audio_vreme.py

from datasets import GeneratorBasedBuilder, DatasetInfo, SplitGenerator, Split, Features, Value, Audio,SplitGenerator, Split
import os
import json
import csv

import datasets
from datasets.utils.py_utils import size_str
from tqdm import tqdm


_BASE_URL = "https://huggingface.co/datasets/iulik-pisik/audio_vreme/resolve/main/"
_AUDIO_URL = _BASE_URL + "audio/{split}.tar"
_TRANSCRIPT_URL = _BASE_URL + "transcript/{split}.tsv"

class AudioVreme(GeneratorBasedBuilder):
    
    def _info(self):
        return DatasetInfo(
            description="Dataset de vreme preluat de la ProTV",
            features=Features({
                "path": Value("string"),
                "audio": Audio(sampling_rate=16000),
                "sentence": Value("string"),
                "gender": Value("string"),
                "name": Value("string"),
            }),
            supervised_keys=("audio", "transcript"),
            homepage="https://huggingface.co/datasets/iulik-pisik/audio_vreme",
            citation="Referința de citare a datasetului",
        )

    def _split_generators(self, dl_manager):
        audio_urls = {
            "train_audio": _AUDIO_URL.format(split="train"),
            "test_audio": _AUDIO_URL.format(split="test"),
            "validation_audio": _AUDIO_URL.format(split="validation"),
        }
        tsv_urls = {
            "train_tsv": _TRANSCRIPT_URL.format(split="train"),
            "test_tsv": _TRANSCRIPT_URL.format(split="test"),
            "validation_tsv": _TRANSCRIPT_URL.format(split="validation"),
        }
    
        downloaded_audio_files = dl_manager.download_and_extract(audio_urls)
        downloaded_tsv_files = dl_manager.download(tsv_urls)
    
        return [
            SplitGenerator(
                name=Split.TRAIN,
                gen_kwargs={
                    "archive_path": downloaded_audio_files["train_audio"],
                    "tsv_path": downloaded_tsv_files["train_tsv"],
                },
            ),
            SplitGenerator(
                name=Split.TEST,
                gen_kwargs={
                    "archive_path": downloaded_audio_files["test_audio"],
                    "tsv_path": downloaded_tsv_files["test_tsv"],
                },
            ),
            SplitGenerator(
                name=Split.VALIDATION,
                gen_kwargs={
                    "archive_path": downloaded_audio_files["validation_audio"],
                    "tsv_path": downloaded_tsv_files["validation_tsv"],
                },
            ),
        ]


    def _generate_examples(self, archive_path, tsv_path):
        with open(tsv_path, encoding="utf-8") as f:
            reader = csv.DictReader(f, delimiter="\t", quoting=csv.QUOTE_NONE)
            for row in tqdm(reader, desc="Se citesc datele..."):
                audio_file_name = row["path"]
                audio_path = os.path.join(archive_path, audio_file_name)
                
                if not os.path.isfile(audio_path):
                    continue
    
                yield audio_file_name, {
                    "path": audio_path,  
                    "audio": audio_path,
                    "sentence": row["sentence"],
                    "gender": row["gender"],
                    "name": row["name"]
                }