Model scoring in pymc

Have you seen the Out-Of-Sample Predictions notebook? If not, I would probably suggest starting there.