nvflare.app_common.ccwf.swarm_client_ctl module

class Gatherer(task_data: Shareable, fl_ctx: FLContext, for_round: int, executor: ClientSideController, aggregator: Aggregator, metric_comparator: MetricComparator, all_clients: list, trainers: list, min_responses_required: int, wait_time_after_min_resps_received: float, timeout)[source]

Bases: FLComponent

Init FLComponent.

The FLComponent is the base class of all FL Components. (executors, controllers, responders, filters, aggregators, and widgets are all FLComponents)

FLComponents have the capability to handle and fire events and contain various methods for logging.

gather(client_name: str, result: Shareable, fl_ctx: FLContext) Shareable[source]
class SwarmClientController(task_name_prefix='swarm', learn_task_name='train', persistor_id='persistor', shareable_generator_id='shareable_generator', aggregator_id='aggregator', metric_comparator_id=None, learn_task_check_interval=1.0, learn_task_abort_timeout=5.0, learn_task_ack_timeout=10, learn_task_timeout=None, final_result_ack_timeout=10, min_responses_required: int = 1, wait_time_after_min_resps_received: float = 10.0)[source]

Bases: ClientSideController

Constructor of a ClientSideController object.

  • task_name_prefix – prefix of task names. All CCWF task names are prefixed with this.

  • learn_task_name – name for the Learning Task (LT)

  • persistor_id – ID of the persistor component

  • shareable_generator_id – ID of the shareable generator component

  • learn_task_check_interval – interval for checking incoming Learning Task (LT)

  • learn_task_ack_timeout – timeout for sending the LT to other client(s)

  • final_result_ack_timeout – timeout for sending final result to participating clients

  • learn_task_abort_timeout – time to wait for the LT to become stopped after aborting it

  • allow_busy_task – whether a new learn task is allowed when working on current learn task

do_learn_task(name: str, task_data: Shareable, fl_ctx: FLContext, abort_signal: Signal)[source]

This is called to do a Learn Task. Subclass must implement this method.

  • name – task name

  • data – task data

  • fl_ctx – FL context of the task

  • abort_signal – abort signal for the task execution


execute(task_name: str, shareable: Shareable, fl_ctx: FLContext, abort_signal: Signal) Shareable[source]

Executes a task.

  • task_name (str) – task name.

  • shareable (Shareable) – input shareable.

  • fl_ctx (FLContext) – fl context.

  • abort_signal (Signal) – signal to check during execution to determine whether this task is aborted.


An output shareable.

handle_event(event_type: str, fl_ctx: FLContext)[source]

Handles events.

  • event_type (str) – event type fired by workflow.

  • fl_ctx (FLContext) – FLContext information.

process_config(fl_ctx: FLContext)[source]

This is called to allow the subclass to process config props.

Returns: None

start_run(fl_ctx: FLContext)[source]
start_workflow(shareable: Shareable, fl_ctx: FLContext, abort_signal: Signal) Shareable[source]

This is called for the subclass to start the workflow. This only happens on the starting_client.

  • shareable – the initial task data (e.g. initial model weights)

  • fl_ctx – FL context

  • abort_signal – abort signal for task execution