olympic_test.py

import os
import datasets
import json


_CITATION = """\
"""

_DESCRIPTION = """\
OlympicArena
"""

_HOMEPAGE = ""

_URL = ""


subject_list = ["Math", "Physics", "Chemistry", "Biology", "Geography", "Astronomy", "CS"]


class OlympicArenaConfig(datasets.BuilderConfig):
    """BuilderConfig"""

    def __init__(self, **kwargs):
        """BuilderConfig
        Args:
          **kwargs: keyword arguments forwarded to super.
        """
        super(OlympicArenaConfig, self).__init__(**kwargs)
        
        
        
class OlympicArena(datasets.GeneratorBasedBuilder):
    BUILDER_CONFIGS = [
        OlympicArenaConfig(name=subject) for subject in subject_list
    ]

    def _info(self):
        features = datasets.Features(
            {
                "id": datasets.Value("string"),
                "problem": datasets.Value("string"),
                "prompt": datasets.Value("string"),
                "figure_urls": datasets.Sequence(datasets.Value("string")),
                "answer": datasets.Sequence(datasets.Value("string")),
                "solution": datasets.Value("string"),
                "answer_type": datasets.Value("string"),
                "unit": datasets.Sequence(datasets.Value("string")),
                "answer_sequence": datasets.Sequence(datasets.Value("string")),
                "type_sequence": datasets.Sequence(datasets.Value("string")),
                "test_cases": datasets.Sequence(
                    {
                        "input": datasets.Value("string"),
                        "output": datasets.Value("string")
                    }
                ),
                "subject": datasets.Value("string"),
                "language": datasets.Value("string"),
                "modality": datasets.Value("string")
            }
        )
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=features,
            homepage=_HOMEPAGE,
            citation=_CITATION,
        )

    def _split_generators(self, dl_manager):
        subject = self.config.name
        return [
            datasets.SplitGenerator(
                name=datasets.Split.TEST,
                gen_kwargs={
                    "filepath": dl_manager.download(os.path.join("test", subject+".json")),
                },
            ),
            datasets.SplitGenerator(
                name=datasets.Split("val"),
                gen_kwargs={
                    "filepath": dl_manager.download(os.path.join("val", subject+".json")),
                },
            ),
        ]

    def _generate_examples(self, filepath):
        with open(filepath, "r") as f:
            data = json.load(f)
        for i, instance in enumerate(data):
            yield i, instance