p2pfl.learning.frameworks.simulation.actor_pool module¶

Actor pool for distributed computing using Ray.

class p2pfl.learning.frameworks.simulation.actor_pool.SuperActorPool(resources=None, actor_list=None)[source]¶

Bases: ActorPool

SuperActorPool extends ActorPool to manage a pool of VirtualLearnerActor instances for asynchronous distributed computing using Ray.

_instance¶

Singleton instance of SuperActorPool.

Type:: SuperActorPool

_lock¶

Lock for thread-safe instance creation.

Type:: threading.Lock

resources¶

Resources for actor creation.

Type:: dict

_addr_to_future¶

Mapping from actor address to future information.

Type:: dict

actor_to_remove¶

Set of actor IDs scheduled for removal.

Type:: set

num_actors¶

Number of active actors in the pool.

Type:: int

lock¶

Reentrant lock for thread-safe operations.

Type:: threading.RLock

initialized¶

Flag indicating initialization status.

Type:: bool

add_actor(num_actors)[source]¶

Add a specified number of actors to the pool.

Parameters:: num_actors (int) – Number of actors to add.
Return type:: None

create_actor()[source]¶

Create a new VirtualLearnerActor instance using provided resources.

Return type:: ActorClass(VirtualLearnerActor)
Returns:: New actor instance.

get_learner_result(addr, timeout)[source]¶

Retrieve the learner result associated with the given address.

Parameters:

addr (str) – Address of the learner result to retrieve.
timeout (Optional[float]) – Timeout for retrieving the result. Defaults to None.

Return type:

Tuple[Any, Any]

Returns:

Address and result of the learner job.

process_unordered_future(timeout=None)[source]¶

Process the next unordered future result from the pool.

Parameters:

timeout (Optional[float]) – Timeout for processing the future. Defaults to None.

Raises:

StopIteration – If no more results are available.
TimeoutError – If the future processing times out.

Return type:

None

submit(fn, value)[source]¶

Submit a task to an idle actor in the pool.

Parameters:

fn (Any) – Function to be executed by the actor.
value (Tuple[str, Learner]) – Tuple containing address and learner information.

Return type:

None

submit_learner_job(actor_fn, job)[source]¶

Submit a learner job to the pool, handling pending submits if no idle actors are available.

Parameters:

actor_fn (Any) – Function to be executed by the actor.
job (Tuple[str, Learner]) – Tuple containing address and learner information.

Return type:

None