def __init__(
self,
name,
env,
dynamics_model,
reward_model=None,
discount=1,
use_cem=False,
n_candidates=1024,
horizon=10,
use_reward_model=False,
num_rollouts=10,
context=False,
mcl_cadm=False,
):
self.dynamics_model = dynamics_model # dynamics_model
self.reward_model = reward_model # None
self.discount = discount # 1
self.n_candidates = n_candidates # 2000
self.horizon = horizon # 30
self.use_cem = use_cem # False
self.env = env # OurHalfCheetahEnv
self.use_reward_model = use_reward_model # False
self.context = context
self.mcl_cadm = mcl_cadm
self.unwrapped_env = env # OurHalfCheetahEnv
while hasattr(self.unwrapped_env, "wrapped_env"):
self.unwrapped_env = self.unwrapped_env.wrapped_env
# make sure that enc has reward function
assert hasattr(self.unwrapped_env, "reward"), "env must have a reward function"
Serializable.quick_init(self, locals())
super(MPCController, self).__init__(env=env)
def __init__(
self,
env,
scale_reward=1.0,
normalize_obs=False,
normalize_reward=False,
obs_alpha=0.001,
reward_alpha=0.001,
normalization_scale=1.0,
dummy_flag=False,
):
Serializable.quick_init(self, locals())
self._scale_reward = 1
self._wrapped_env = env
self._normalize_obs = normalize_obs
self._normalize_reward = normalize_reward
self._obs_alpha = obs_alpha
self._obs_mean = np.zeros(self.observation_space.shape)
self._obs_var = np.ones(self.observation_space.shape)
self._reward_alpha = reward_alpha
self._reward_mean = 0.0
self._reward_var = 1.0
self._normalization_scale = normalization_scale
self._dummy_flag = dummy_flag
def __init__(self, observation_space, action_space):
"""
:type observation_space: Space
:type action_space: Space
"""
Serializable.quick_init(self, locals())
self._observation_space = observation_space
self._action_space = action_space
def __init__(self, *args, **kwargs):
# store the init args for serialization and call the super constructors
Serializable.quick_init(self, locals())
Layer.__init__(self, *args, **kwargs)
self.build_graph()
def __init__(self, env):
Serializable.quick_init(self, locals())
self.env = env
while hasattr(self.env, "wrapped_env"):
self.env = self.env.wrapped_env
def __setstate__(self, d):
Serializable.__setstate__(self, d)
self._obs_mean = d["_obs_mean"]
self._obs_var = d["_obs_var"]
def __getstate__(self):
d = Serializable.__getstate__(self)
d["_obs_mean"] = self._obs_mean
d["_obs_var"] = self._obs_var
return d