MegaWrapper

class MegaWrapper(env, image_shape: tuple[int, int] = (84, 84), pixels_transform: Callable | None = None, goal_transform: Callable | None = None, required_keys: Iterable | None = None, separate_goal: Iterable = True, history_size: int = 1, frame_skip: int = 1)

Bases: Wrapper

Combines multiple wrappers for comprehensive environment preprocessing.

Applies in sequence:

AddPixelsWrapper → EverythingToInfoWrapper → EnsureInfoKeysWrapper → EnsureGoalInfoWrapper → ResizeGoalWrapper → StackedWrapper

This provides a complete preprocessing pipeline with rendered pixels, unified info dict, key validation, goal checking, goal resizing, and temporal stacking.

Parameters:

• env – The Gymnasium environment to wrap.

• image_shape – Target (height, width) for pixels and goal. Defaults to (84, 84).

• pixels_transform – Optional torchvision transform for rendered pixels.

• goal_transform – Optional torchvision transform for goal images.

• required_keys – Additional regex patterns for keys that must be in info. Pattern ^pixels(?:\..*)?$ is always added.

• separate_goal – If True, validates ‘goal’ is present in info. Defaults to True.

• n_stacks – Number of steps to stack (passed to StackedWrapper).

reset(*args, **kwargs)

Uses the reset() of the env that can be overwritten to change the returned data.

step(action)

Uses the step() of the env that can be overwritten to change the returned data.