A subset of datasets is used for all datasets, there should be a global datafolder in zenodo, with content that gets loaded lazily