I have a model that I am sampling from as follows: with m.model: m.trace = pm.sample(cores=1, chains=1) This works fine (well, there are divergences, so not fine, but it works). But when I remove the cores=1, chains=1, I get a bad initial energy error: Multiprocess sampling (4 chains…

NUTS sampler gets "bad initial energy" only with parallel sampling

rpgoldman May 28, 2019, 8:48pm 5

This seems to be a problem with the search and my keyword guessing. Here’s the key post.

Topic		Replies	Views
SamplingError: Bad initial energy Questions	1	2789	September 10, 2020
Getting 'Bad initial energy: inf' when trying to sample simple model Questions	4	13950	May 21, 2018
Sample Binomial module initial error Questions theano	6	1530	November 20, 2022
Bad Initial Energy Questions	9	1479	November 17, 2021
Bad Initial Energy when observed var contains nan Questions	4	1040	September 23, 2019