Method __init__

rl3/a2c/atari_wrappers.py:10–18

Sample initial states by taking random number of no-ops on reset. No-op is assumed to be action 0.

(self, env, noop_max=30)

class NoopResetEnv(gym.Wrapper):
    def __init__(self, env, noop_max=30):
        """Sample initial states by taking random number of no-ops on reset.
        No-op is assumed to be action 0.
        """
        gym.Wrapper.__init__(self, env)
        self.noop_max = noop_max
        self.override_num_noops = None
        self.noop_action = 0
        assert env.unwrapped.get_action_meanings()[0] == 'NOOP'

    def reset(self, **kwargs):
        """ Do no-op action for a number of steps in [1, noop_max]."""