Parallelization

This page contains the reference documentation for the parallelization module.

Vectorized Environments

These are interfaces, classes and utility functions for vectorized environments:

VectorEnv	Abstract base class for vectorised environments.
StructuredVectorEnv	Common superclass for the structured vectorised env implementations in Maze.
SequentialVectorEnv	Creates a simple wrapper for multiple environments, calling each environment in sequence on the current Python process.
SubprocVectorEnv	Creates a multiprocess wrapper for multiple environments, distributing each environment to its own process.
CloudpickleWrapper	Uses cloudpickle to serialize contents (otherwise multiprocessing tries to use pickle).
SinkHoleConsumer	Sink hole statistics consumer.
disable_epoch_level_stats	Disable collection of statistics on epoch level to save memory.

Distributed Actors

These are interfaces, classes and utility functions for distributed actors:

DistributedActors	The base class for all distributed actors.
SequentialDistributedActors	Dummy implementation of distributed actors creates the actors as a list. Once the outputs are to
SubprocDistributedActors	Basic Distributed-Actors-Module using python multiprocessing.Process
BaseDistributedWorkersWithBuffer	The base class for all distributed workers with buffer.
DummyDistributedWorkersWithBuffer	Dummy implementation of distributed workers with buffer creates the workers as a list.

Utilities

Reusable components used in multiple distribution scenarios:

BroadcastingContainer	Synchronizes policy updates and other information across workers on local machine.
BroadcastingManager	A wrapper around BaseManager, used for managing the broadcasting container in multiprocessing scenarios.