train_with_context/dataset_dict.json

{"splits": ["train"]}