nexuml_library.data.audioset.audioset_hf_download¶
Hugging Face-only AudioSet downloader.
is_hf_audioset_complete(root: str | Path) -> bool
Return whether the Hugging Face AudioSet layout exists.
expected_hf_audioset_layout(root: str | Path) -> str
Return the expected HuggingFace AudioSet directory layout path.
download_hf_audioset(
root: str | Path, decode: bool = True
) -> None
Download AudioSet from Hugging Face into root.
Uses only agkphysics/AudioSet config full and splits
bal_train, eval, and unbal_train. No YouTube or yt-dlp code is
imported or invoked.
Raises:
| Type | Description |
|---|---|
RuntimeError
|
If required packages (datasets, torchaudio, tqdm) are not installed. |