kd.data.py.Tfds

class kauldron.data.py.Tfds(name: str, *, _fake_refs: type[_FakeRefsUnset] | dict[str, _FakeRootCfg] = <class 'kauldron.utils.config_util._FakeRefsUnset'>, batch_size: int | None = None, seed: int | collections.abc.Sequence[int] | numpy.ndarray | jaxtyping.UInt32[Array, '2'] | jaxtyping.UInt32[ndarray, '2'] | jax.Array | None = _FakeRootCfg('cfg.seed'), transforms: tr_normalize.Transformations = <factory>, num_epochs: Optional[int] = None, batch_drop_remainder: bool = True, num_workers: int = 16, read_options: grain.ReadOptions | None = None, enable_profiling: bool = False, per_worker_buffer_size: int = 1, worker_init_fn: Callable[[int, int], None] | None = None, shuffle: bool, split: str, data_dir: epath.PathLike | None = None, decoders: Optional[Mapping[str, Any]] = None)

Bases: kauldron.data.py.base.DataSourceBase

Base TFDS loader.

Convenience wrapper around tfds.data_source.

name: str

split: str

data_dir: epath.PathLike | None = None

decoders: Mapping[str, Any] | None = None

property data_source: grain._src.python.data_sources.RandomAccessDataSource