predict_f() outputs multidimensional tensor when a scalar value is expected?

Question

Hello GPflow community!

I\'ve just started using the library recently and I\'m working on a basic GPR model with 24 training examples, where x_train has shape (24, 36