def model_vector_env(env: EnvType) -> BaseEnv:
"""Returns a VectorizedEnv wrapper around the given environment.
To obtain worker configs, one can call get_global_worker().
Args:
env (EnvType): The input environment (of any supported environment
type) to be convert to a _VectorizedModelGymEnv (wrapped as
an RLlib BaseEnv).
Returns:
BaseEnv: The BaseEnv converted input `env`.
"""
worker = get_global_worker()
worker_index = worker.worker_index
if worker_index:
env = _VectorizedModelGymEnv(
make_env=worker.make_sub_env_fn,
existing_envs=[env],
num_envs=worker.num_envs,
observation_space=env.observation_space,
action_space=env.action_space,
)
return convert_to_base_env(
env,
make_env=worker.make_sub_env_fn,
num_envs=worker.num_envs,
remote_envs=False,
remote_env_batch_wait_ms=0,
)
def test_nested_tuple_async(self):
self.do_test_nested_tuple(lambda _: convert_to_base_env(NestedTupleEnv()))
def test_nested_dict_async(self):
self.do_test_nested_dict(lambda _: convert_to_base_env(NestedDictEnv()))
def test_check_correct_env(self):
env = self._make_base_env()
check_env(env)
env = gym.make("CartPole-v0")
env = convert_to_base_env(env)
check_env(env)