Source code for garage.envs.step

"""Environment step data type."""
import collections


[docs]def Step(observation, reward, done, **kwargs): # noqa: N802 """Create a namedtuple from the results of environment.step(action). Provides the option to put extra diagnostic info in the kwargs (if it exists) without demanding an explicit positional argument. Args: observation (object): Agent's observation of the current environment reward (float) : Amount of reward returned after previous action done (bool): Whether the episode has ended, in which case further step() calls will return undefined results kwargs: Keyword args Returns: collections.namedtuple: A named tuple of the arguments. """ return _Step(observation, reward, done, kwargs)
_Step = collections.namedtuple('Step', ['observation', 'reward', 'done', 'info'])