def make_env(self, env_id, seed, logger_dir=None, reward_scale=1.0, mpi_rank=0, subrank=0, info_keywords=()):
"""
Create a wrapped, monitored gym.Env for safety.
"""
scenario = scenarios.load('{}.py'.format(env_id)).Scenario()
world = scenario.make_world()
env_dict = {
"world": world,
'reset_callback': scenario.reset_world,
'reward_callback': scenario.reward,
'observation_callback': scenario.observation,
'info_callback': None,
'done_callback': scenario.done,
'shared_viewer': True
}
env = gym.make('MultiAgent-v0', **env_dict)
env.seed(seed + subrank if seed is not None else None)
env = Monitor(env,
logger_dir and os.path.join(logger_dir, str(mpi_rank) + '.' + str(subrank)),
allow_early_resets=True,
info_keywords=info_keywords)
env = ClipActionsWrapper(env)
if reward_scale != 1.0:
from baselines.common.retro_wrappers import RewardScaler
env = RewardScaler(env, reward_scale)
return env
def wrap_deepmind_n64(env, reward_scale=1 / 100.0, frame_stack=1, grayscale=False):
env = MaxAndSkipEnv(env, skip=4)
env = WarpFrame(env, width=150, height=100, grayscale=grayscale)
env = FrameStack(env, frame_stack)
env = ScaledFloatFrame(env)
env = RewardScaler(env, scale=1 / 100.0)
return env
def wrap_deepmind_n64(env, reward_scale=1 / 100.0, frame_stack=1, normalize_observations=True):
env = MaxAndSkipEnv(env, skip=4)
env = WarpFrame(env, width=450, height=300, grayscale=False)
env = ScaledFloatFrame(env)
if normalize_observations:
env = ImageNormalizer(env, mean=SSB64_IMAGE_MEAN)
env = RewardScaler(env, scale=reward_scale)
return env
def _thunk():
env = make_atari(env_id) if env_type == 'atari' else gym.make(env_id)
env.seed(seed + 10000*mpi_rank + rank if seed is not None else None)
env = Monitor(env,
logger.get_dir() and os.path.join(logger.get_dir(), str(mpi_rank) + '.' + str(rank)),
allow_early_resets=True)
if env_type == 'atari': return wrap_deepmind(env, **wrapper_kwargs)
elif reward_scale != 1: return RewardScaler(env, reward_scale)
else: return env
def make_env(env_id, seed, train=True, logger_dir=None, reward_scale=1.0):
"""
Create a wrapped, monitored gym.Env for safety.
"""
env = gym.make(env_id, **{"train":train})
env = Monitor(env, logger_dir, allow_early_resets=True, info_keywords=tuple("s"))
env.seed(seed)
if isinstance(env.action_space, gym.spaces.Box):
env = ClipActionsWrapper(env)
if reward_scale != 1.0:
from baselines.common.retro_wrappers import RewardScaler
env = RewardScaler(env, reward_scale)
return env
def make_mujoco_env(env_id, seed, reward_scale=1.0):
"""
Create a wrapped, monitored gym.Env for MuJoCo.
"""
rank = MPI.COMM_WORLD.Get_rank()
myseed = seed + 1000 * rank if seed is not None else None
set_global_seeds(myseed)
env = gym.make(env_id)
env = Monitor(env, os.path.join(logger.get_dir(), str(rank)), allow_early_resets=True)
env.seed(seed)
if reward_scale != 1.0:
from baselines.common.retro_wrappers import RewardScaler
env = RewardScaler(env, reward_scale)
return env
def wrap_n64(env,
reward_scale=1 / 100.0,
frame_skip=4,
width=150,
height=100,
grayscale=True,
normalize_observations=True):
env = MaxAndSkipEnv(env, skip=frame_skip)
env = WarpFrame(env, width=width, height=height, grayscale=grayscale)
env = ScaledFloatFrame(env)
if normalize_observations:
env = ImageNormalizer(env, mean=SSB64_IMAGE_MEAN)
env = RewardScaler(env, scale=1 / 100.0)
return env
def make_env(env_id, seed, train=True, logger_dir=None, reward_scale=1.0, mpi_rank=0, subrank=0):
"""
Create a wrapped, monitored gym.Env for safety.
"""
env = gym.make(env_id, **{"train":train})
env.seed(seed + subrank if seed is not None else None)
env = Monitor(env,
logger_dir and os.path.join(logger_dir, str(mpi_rank) + '.' + str(subrank)),
allow_early_resets=True)
env.seed(seed)
env = ClipActionsWrapper(env)
if reward_scale != 1.0:
from baselines.common.retro_wrappers import RewardScaler
env = RewardScaler(env, reward_scale)
return env
def make_mujoco_env(env_id, seed, reward_scale=1.0):
"""Create a wrapped, monitored gym.Env for MuJoCo.
Daniel: doesn't seem to get called, at least for DDPG on HalfCheetah-v2.
"""
rank = MPI.COMM_WORLD.Get_rank()
myseed = seed + 1000 * rank if seed is not None else None
set_global_seeds(myseed)
env = gym.make(env_id)
logger_path = None if logger.get_dir() is None else os.path.join(
logger.get_dir(), str(rank))
env = Monitor(env, logger_path, allow_early_resets=True)
env.seed(seed)
if reward_scale != 1.0:
from baselines.common.retro_wrappers import RewardScaler
env = RewardScaler(env, reward_scale)
return env
def _thunk():
if env_type == 'unity':
from gym_unity.envs import UnityEnv
import random; r=random.randint(64,164)
print ("***** HELLO", mpi_rank + r)
env = UnityEnv(env_id, mpi_rank + r)
else:
env = make_atari(env_id) if env_type == 'atari' else gym.make(env_id)
env.seed(seed + 10000*mpi_rank + rank if seed is not None else None)
env = Monitor(env,
logger.get_dir() and os.path.join(logger.get_dir(), str(mpi_rank) + '.' + str(rank)),
allow_early_resets=True)
if env_type == 'atari': return wrap_deepmind(env, **wrapper_kwargs)
elif reward_scale != 1: return RewardScaler(env, reward_scale)
else: return env
def _thunk():
env = ProstheticsEnv(visualize=False)
env.seed(seed + 10000 * mpi_rank +
rank if seed is not None else None)
env = ForceDictObservation(env)
env = DictToListFull(env)
env = JSONable(env)
env = Monitor(
env,
logger.get_dir()
and os.path.join(logger.get_dir(),
str(mpi_rank) + '.' + str(rank)))
if reward_scale != 1: return RewardScaler(env, reward_scale)
else: return env
def make_env(env_id,
seed,
train=True,
logger_dir=None,
mpi_rank=0,
subrank=0,
reward_scale=1.0):
"""
Create a wrapped, monitored gym.Env for safety.
"""
env = gym.make(env_id, **{"train": train, "penalty": penalty})
env = Monitor(env,
logger_dir
and os.path.join(logger_dir,
str(mpi_rank) + '.' + str(subrank)),
allow_early_resets=True)
env.seed(seed)
if isinstance(env.action_space, gym.spaces.Box):
env = ClipActionsWrapper(env)
if reward_scale != 1.0:
from baselines.common.retro_wrappers import RewardScaler
env = RewardScaler(env, reward_scale)
return env