We are currently storing all likelihoods and all parameters for every temperature rung. I've found that when running MCMCs with large numbers of parameters this makes the output object very large, when we rarely (if ever) need to see the parameter values for hot rungs. We do, however, need to see the likelihood and prior for all rungs. So I think we need to split the output so that likelihood & prior are stored in one object over all rungs, and parameters are stored in another object just for the cold rung.
We are currently storing all likelihoods and all parameters for every temperature rung. I've found that when running MCMCs with large numbers of parameters this makes the output object very large, when we rarely (if ever) need to see the parameter values for hot rungs. We do, however, need to see the likelihood and prior for all rungs. So I think we need to split the output so that likelihood & prior are stored in one object over all rungs, and parameters are stored in another object just for the cold rung.