espnet3.components.data.data_organizer.DatasetConfig

source

class espnet3.components.data.data_organizer.DatasetConfig(name: str | None = None, data_src: str | None = None, data_src_args: Dict[str, Any] | None = None, transform: Dict[str, Any] | None = None, split: str | None = None)

Bases: object

Configuration class for dataset metadata and construction.

This class encapsulates the necessary fields to define and instantiate a dataset. Used with Hydra to allow modular and flexible configuration via YAML or dictionaries.

name

Name identifier for the dataset.

• Type: str

data_src

Optional dataset source reference resolved via espnet3.components.data.dataset_module.

• Type: Optional[str]

data_src

_args

Keyword arguments passed to the recipe Dataset class.

• Type: Optional[Dict[str, Any]]

transform

A dictionary for Hydra instantiation of a transform applied to each sample after loading.

• Type: Optional[Dict[str, Any]]

split

Optional split label kept for compatibility with configs that still carry it as metadata.

• Type: Optional[str]

Examples

Recipe-backed dataset entry. : ```python

config_dict = { ... "name": "custom", ... "data_src": "mini_an4/asr", ... "data_src_args": {"split": "test"}, ... "transform": { ... "target": "my_project.transforms.uppercase_transform", ... }, ... } config = DatasetConfig(**config_dict) config.data_src 'mini_an4/asr'

**Local recipe dataset entry.**
: ```python
>>> config = DatasetConfig(
...     name="local_eval",
...     data_src_args={"split": "eval"},
... )
>>> config.data_src is None
True

data_src

*: str | None* *= None*

data_src

_args *: Dict[str, Any] | None* *= None*

name

*: str | None* *= None*

split

*: str | None* *= None*

transform

*: Dict[str, Any] | None* *= None*