Covariance matrix in Multivariate Normal dist

Hello, I am new to ‘pyro’, and I am trying to implement SVI using GP. So I am trying to make use of
‘pyro.distributions.MultivariateNormal’ but I keep getting error when I specify the Squared Exponential covariance matrix by myself. I am not sure what I am missing. Attached picture is the code and the error I ran. the data I have is x: 50x1 and z: 50x1

I think you can check if sigma_eta is a positive definite matrix. Sometimes, you need to add a small epsilon at the diagonal part to make it “numerically” positive definite.