def __init__(self, venv, nstack, device=None):
self.venv = venv
self.nstack = nstack
wos = venv.observation_space # wrapped ob space
wos = obs_to_dict(wos)
self.stacked_obs = {}
new_observation_spaces = {}
self.shape_dim0 = {}
for k in wos.spaces:
self.shape_dim0[k] = wos.spaces[k].shape[0]
low = np.repeat(wos.spaces[k].low, self.nstack, axis=0)
high = np.repeat(wos.spaces[k].high, self.nstack, axis=0)
if device is None:
device = torch.device('cpu')
self.stacked_obs[k] = torch.zeros((venv.num_envs, ) +
low.shape).to(device)
new_observation_spaces[k] = gym.spaces.Box(low=low,
high=high,
dtype=np.float32)
if set(new_observation_spaces.keys()) == {None}:
VecEnvWrapper.__init__(
self, venv, observation_space=new_observation_spaces[None])
else:
VecEnvWrapper.__init__(
self,
venv,
observation_space=gym.spaces.Dict(new_observation_spaces))
def __init__(self,
venv,
ob=True,
ret=False,
clipob=5.,
cliprew=10.,
gamma=0.99,
epsilon=1e-8,
use_tf=False):
VecEnvWrapper.__init__(self, venv)
if use_tf:
from running_mean_std import TfRunningMeanStd
self.ob_rms = TfRunningMeanStd(shape=self.observation_space.shape,
scope='ob_rms') if ob else None
self.ret_rms = TfRunningMeanStd(shape=(),
scope='ret_rms') if ret else None
else:
from running_mean_std import RunningMeanStd
self.ob_rms = RunningMeanStd(
shape=self.observation_space.shape) if ob else None
self.ret_rms = RunningMeanStd(shape=()) if ret else None
self.clipob = clipob
self.cliprew = cliprew
self.ret = np.zeros(self.num_envs)
self.gamma = gamma
self.epsilon = epsilon
def __init__(self,
venv,
ob=True,
ret=True,
train=True,
noclip=False,
has_timestep=False,
ignore_mask=None,
freeze_mask=None,
time_scale=1e-3,
clipob=10.,
cliprew=10.,
gamma=0.99,
epsilon=1e-8):
VecEnvWrapper.__init__(self, venv)
self.ob_rms = RunningMeanStd(
shape=self.observation_space.shape) if ob else None
self.ret_rms = RunningMeanStd(shape=()) if ret else None
self.clipob = clipob
self.cliprew = cliprew
self.ret = np.zeros(self.num_envs)
self.train = train
self.gamma = gamma
self.epsilon = epsilon
self.noclip = noclip
self.ignore_mask = ignore_mask
self.freeze_mask = freeze_mask
self.has_timestep = has_timestep
self.time_scale = time_scale
def __init__(self, venv, nstack):
self.venv = venv
self.nstack = nstack
wos = venv.observation_space # wrapped ob space
low = np.repeat(wos.low, self.nstack, axis=-1)
high = np.repeat(wos.high, self.nstack, axis=-1)
self.stackedobs = np.zeros((venv.num_envs,) + low.shape, low.dtype)
observation_space = spaces.Box(low=low, high=high, dtype=venv.observation_space.dtype)
VecEnvWrapper.__init__(self, venv, observation_space=observation_space)
def __init__(self, venv, nstack, device):
self.venv = venv
self.nstack = nstack
wos = venv.observation_space # wrapped ob space
self.shape_dim0 = wos.low.shape[0]
low = np.repeat(wos.low, self.nstack, axis=0)
high = np.repeat(wos.high, self.nstack, axis=0)
self.stackedobs = np.zeros((venv.num_envs, ) + low.shape)
self.stackedobs = torch.from_numpy(self.stackedobs).float()
self.stackedobs = self.stackedobs.to(device)
observation_space = gym.spaces.Box(low=low,
high=high,
dtype=venv.observation_space.dtype)
VecEnvWrapper.__init__(self, venv, observation_space=observation_space)
def __init__(self, venv, nstack, device=None):
self.venv = venv
self.nstack = nstack
wos = venv.observation_space # wrapped ob space
self.shape_dim0 = wos.shape[0]
low = np.repeat(wos.low, self.nstack, axis=0)
high = np.repeat(wos.high, self.nstack, axis=0)
if device is None:
device = torch.device('cpu')
self.stacked_obs = torch.zeros((venv.num_envs, ) +
low.shape).to(device)
observation_space = gym.spaces.Box(low=low,
high=high,
dtype=venv.observation_space.dtype)
VecEnvWrapper.__init__(self, venv, observation_space=observation_space)