Thanks! Its finally working.
I actually tried running the same code using pm.gp.GP. However, I got the following error:
ValueError: cov_func must be a subclass of Covariance. The only change was using pm.gp.GP instead of pm.MvNormal. I am trying to use pm.gp.GP instead of pm.MvNormal because it would be a lot easier to use pm.gp.sample_gp to draw realizations of the GP from the predictive distribution. Anyone has any recommendations or workarounds?