nvflare.app_common.workflows.model_controller module

class ModelController(*args, persistor_id: str = 'persistor', **kwargs)[source]

Bases: BaseModelController, ABC

Model Controller API for FLModel-based Controller.

Parameters:

persistor_id (str, optional) – ID of the persistor component. Defaults to AppConstants.DEFAULT_PERSISTOR_ID (“persistor”).

load_model() → FLModel[source]

Load initial model from persistor. If persistor is not configured, returns empty FLModel.

Returns:

FLModel

abstract run()[source]

Main run routine for the controller workflow.

sample_clients(num_clients: int | None = None) → List[str][source]

Returns a list of num_clients clients.

Parameters:

num_clients – number of clients to return. If None or > number available clients, returns all available clients. Defaults to None.

Returns:

A list of clients names.

save_model(model: FLModel) → None[source]

Saves model with persistor. If persistor is not configured, does not save.

Parameters:

model (FLModel) – model to save.

Returns:

None

send_model(task_name: str = 'train', data: FLModel | None = None, targets: List[str] | None = None, min_responses: int | None = None, timeout: int = 0, callback: Callable[[FLModel], None] | None = None) → None[source]

Send a task with data to targets (non-blocking). Callback is called when a result is received.

Parameters:

  • task_name (str, optional) – name of the task. Defaults to “train”.

  • data (FLModel, optional) – FLModel to be sent to clients. Defaults to None.

  • targets (List[str], optional) – the list of target client names or None (all clients). Defaults to None.

  • min_responses (int, optional) – the minimum number of responses expected. If None, must receive responses from all clients that the task has been sent to. Defaults to None.

  • timeout (int, optional) – time to wait for clients to perform task. Defaults to 0 (never time out).

  • callback (Callable[[FLModel], None], optional) – callback when a result is received. Defaults to None.

Returns:

None

send_model_and_wait(task_name: str = 'train', data: FLModel | None = None, targets: List[str] | None = None, min_responses: int | None = None, timeout: int = 0) → List[FLModel][source]

Send a task with data to targets and wait for results.

Returns list of FLModel results from clients once task is completed (min_responses received or timed out). Results received from any clients after task is completed will be discarded.

Parameters:

  • task_name (str, optional) – name of the task. Defaults to “train”.

  • data (FLModel, optional) – FLModel to be sent to clients. Defaults to None.

  • targets (List[str], optional) – the list of target client names or None (all clients). Defaults to None.

  • min_responses (int, optional) – the minimum number of responses expected. If None, must receive responses from all clients that the task has been sent to. Defaults to None.

  • timeout (int, optional) – time to wait for clients to perform task. Defaults to 0 (never time out).

Returns:

List[FLModel]