dataset_loader.py

import os
import json
from datasets import DatasetBuilder, SplitGenerator, DownloadManager, DatasetInfo, Features, Value, Sequence

class MultiFileDataset(DatasetBuilder):
    """Example of a Hugging Face dataset script for handling multiple files per split."""

    VERSION = "1.0.0"

    def _info(self):
        return DatasetInfo(
            description="This dataset includes multiple JSON lines files for text classification.",
            features=Features({
                "question": Value("string"),
                "options": Sequence(Value("string")),
                "answer": Value("string"),
                "prompt": Value("string"),
                "num_options": Value("string"),  # Assuming numeric values are expected for this field
                "question_type": Value("string"),
                "exam_id": Value("string")  # Assuming all records will have an exam_id; use default value otherwise
            }),
            homepage="https://www.example.com/mydataset",
            citation="Cite the source here if applicable.",
        )

    def _split_generators(self, dl_manager):
        """Returns SplitGenerators."""
        # Assuming `data_dir` is your dataset's root directory on Hugging Face
        return [
            SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={"filepath": os.path.join(self.config.data_dir, "train/mcq")},
            ),
            SplitGenerator(
                name=datasets.Split.TEST,
                gen_kwargs={"filepath": os.path.join(self.config.data_dir, "test/mcq")},
            ),
        ]

    def _generate_examples(self, filepath):
        """Yields examples from multiple files."""
        for file in sorted(os.listdir(filepath)):  # Ensures consistent order
            full_file_path = os.path.join(filepath, file)
            if full_file_path.endswith('.jsonl'):  # Ensure to process only JSON Lines files
                with open(full_file_path, encoding="utf-8") as f:
                    for id, line in enumerate(f):
                        data = json.loads(line)
                        yield f"{file}_{id}", {
                            "question": data["question"],
                            "options": data["options"],
                            "answer": data["answer"],
                            "prompt": data["prompt"],
                            "num_options": int(data["num_options"]),  # Convert to int if it's a string in JSON
                            "question_type": data["question_type"],
                            "exam_id": data.get("exam_id", "")  # Check if this optional field exists
                        }