data/processed_subset/dataset_dict.json

{"splits": ["train", "validation"]}