I've been running some fairly expensive models with a long burn-in time. I'd like to save & restore the state of the AdaptiveMetropolis step methods (at least) to reduce the burn-in time. Is there a recommended way to do this, or should I just pickle the step_method.current_state() dict and try to update step_method.C, step_method.proposal_sd, and maybe others from pickled data? If there's example code hiding in the docs, I'd appreciate being pointed in the right direction.
Thanks,
Adam