Supporting Data for "More converged, less accurate? Reassessing standard choices for ab initio water using machine learning potentials"

The repository contains the input files for the electronic structure methods, the training datasets and the committee models and is sorted by the electronic structure method. The names of the directory follow the format "<method>_<basis set>_<potential>". In directory can be found: A file named "force_eval.inp", containing the FORCE_EVAL section of the CP2K input file. A file name "training_set.data", containing the full training data used to train the committee model in RuNNer-style file format. Note that atomic units are used. A directory named "committee-model", with 8 subdirectories containing the N2P2 input file and weights for one committee member each.