I have an ELBO loss that looks as follows:
My first question is whether this loss optimization looks reasonable?
Secondly, is there some mechanism within pyro to detect convergence and to query the optimum parameters and the associated loss with it?