A set of convenience functions for NumPy
npu.io.load_dir(path, prefix=None) loads all files (starting with optional prefix) from a path into a single array.
npu.array.unique_rows(array) returns unique rows of an array
npu.array.groupby(array, column, presorted=True) returns a list of slices of array grouped by the values in the specified column