Hi @danchurch could you provide the model data and priors in one of those failing runs that lead to bad initial energy?
Since you are fitting your model in data that you know was generated like this it should be possible to have it sample okayish.
Hi @danchurch could you provide the model data and priors in one of those failing runs that lead to bad initial energy?
Since you are fitting your model in data that you know was generated like this it should be possible to have it sample okayish.