I have a model that uses a black-box likelihood op. This likelihood is stateful, and I would love to be able to access its state at the end of sampling.
I can do this if MCMC sampling in only one process, but if using multiple processes and chains, I haven’t been able to figure out how to access the model or op in each chain once it is copied and sent to different processes.
Is there currently any way to do this in pymc?