Summary:

It'd be nice to have a builder pattern for var contexts to make them easy to construct for testing. Something that could be used like this:

MatrixXd m(3, 2); 
...
var_context vc
  = var_context::builder()
    .matrix("a", m)
    .real("f", 2.3)
    .build();

Current Version:

v2.23.0

Feature request

Our Scipy optimizer wrapper should explicitly handle unconnected gradients.

Motivation

Currently, the GPflow Scipy optimizer wrapper (gpflow.optimizers.Scipy) simply errors (with one of the typically unhelpful TensorFlow error messages) when one of the passed-in variables is not connected to the loss to be minimized. This can easily happen in a valid case, for exa