BUG: pymc5 posterior predictive sampling

I am learning Bayesian Analysis Chapter 7 and was trying to use pymc5 to reproduce some modelling examples.

When I tried to do posterior predictive sampling, I got an error message saying “The input matrix must be symmetric positive semidefinite”.

Attached is the file with printed codes of the model and the error messages. I also attached another file with the model with which I was able to do the posterior predictive sampling.

posterior_predictive_sampling_working.pdf (156.4 KB)
posterior_predictive_sampling_issue.pdf (1.9 MB)

The full error messages did give me some hints to fix this issue, but I do not fully understand them so not sure what steps I should take…

Could anyone suggest some solutions?

Thank you very much!