my_dataset.py

from datasets import DatasetBuilder, DownloadManager, DatasetInfo, Features, Value, SplitGenerator, Split
import csv
import os

class WeatherDataset(DatasetBuilder):
    VERSION = "1.0.0"
    BUILDER_CONFIGS = []

    def _info(self):
        return DatasetInfo(
            description="This dataset contains hourly weather information from 2023/3/19 to 2024/3/16 in ten major stations in the state of North Carolina.",
            features=Features({
                "dt": Value("int64"),
                "main.temp": Value("float"),
                "main.feels_like": Value("float"),
                "main.pressure": Value("int64"),
                "main.humidity": Value("int64"),
                "main.temp_min": Value("float"),
                "main.temp_max": Value("float"),
                "wind.speed": Value("float"),
                "wind.deg": Value("int64"),
                "wind.gust": Value("float"),
                "clouds.all": Value("int64"),
                "latitude": Value("float"),
                "longitude": Value("float"),
                "date": Value("string"),
                "rain.1h": Value("float"),
                "weather_id": Value("int64"),
                "weather_main": Value("string"),
                "weather_description": Value("string"),
                "weather_icon": Value("string"),
                "city": Value("string"),
                "snow.1h": Value("float").nullable,
                "rain.3h": Value("float").nullable,
            }),
            supervised_keys=None,
            citation = """@misc{weather_dataset_2024, author = {Katherine Tian}, title = {Weather Dataset for NC City Analysis}, year = {23-24}, howpublished = {\url{https://github.com/yourusername/weather-dataset}} }""",
        )

    def _split_generators(self, dl_manager: DownloadManager):
        data_dir = dl_manager.download_and_extract("https://huggingface.co/datasets/Katherinetian/weather_data_NC/raw/main/weather_data.csv")
        return [
            SplitGenerator(
                name=Split.TRAIN,
                gen_kwargs={"filepath": os.path.join(data_dir, "weather_data.csv")},
            ),
        ]

    def _generate_examples(self, filepath):
        with open(filepath, encoding="utf-8") as csv_file:
            reader = csv.DictReader(csv_file)
            for id_, row in enumerate(reader):
                # Process nullable fields
                row['snow.1h'] = float(row['snow.1h']) if row.get('snow.1h') else None
                row['rain.3h'] = float(row['rain.3h']) if row.get('rain.3h') else None
                # Convert fields to their appropriate types
                row['rain.1h'] = float(row['rain.1h'])
                yield id_, row