def __init__(self, rl2_max_path_length, meta_batch_size, task_sampler,
**inner_algo_args):
self._inner_algo = RL2NPO(**inner_algo_args)
self._rl2_max_path_length = rl2_max_path_length
self._env_spec = self._inner_algo.env_spec
self._flatten_input = self._inner_algo.flatten_input
self._policy = self._inner_algo.policy
self._discount = self._inner_algo.discount
self._meta_batch_size = meta_batch_size
self._task_sampler = task_sampler
def __init__(self, rl2_max_path_length, meta_batch_size, task_sampler,
meta_evaluator, n_epochs_per_eval, **inner_algo_args):
self._inner_algo = RL2NPO(**inner_algo_args)
self._rl2_max_path_length = rl2_max_path_length
self.env_spec = self._inner_algo._env_spec
self._n_epochs_per_eval = n_epochs_per_eval
self._policy = self._inner_algo.policy
self._discount = self._inner_algo._discount
self._meta_batch_size = meta_batch_size
self._task_sampler = task_sampler
self._meta_evaluator = meta_evaluator
def __init__(self, env_spec, episodes_per_trial, meta_batch_size,
task_sampler, meta_evaluator, n_epochs_per_eval,
**inner_algo_args):
self._env_spec = env_spec
_inner_env_spec = EnvSpec(
env_spec.observation_space, env_spec.action_space,
episodes_per_trial * env_spec.max_episode_length)
self._inner_algo = RL2NPO(env_spec=_inner_env_spec, **inner_algo_args)
self._rl2_max_episode_length = self._env_spec.max_episode_length
self._n_epochs_per_eval = n_epochs_per_eval
self._policy = self._inner_algo.policy
self._discount = self._inner_algo._discount
self._meta_batch_size = meta_batch_size
self._task_sampler = task_sampler
self._meta_evaluator = meta_evaluator