Navigating The Metric Zoo: Towards A More Coherent Model For Quantitative Evaluation Of Generative Ml Models