def load(reward_scale, seed):
"""Load a cartpole experiment with the prescribed settings."""
env = wrappers.RewardScale(env=cartpole.Cartpole(seed=seed),
reward_scale=reward_scale,
seed=seed)
env.bsuite_num_episodes = sweep.NUM_EPISODES
return env
def load(reward_scale, seed):
"""Load a bandit_scale experiment with the prescribed settings."""
env = wrappers.RewardScale(env=mnist.MNISTBandit(seed=seed),
reward_scale=reward_scale,
seed=seed)
env.bsuite_num_episodes = sweep.NUM_EPISODES
return env
def load(reward_scale: float, seed: int):
"""Load a mountain_car experiment with the prescribed settings."""
env = wrappers.RewardScale(env=mountain_car.MountainCar(seed=seed),
reward_scale=reward_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 main(unused_arg):
env = catch.Catch(seed=FLAGS.seed)
env = wrappers.RewardScale(env, reward_scale=FLAGS.reward_scale)
agent = PopArtAgent(
observation_spec=env.observation_spec(),
action_spec=env.action_spec(),
num_hidden_units=FLAGS.num_hidden_units,
epsilon=FLAGS.epsilon,
learning_rate=FLAGS.learning_rate,
pop_art_step_size=FLAGS.pop_art_step_size,
)
accumulator = TransitionAccumulator()
experiment.run_loop(
agent=agent,
environment=env,
accumulator=accumulator,
seed=FLAGS.seed,
batch_size=1,
train_episodes=FLAGS.train_episodes,
evaluate_every=FLAGS.evaluate_every,
eval_episodes=FLAGS.eval_episodes,
)
