test_repo.py

import json
import os
import datasets
import ast

# Metadata and descriptions for the dataset
_CITATION = """\
@InProceedings{huggingface:dataset,
    title = {Test Repo Dataset},
    author={huggingface, Inc.},
    year={2020}
}
"""

_DESCRIPTION = """\
The Test Repo dataset includes multiple choice questions tailored for NLP research and testing.
"""

_HOMEPAGE = "https://huggingface.co/datasets/anand-s/test_repo"

_LICENSE = "Apache License 2.0"

# Define URLs for different parts of the dataset if applicable
_URLS = {
    "mcq_domain": "https://huggingface.co/datasets/anand-s/test_repo/resolve/main/train_mcq.zip",
}

class TestRepo(datasets.GeneratorBasedBuilder):
    """Dataset for multiple choice questions from Test Repo."""

    VERSION = datasets.Version("1.0.0")

    BUILDER_CONFIGS = [
        datasets.BuilderConfig(name="mcq_domain", version=VERSION, description="This configuration covers multiple choice questions."),
    ]

    DEFAULT_CONFIG_NAME = "mcq_domain"

    def _info(self):
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=datasets.Features({
                "prompt": datasets.Value("string"),
                "question": datasets.Value("string"),
                "options": datasets.Sequence(datasets.Value("string")),
                "answer": datasets.Value("string"),
                "context": datasets.Value("string"),  # Assuming all data includes context
                "num_options": datasets.Value("string"),
                "question_type": datasets.Value("string"),
                "dataset_name": datasets.Value("string"),
            }),
            homepage=_HOMEPAGE,
            license=_LICENSE,
            citation=_CITATION,
        )

    def _split_generators(self, dl_manager):
        # Download and extract all the files in the directory
        data_dir = dl_manager.download_and_extract(_URLS[self.config.name])
        print(data_dir)
        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={"directory": data_dir, "split": "train"},
            ),
        ]

    def _generate_examples(self, directory, split):
        # Iterate over each file in the directory
        key_idx = 0
        for filename in os.listdir(directory):
            filepath = os.path.join(directory, filename)
            if filepath.endswith(".jsonl"):
                with open(filepath, encoding="utf-8") as f:
                    for key, row in enumerate(f):
                        data = json.loads(row)
                        yield key_idx, {
                            "prompt": data.get("prompt", ""),
                            "question": data["question"],
                            "options": data["options"],
                            "answer": data.get("answer", ""),
                            "context": data.get("context", ""),
                            "num_options": data.get("num_options", ""),
                            "question_type": data.get("question_type", ""),
                            "dataset_name": os.path.split(filepath)[-1].replace(".jsonl","")
                        }
                        key_idx +=1