ACLocalRunner

class maze.train.trainers.common.actor_critic.actor_critic_runners.ACLocalRunner(state_dict_dump_file: str, dump_interval: int | None, spaces_config_dump_file: str, normalization_samples: int, trainer_class: str | type, concurrency: int, eval_concurrency: int)

Runner for locally distributed training, based on SubprocVectorEnv.

create_distributed_env(env_factory: Callable[[], MazeEnv | StructuredEnv], concurrency: int, logging_prefix: str) → SubprocVectorEnv

create multi-process env distribution