When generating training data on specific systems the MLIP/NNP community needs to get organized and agree on one particular level of theory, settings and data storage standards as much as possible so we can pool all the data for training foundation/universal models right?
