Fetch LibriLightLimited 10h fine-tuning dataset as TSV file

librilight10h_tsv.py

import torchaudio

import pandas as pd

from pathlib import Path
from tqdm import tqdm

llight10h =  torchaudio.datasets.LibriLightLimited(root="tmp/", subset="10h", download=True)

manifest_rows = []

for (audio_samples, sr, text, spk_id, chap_id, utt_id) in tqdm(llight10h):
    filename = f"{spk_id}-{chap_id}-{utt_id}.wav"

    torchaudio.save(
        str(Path("data/LibriLight10h/clips/") / filename),
        audio_samples,
        sr
    )

    manifest_rows.append({'path':"clips/" + filename, 'text': text.lower()})

manifest_df = pd.DataFrame(manifest_rows)
manifest_df
manifest_df.to_csv("data/LibriLight10h/_all.tsv", sep="\t", index=False)