How can I terminate pm.sample when a convergence criterion is met (e.g. r_hat < 1.05)

Questions v5
2

Hey,

I have encountered the same issue on my model (using PyMC v4.2.0) and could reproduce the error you got.

I do not have a solution to the problem, but what I’ve found is that the error only appears when using multiple chains. If you need the callback, a workaround (not a solution) may be to set chains=1 and run the sampling process multiple times.