When I am tring to implement latent variable GP using Numpyro, I usually use a non-center parameterization in which the cholesky deompostion of the cov matrix is multiplied with a vector of standard normal variables. But I am wondering if I can use the reparam handler in numpyro to automatically reparamterize multivairate normal distribution? I see the doc of multivairate normal dist where there is a list of reparamized parameters, But I don’t know how to implement that.

Thank you