hive.envs.minatar.minatar module

class hive.envs.minatar.minatar.MinAtarEnv(env_name, sticky_action_prob=0.1, difficulty_ramping=True)[source]

Bases: BaseEnv

Class for loading MinAtar environments. See https://github.com/kenjyoung/MinAtar.

Parameters

  • env_name (str) – Name of the environment

  • sticky_actions (bool) – Whether to use sticky_actions as per Machado et al.

  • difficulty_ramping (bool) – Whether to periodically increase difficulty.

create_env_spec(env_name)[source]
reset()[source]

Resets the state of the environment.

Returns

The initial observation of the new episode. turn (int): The index of the agent which should take turn.

Return type

observation

seed(seed=None)[source]

Reseeds the environment.

Parameters

seed (int) – Seed to use for environment.

step(action=None)[source]
Remarks:

  • Execute self.frame_skips steps taking the action in the the environment.

  • This may execute fewer than self.frame_skip steps in the environment, if the done state is reached.

  • Furthermore, in this case the returned observation should be ignored.