dataset.py

# dataset.py

from datasets import DatasetBuilder, DatasetInfo, SplitGenerator, Features, Value, Audio
import os
import pandas as pd

class MarathiTTSDataset(DatasetBuilder):
    def _info(self):
        return DatasetInfo(
            description="Marathi Text-to-Speech dataset.",
            features=Features({
                'audio': Audio(sampling_rate=22050),  # Adjust if your audio has a different sampling rate
                'text': Value('string'),
            }),
            supervised_keys=None,
        )

    def _split_generators(self, dl_manager):
        data_dir = os.path.abspath(".")
        return [
            SplitGenerator(
                name="train",
                gen_kwargs={
                    "filepath": os.path.join(data_dir, "train.csv"),
                },
            )
        ]

    def _generate_examples(self, filepath):
        df = pd.read_csv(filepath)
        for idx, row in df.iterrows():
            yield idx, {
                'audio': row['audio'],
                'text': row['text'],
            }