SACDevRunner

class maze.train.trainers.sac.sac_runners.SACDevRunner(state_dict_dump_file: str, dump_interval: int | None, spaces_config_dump_file: str, normalization_samples: int, eval_concurrency: int, initial_demonstration_trajectories: omegaconf.DictConfig)

Runner for single-threaded training, based on SequentialVectorEnv.

create_distributed_eval_env(env_factory: Callable[[], MazeEnv], eval_concurrency: int, logging_prefix: str) → SequentialVectorEnv

(overrides SACRunner)

create single-threaded env distribution

create_distributed_rollout_workers(env_factory: Callable[[], MazeEnv], worker_policy: TorchPolicy, n_rollout_steps: int, n_workers: int, batch_size: int, rollouts_per_iteration: int, split_rollouts_into_transitions: bool, env_instance_seeds: List[int], replay_buffer: BaseReplayBuffer) → DummyDistributedWorkersWithBuffer

(overrides SACRunner)

Create dummy (sequentially-executed) actors.