def load(noise_scale, seed):
"""Load a cartpole experiment with the prescribed settings."""
env = wrappers.RewardNoise(env=cartpole.Cartpole(seed=seed),
noise_scale=noise_scale,
seed=seed)
env.bsuite_num_episodes = sweep.NUM_EPISODES
return env
def load(noise_scale, seed):
"""Load a mountain_car experiment with the prescribed settings."""
env = wrappers.RewardNoise(env=mountain_car.MountainCar(seed=seed),
noise_scale=noise_scale,
seed=seed)
env.bsuite_num_episodes = sweep.NUM_EPISODES
return env
def load(noise_scale, seed):
"""Load a bandit_noise experiment with the prescribed settings."""
env = wrappers.RewardNoise(env=bandit.SimpleBandit(seed=seed),
noise_scale=noise_scale,
seed=seed)
env.bsuite_num_episodes = sweep.NUM_EPISODES
return env
def test_unwrap(self):
raw_env = FakeEnvironment([dm_env.restart([])])
scale_env = wrappers.RewardScale(raw_env, reward_scale=1.)
noise_env = wrappers.RewardNoise(scale_env, noise_scale=1.)
logging_env = wrappers.Logging(noise_env, logger=None) # pytype: disable=wrong-arg-types
unwrapped = logging_env.raw_env
self.assertEqual(id(raw_env), id(unwrapped))
def load(mean_rewards=None, seed=None):
"""Load a bandit_noise experiment with the prescribed settings."""
env = wrappers.RewardNoise(env=bandit.SimpleBandit(rewards=mean_rewards,
seed=seed),
bernoulli=True,
seed=seed)
env.bsuite_num_episodes = sweep.NUM_EPISODES
return env
