Module ilpyt.utils.env_utils
Expand source code
import gym
from ilpyt.envs.dummy_vec_env import DummyVecEnv
from ilpyt.envs.subproc_vec_env import SubprocVecEnv
from ilpyt.envs.vec_env import VecEnv
import gym
from ilpyt.envs.dummy_vec_env import DummyVecEnv
from ilpyt.envs.subproc_vec_env import SubprocVecEnv
from ilpyt.envs.vec_env import VecEnv
def build_env(
env_id: str, num_env: int = 16, seed: int = 24, vecenv_type: str = "dummy"
) -> VecEnv:
"""
Build vectorized environment. Adapted from: https://github.com/openai/baselines/
Parameters
----------
env_id: str
Name of registered OpenAI Gym environment
num_env: int, default=16
number of parallel environments to initialize
seed: int, default=24
random seed for environments
vecenv_type: str, default="dummy"
Vectorized environment type; choose from ["dummy", "subproc"];
"dummy" will create a serialized environment, which is useful for debugging;
"subproc" will create parallel environments for high-throughput training
Returns
-------
VecEnv:
vectorized environment
"""
def make_env(env_id, rank):
env = gym.make(env_id)
env.seed(seed + rank)
env.action_space.seed(seed + rank)
return env
if vecenv_type == 'dummy':
vec_env = DummyVecEnv(
[lambda rank=i: make_env(env_id, rank) for i in range(num_env)]
) # type: VecEnv
elif vecenv_type == 'subproc':
vec_env = SubprocVecEnv(
[lambda rank=i: make_env(env_id, rank) for i in range(num_env)]
)
return vec_env
def list_all_envs() -> None:
"""
List all the environments available on this system.
Please see the official list of OpenAI Gym environments for more details: https://github.com/openai/gym/wiki/Table-of-environments
"""
env_ids = [env_spec.id for env_spec in gym.envs.registry.all()]
for env_id in env_ids:
try:
env = gym.make(env_id)
print(env_id, env.observation_space, env.action_space)
except Exception: # Mujoco
print(env_id, "not available on this system.")Functions
def build_env(env_id: str, num_env: int = 16, seed: int = 24, vecenv_type: str = 'dummy') ‑> VecEnv-
Build vectorized environment. Adapted from: https://github.com/openai/baselines/
Parameters
env_id:str- Name of registered OpenAI Gym environment
num_env:int, default=16- number of parallel environments to initialize
seed:int, default=24- random seed for environments
vecenv_type:str, default="dummy"- Vectorized environment type; choose from ["dummy", "subproc"]; "dummy" will create a serialized environment, which is useful for debugging; "subproc" will create parallel environments for high-throughput training
Returns
Vecenv
vectorized environment
Expand source code
def build_env( env_id: str, num_env: int = 16, seed: int = 24, vecenv_type: str = "dummy" ) -> VecEnv: """ Build vectorized environment. Adapted from: https://github.com/openai/baselines/ Parameters ---------- env_id: str Name of registered OpenAI Gym environment num_env: int, default=16 number of parallel environments to initialize seed: int, default=24 random seed for environments vecenv_type: str, default="dummy" Vectorized environment type; choose from ["dummy", "subproc"]; "dummy" will create a serialized environment, which is useful for debugging; "subproc" will create parallel environments for high-throughput training Returns ------- VecEnv: vectorized environment """ def make_env(env_id, rank): env = gym.make(env_id) env.seed(seed + rank) env.action_space.seed(seed + rank) return env if vecenv_type == 'dummy': vec_env = DummyVecEnv( [lambda rank=i: make_env(env_id, rank) for i in range(num_env)] ) # type: VecEnv elif vecenv_type == 'subproc': vec_env = SubprocVecEnv( [lambda rank=i: make_env(env_id, rank) for i in range(num_env)] ) return vec_env def list_all_envs() ‑> NoneType-
List all the environments available on this system. Please see the official list of OpenAI Gym environments for more details: https://github.com/openai/gym/wiki/Table-of-environments
Expand source code
def list_all_envs() -> None: """ List all the environments available on this system. Please see the official list of OpenAI Gym environments for more details: https://github.com/openai/gym/wiki/Table-of-environments """ env_ids = [env_spec.id for env_spec in gym.envs.registry.all()] for env_id in env_ids: try: env = gym.make(env_id) print(env_id, env.observation_space, env.action_space) except Exception: # Mujoco print(env_id, "not available on this system.")