RandomResetWrapper

class maze.core.wrappers.random_reset_wrapper.RandomResetWrapper(env: StructuredEnvSpacesMixin | MazeEnv, min_skip_steps: int, max_skip_steps: int)

A wrapper skipping the first few steps by taking random actions. This is useful for skipping irrelevant initial parts of a trajectory or for introducing randomness in the training process.

Parameters:

• env – Environment/wrapper to wrap.

• min_skip_steps – Minimum number of steps to skip.

• max_skip_steps – Maximum number of steps to skip.

clone_from(env: RandomResetWrapper) → None

(overrides SimulatedEnvMixin)

implementation of SimulatedEnvMixin.

get_observation_and_action_dicts(maze_state: Any | None, maze_action: Any | None, first_step_in_episode: bool) → Tuple[Dict[int | str, Any] | None, Dict[int | str, Any] | None]

(overrides Wrapper)

This wrapper does not modify observations and actions.

reset() → Any

(overrides BaseEnv)

Override BaseEnv.reset to reset the step count.

seed(seed: int) → None

(overrides StructuredEnv)

Apply seed to wrappers rng, and pass the seed forward to the env