ImpalaLocalRunner

class maze.train.trainers.impala.impala_runners.ImpalaLocalRunner(state_dict_dump_file: str, dump_interval: int | None, spaces_config_dump_file: str, normalization_samples: int, eval_concurrency: int, start_method: str)

Runner for locally distributed training, based on SubprocVectorEnv.

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

create multi-process env distribution

create_distributed_rollout_actors(env_factory: Callable[[], StructuredEnv | StructuredEnvSpacesMixin | LogStatsEnv], policy: TorchPolicy, n_rollout_steps: int, n_actors: int, batch_size: int, queue_out_of_sync_factor: float, env_instance_seeds: List[int], agent_instance_seeds: List[int]) → SubprocDistributedActors

Create dummy (sequentially-executed) actors.

start_method: str

Type of start method used for multiprocessing (‘forkserver’, ‘spawn’, ‘fork’).