The Deep Kernel Learning Notebook from readthedeocs.io - How to incorporate a combination kernel instead of a single kernel in the existing codebase?

[GautamV234]

The following is the notebook being referred to in the doubt below: https://gpjax.readthedocs.io/en/latest/examples/haiku.html#Deep-Kernel-Learning

Hi all, My Doubt is regarding the utilization of a kernel combination such as jk.RBF() + jk.RBF() instead of a single kernel and keep the remaining architecture the same i.e. same neural network configuration in Haiku and the same way of incorporation of the linear weights in the parameter state. When I try to replace the jk.RBF() with the combination jk.RBF() + jk.RBF(), I get the following error in calculating the gradient of the MLL

It would be great if someone could please clarify how to go about using a combination kernel along with additional parameters such as the linear weights and biases in this case. Thanks a lot!