tff.simulation.datasets.save_to_sql_client_data

Serialize a federated dataset into a SQL database compatible with SqlClientData.

tff.simulation.datasets.save_to_sql_client_data(
    client_ids: list[str],
    dataset_fn: Callable[[str], tf.data.Dataset],
    database_filepath: str,
    allow_overwrite: bool = False,
    split_names_by_client_id: Optional[Mapping[str, str]] = None
) -> None

Note: All the clients must share the same dataset.element_spec of type Mapping[str, TensorSpec].

Args
client_ids	A list of string identifiers for clients in this dataset.
dataset_fn	A callable that accepts a str as an argument and returns a tf.data.Dataset instance. Unlike from_client_and_tf_dataset_fn, it does not require a TF serializable dataset function.
database_filepath	A str filepath to the SQL database.
allow_overwrite	A boolean indicating whether to allow overwriting if file already exists at dataset_filepath.
split_names_by_client_id	An optional mapping from client identifier to a str identifier for the split (e.g. "train") the client belongs to.

Raises
FileExistsError	if file exists at dataset_filepath and allow_overwrite is False. If overwriting is intended, please use allow_overwrite = True.
ElementSpecCompatibilityError	if the element_spec of local datasets are not identical across clients, or if the element_spec of datasets are not of type Mapping[str, TensorSpec].
ValueError	If split_names_by_client_id is passed and is missing an entry for a client in client_ids.