PreProcessor

class maze.core.wrappers.observation_preprocessing.preprocessors.base.PreProcessor(observation_space: gym.spaces.Space, **kwargs)

Interface for observation pre-processors. Pre-processors implementing this interface can be used in combination with the PreProcessingWrapper.

Parameters
  • observation_space – The observation space to pre-process.

  • kwargs – Arguments to be passed on to preprocessor’s constructor.

abstract process(observation: numpy.ndarray)numpy.ndarray

Pre-processes the observation.

Parameters

observation – The observation to pre-process.

Returns

The pre-processed observation.

abstract processed_shape() → Tuple[int, ]

Computes the observation’s shape after pre-processing.

Returns

The resulting shape.

abstract processed_space() → gym.spaces.Box

Modifies the given observation space according to the respective pre-processor.

Returns

The updated observation space.

tag()str

Returns a tag identifying the pre-processed feature.

Returns

The pre-processor’s tag.