Split inference of calib vs valid

Running BayesInference.create_inference with name='valid' vs 'calib' leads to two different use cases. Split the code to reflect this?