tedrics-data.py

import json
import datasets

_REPO_NAME = "TeDriCS/tedrics-data"

_DESCRIPTION = ""

_HOMEPAGE = ""

_CITATION = """\
@misc{,
      title={ },
      author={},
      year={2022}
}
"""

_LICENSE = 'CC BY-SA'

_SUBSETS = ["tasks", "testcases", "codefunctions"]

_DATA_URLS = {
    "tasks": {
        "train": "tedrics_data_tasks.json"
    },
    "testcases": {
        "train": "tedrics_data_testcases.json",
        "validation": "tedrics_data_testcases_val.json"
    },
    "codefunctions": {
        "train": "tedrics_data_codefunctions.json",
        "validation": "tedrics_data_codefunctions_val.json"
    }
}

class TeDriCSData(datasets.GeneratorBasedBuilder):
    BUILDER_CONFIGS = [
        datasets.BuilderConfig(
            name=f"{subset}",
            version=datasets.Version("1.1.0"),
            description=_DESCRIPTION,
        )
        for subset in _SUBSETS
    ]

    DEFAULT_CONFIG_NAME = "testcases"

    def _info(self):
        if self.config.name == "tasks":
            features = datasets.Features(
                {
                    "task_id": datasets.Value("int32"),
                    "mbpp_task_id": datasets.Value("int32"),
                    "source": datasets.Value("string"),
                    "licence": datasets.Value("string"),
                    "task": datasets.Value("string"),
                }
            )

        if self.config.name == "testcases":
            features = datasets.Features(
                {
                    "task_id": datasets.Value("int32"),
                    "mbpp_task_id": datasets.Value("int32"),
                    "task": datasets.Value("string"),
                    "test_cases": datasets.Sequence(
                        {
                            "test_case_id": datasets.Value("int32"),
                            "cot": datasets.Value("string"),
                            "input": datasets.Value("string"),
                            "output": datasets.Value("string")
                        }
                    )
                }
            )

        if self.config.name == "codefunctions":
            features = datasets.Features(
                {
                    "task_id": datasets.Value("int32"),
                    "mbpp_task_id": datasets.Value("int32"),
                    "description": datasets.Value("string"),
                    "cot": datasets.Value("string"),
                    "imports": datasets.Value("string"),
                    "function_head": datasets.Value("string"),
                    "function_body": datasets.Value("string")
                }
            )

        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=features,
            supervised_keys=None,
            homepage=_HOMEPAGE,
            license=_LICENSE,
            citation=_CITATION,
        )

    def _split_generators(self, dl_manager):
        urls = _DATA_URLS[self.config.name]
        data = dl_manager.download_and_extract(urls)

        splits = []

        if self.config.name == "tasks":
            splits = [datasets.Split.TRAIN]

        if self.config.name == "testcases" or self.config.name == "codefunctions":
            splits = [datasets.Split.TRAIN, datasets.Split.VALIDATION]

        return [
            datasets.SplitGenerator(
                name=split,
                gen_kwargs={
                    "filepath": data[split],
                },
            )
            for split in splits
        ]

    def _generate_examples(self, filepath):
        with open(filepath, encoding="utf-8") as file:
            data = json.load(file)
            id_ = 0
            for sample in data:
                yield id_, sample
                id_ += 1

####Zum Testen der Implementierung
# from datasets import load_dataset
# def main():
#     dataset = load_dataset('C:\\Users\\klaud\\TeDriCSProj\\tedrics-data\\tedrics-data.py')
#     print(dataset["train"])
#     print(dataset["train"][0])
#     print(dataset["validation"])
#     print(dataset["validation"][0])

# if __name__ == "__main__":
#    main()

#### Zum Generieren der dataset_infos.json (im Terminal)
# (hf-gptneox-cpu) PS C:\Users\klaud\TeDriCSProj\tedrics-data> datasets-cli test C:\Users\klaud\TeDriCSProj\tedrics-data\tedrics-data.py --save_infos --all_configs