Thanks for the reply. I hadn’t thought of it. Maybe I should give a try to sunode, I guess just adding the parameters for the SDEs as priors may work. I’m trying with scan, but I tend to get confused when trying to recover the rngs. I couldn’t figure out how to use “pytensorf.collect_default_updates”. I’m looking at this thread: Simple markov chain with aesara scan - #10 by ricardoV94 . So maybe that’ll help to solve the issue.