BroadcastingContainer

class maze.train.parallelization.distributed_actors.broadcasting_container.BroadcastingContainer

Synchronizes policy updates and other information across actors on local machine.

Used for dummy and sub-process distribution scenarios.

The BroadcastingContainer object can be read by all actor workers in order to update their policy, and it can be accessed by the main Thread to write the updated policy from the learner into it.

policy_state_dict() → Dict

Return the current policy state dict.

policy_version()int

Return the current policy version number (to check whether fetching a new state dict is necessary).

set_policy_state_dict(state_dict: Dict) → NoReturn

Store new policy version.

Parameters

state_dict – New state dict to store

set_stop_flag() → NoReturn

Signal to the workers to exit after they finish the current rollout.

stop_flag()bool

True if workers should exit.