from datasets import DatasetInfo, Features, Split, SplitGenerator, GeneratorBasedBuilder, Value, Sequence import json class MyDataset(GeneratorBasedBuilder): def _info(self): return DatasetInfo( features=Features({ "questions": Sequence(Value("string")), "answers": Sequence(Value("string")) }), supervised_keys=("questions", "answers"), homepage="https://github.com/FreedomIntelligence/HuatuoGPT", citation="...", ) def _split_generators(self, dl_manager): train_path = "train_datasets.jsonl" validation_path = "validation_datasets.jsonl" test_path = "test_datasets.jsonl" return [ SplitGenerator(name=Split.TRAIN, gen_kwargs={"filepath": train_path}), SplitGenerator(name=Split.VALIDATION, gen_kwargs={"filepath": validation_path}), SplitGenerator(name=Split.TEST, gen_kwargs={"filepath": test_path}), ] def _generate_examples(self, filepath): with open(filepath, encoding="utf-8") as f: for id_, row in enumerate(f): # Process your data here and create a dictionary with the features. # For example, if your data is in JSON format: data = json.loads(row) yield id_, { "questions": data["questions"], "answers": data["answers"], } if __name__ == '__main__': from datasets import load_dataset dataset = load_dataset("my_dataset.py") print()