape, name)
    825       return _result
    826     except _core._NotOkStatusException as e:
--> 827       _ops.raise_from_not_ok_status(e, name)
    828     except _core._FallbackException:
    829       pass

/usr/local/lib/python3.7/dist-packages/tensorflow/python/framework/ops.py in raise_from_not_ok_status(e, name)
   6860   message = e.message + (" name: " + name if name is not None else "")
   6861   # pylint: disable=protected-access
-> 6862   six.raise_from(core._status_to_exception(e.code, message), None)
   6863   # pylint: enable=protected-access
   6864 

/usr/local/lib/python3.7/dist-packages/six.py in raise_from(value, from_value)

InvalidArgumentError: Rank of input (4) must be no greater than rank of output shape (3). [Op:BroadcastTo]

Expected behavior

Return the mean and variance.

System information

• GPflow version: 2.1.4
• GPflow installed from: git commit 405eb97
• TensorFlow version: 2.3.0
• Python version: 3.7.5
• Operating system: Ubuntu 18.04.5 LTS

Additional context