def __init__(self,
n_input_channels,
n_dim_action,
n_hidden_channels,
n_hidden_layers,
action_space,
scale_mu=True,
normalize_input=True):
self.n_input_channels = n_input_channels
self.n_hidden_layers = n_hidden_layers
self.n_hidden_channels = n_hidden_channels
assert action_space is not None
self.scale_mu = scale_mu
self.action_space = action_space
super().__init__()
with self.init_scope():
assert n_hidden_layers >= 1
self.hidden_layers = MLPBN(in_size=n_input_channels,
out_size=n_hidden_channels,
hidden_sizes=[n_hidden_channels] *
(n_hidden_layers - 1),
normalize_input=normalize_input)
self.v = L.Linear(n_hidden_channels, 1)
self.mu = L.Linear(n_hidden_channels, n_dim_action)
self.mat_diag = L.Linear(n_hidden_channels, n_dim_action)
non_diag_size = n_dim_action * (n_dim_action - 1) // 2
if non_diag_size > 0:
self.mat_non_diag = L.Linear(n_hidden_channels, non_diag_size)
def __init__(self, n_input_channels, n_dim_action, n_hidden_channels,
n_hidden_layers, action_space, scale_mu=True,
normalize_input=True):
self.n_input_channels = n_input_channels
self.n_hidden_layers = n_hidden_layers
self.n_hidden_channels = n_hidden_channels
assert action_space is not None
self.scale_mu = scale_mu
self.action_space = action_space
layers = {}
assert n_hidden_layers >= 1
layers['hidden_layers'] = MLPBN(
in_size=n_input_channels, out_size=n_hidden_channels,
hidden_sizes=[n_hidden_channels] * (n_hidden_layers - 1),
normalize_input=normalize_input)
layers['v'] = L.Linear(n_hidden_channels, 1)
layers['mu'] = L.Linear(n_hidden_channels, n_dim_action)
layers['mat_diag'] = L.Linear(n_hidden_channels, n_dim_action)
non_diag_size = n_dim_action * (n_dim_action - 1) // 2
if non_diag_size > 0:
layers['mat_non_diag'] = L.Linear(n_hidden_channels, non_diag_size)
super().__init__(**layers)
def __init__(self,
n_dim_obs,
n_dim_action,
n_hidden_channels,
n_hidden_layers,
normalize_input=True,
nonlinearity=F.relu,
last_wscale=1.):
assert n_hidden_layers >= 1
self.n_input_channels = n_dim_obs + n_dim_action
self.n_hidden_layers = n_hidden_layers
self.n_hidden_channels = n_hidden_channels
self.normalize_input = normalize_input
self.nonlinearity = nonlinearity
super().__init__()
with self.init_scope():
# No need to pass nonlinearity to obs_mlp because it has no
# hidden layers
self.obs_mlp = MLPBN(in_size=n_dim_obs,
out_size=n_hidden_channels,
hidden_sizes=[],
normalize_input=normalize_input,
normalize_output=True)
self.mlp = MLP(
in_size=n_hidden_channels + n_dim_action,
out_size=1,
hidden_sizes=([self.n_hidden_channels] *
(self.n_hidden_layers - 1)),
nonlinearity=nonlinearity,
last_wscale=last_wscale,
)
self.output = self.mlp.output
def __init__(self, n_input_channels, n_hidden_layers,
n_hidden_channels, action_size,
min_action=None, max_action=None, bound_action=True,
normalize_input=True,
nonlinearity=F.relu,
last_wscale=1.):
self.n_input_channels = n_input_channels
self.n_hidden_layers = n_hidden_layers
self.n_hidden_channels = n_hidden_channels
self.action_size = action_size
self.min_action = min_action
self.max_action = max_action
self.bound_action = bound_action
self.normalize_input = normalize_input
if self.bound_action:
def action_filter(x):
return bound_by_tanh(
x, self.min_action, self.max_action)
else:
action_filter = None
super().__init__(
model=MLPBN(n_input_channels,
action_size,
(n_hidden_channels,) * n_hidden_layers,
normalize_input=self.normalize_input,
nonlinearity=nonlinearity,
last_wscale=last_wscale,
),
action_filter=action_filter)
def __init__(self,
n_input_channels,
n_hidden_layers,
n_hidden_channels,
action_size,
min_action=None,
max_action=None,
bound_action=True,
normalize_input=True):
self.n_input_channels = n_input_channels
self.n_hidden_layers = n_hidden_layers
self.n_hidden_channels = n_hidden_channels
self.action_size = action_size
self.min_action = min_action
self.max_action = max_action
self.bound_action = bound_action
self.normalize_input = normalize_input
if self.bound_action:
action_filter = lambda x: bound_by_tanh(x, self.min_action, self.
max_action)
else:
action_filter = None
super().__init__(model=MLPBN(n_input_channels,
action_size,
(n_hidden_channels, ) * n_hidden_layers,
normalize_input=self.normalize_input),
action_filter=action_filter)
def __init__(self,
ndim_obs,
n_actions,
n_hidden_channels,
n_hidden_layers,
nonlinearity=F.relu,
last_wscale=1.0):
super(QFunc,
self).__init__(model=MLPBN(in_size=ndim_obs,
out_size=n_actions,
hidden_sizes=[n_hidden_channels] *
n_hidden_layers))
def __init__(self,
n_dim_obs,
n_dim_action,
n_hidden_channels,
n_hidden_layers,
normalize_input=True):
self.n_input_channels = n_dim_obs + n_dim_action
self.n_hidden_layers = n_hidden_layers
self.n_hidden_channels = n_hidden_channels
self.normalize_input = normalize_input
super().__init__(obs_mlp=MLPBN(in_size=n_dim_obs,
out_size=n_hidden_channels,
hidden_sizes=[],
normalize_input=normalize_input,
normalize_output=True),
mlp=MLP(in_size=n_hidden_channels + n_dim_action,
out_size=1,
hidden_sizes=([self.n_hidden_channels] *
(self.n_hidden_layers - 1))))
self.output = self.mlp.output