outputs/gpt2cocoDSA/tokenized_data_DSA/dataset_dict.json

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