def __init__(
self,
goal=(1, 1),
random_start=False,
show_traces=True,
completion_bonus=0.,
never_done=False,
action_scale=1.,
):
Serializable.quick_init(self, locals())
Parameterized.__init__(self)
self._goal = np.array(goal, dtype=np.float32)
self._point = np.zeros(2)
self._completion_bonus = completion_bonus
self._never_done = never_done
self._action_scale = action_scale
self.screen = None
self.screen_width = 500
self.screen_height = 500
self.zoom = 50.
self.show_traces = show_traces
self.random_start = random_start
self._traces = deque(maxlen=MAX_SHOWN_TRACES)
def __init__(self, env=None, env_name=""):
if env_name:
super().__init__(gym.make(env_name))
else:
super().__init__(env)
Parameterized.__init__(self)
Serializable.quick_init(self, locals())
def __init__(self, wrapped_env=None, wrapped_policy=None):
assert isinstance(wrapped_policy, MultitaskPolicy)
Serializable.quick_init(self, locals())
Parameterized.__init__(self)
self._wrapped_env = wrapped_env
self._wrapped_policy = wrapped_policy
self._last_obs = None
def __init__(self, env=None, env_name=""):
if env_name:
super().__init__(gym.make(env_name))
else:
super().__init__(env)
self.action_space = self._to_akro_space(self.env.action_space)
self.observation_space = self._to_akro_space(
self.env.observation_space)
Parameterized.__init__(self)
Serializable.quick_init(self, locals())
def __init__(self,
wrapped_env=None,
wrapped_policy=None,
latents=None,
normalize=False):
assert isinstance(wrapped_policy, MultitaskPolicy)
Serializable.quick_init(self, locals())
Parameterized.__init__(self)
self._wrapped_env = wrapped_env
self._wrapped_policy = wrapped_policy
self._latents = np.array(latents)
self._last_obs = None
self._normalize = normalize
def __init__(self,
task_selection_strategy=round_robin,
task_env_cls=None,
task_args=None,
task_kwargs=None):
Serializable.quick_init(self, locals())
Parameterized.__init__(self)
self._task_envs = [
task_env_cls(*t_args, **t_kwargs)
for t_args, t_kwargs in zip(task_args, task_kwargs)
]
self._task_selection_strategy = task_selection_strategy
self._active_task = None
def __init__(self,
wrapped_env=None,
wrapped_policy=None,
latents=None,
skip_steps=1,
deterministic=True):
assert isinstance(wrapped_policy, MultitaskPolicy)
assert isinstance(latents, list)
Serializable.quick_init(self, locals())
Parameterized.__init__(self)
self._wrapped_env = wrapped_env
self._wrapped_policy = wrapped_policy
self._latents = latents
self._last_obs = None
self._skip_steps = skip_steps
self._deterministic = deterministic
def __init__(self, wrapped_env=None, wrapped_policy=None):
assert isinstance(wrapped_policy, MultitaskPolicy)
Serializable.quick_init(self, locals())
Parameterized.__init__(self)
self._wrapped_env = wrapped_env
self._wrapped_policy = wrapped_policy
self._last_obs = None
n_task = self._wrapped_policy.task_space.flat_dim
one_hots = np.identity(n_task)
latents, infos = self._wrapped_policy._embedding.get_latents(one_hots)
latents_means = infos["mean"]
self._latents_combination_hash = list()
for i in range(n_task):
for j in range(i + 1, n_task):
self._latents_combination_hash.append(
(latents_means[i, ...], latents_means[j, ...]))
self._latents_combination_hash = tuple(self._latents_combination_hash)
self._n_skills = n_task
def __init__(self, env_spec):
Parameterized.__init__(self)
self._env_spec = env_spec
def __init__(self, env):
Serializable.quick_init(self, locals())
Parameterized.__init__(self)
super().__init__(env)
def __init__(self, *args, **kwargs):
Serializable.quick_init(self, locals())
Parameterized.__init__(self)
super().__init__(SawyerReachXYZEnv(*args, **kwargs))
def __init__(self, env):
Serializable.quick_init(self, locals())
Parameterized.__init__(self)
NormalizedEnv.__init__(self, env)
