nexuml_library.scenarios.data.audioset¶
AudioSet data specification builders.
audioset_data(
data_root: str = "audioset_hf/full",
train_subset: Literal[
"bal_train", "unbal_train"
] = "bal_train",
download: bool = False,
sample_rate: int = 16000,
clip_num_samples: int = 160000,
batch_size: int = 128,
num_workers: int = 4,
num_classes: int = 527,
perform_checks: bool = False,
validate_layout: bool = True,
max_samples: int | None = None,
) -> DataSpec
AudioSet balanced-train + eval data spec.
Returns:
| Name | Type | Description |
|---|---|---|
DataSpec |
DataSpec
|
Audio dataset specification with fit and test splits. |
Raises:
| Type | Description |
|---|---|
FileNotFoundError
|
If AudioSet data is missing or incomplete after validation. |