ACRunner

class maze.train.trainers.common.actor_critic.actor_critic_runners.ACRunner(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)

Abstract baseclass of AC runners.

concurrency: int

Number of concurrently executed environments.

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

The dev and local runner implement the setup of the distribution env

eval_concurrency: int

Number of concurrent evaluation envs

setup(cfg: omegaconf.DictConfig) → None

(overrides TrainingRunner)

See setup().

trainer_class: str | type

The actor critic trainer class to be used.