def __init__(self,
action_dim: int,
history_length: int = 4,
fc_layers: Tuple = (256, )):
super(DuelingDQNValueCNN, self).__init__()
self.action_dim = action_dim
self.conv, output_size = cnn((history_length, 16, 32))
self.advantage = mlp([output_size] + list(fc_layers) + [action_dim])
self.value = mlp([output_size] + list(fc_layers) + [1])
def __init__(self,
framestack: int,
action_dim: int,
hidden: Tuple = (32, 32),
discrete: bool = True,
*args,
**kwargs):
super(CNNPolicy, self).__init__(framestack, action_dim, hidden,
discrete, **kwargs)
self.action_dim = action_dim
self.conv, output_size = cnn((framestack, 16, 32))
self.fc = mlp([output_size] + list(hidden) + [action_dim],
sac=self.sac)
def __init__(self,
state_dim,
action_dim,
hidden=(32, 32),
disc=True,
*args,
**kwargs):
super(MlpPolicy, self).__init__(state_dim, action_dim, hidden, disc,
**kwargs)
self.state_dim = state_dim
self.action_dim = action_dim
self.model = mlp([state_dim] + list(hidden) + [action_dim],
sac=self.sac)
def __init__(self,
state_dim: int,
action_dim: int,
hidden: Tuple = (32, 32),
discrete: bool = True,
*args,
**kwargs):
super(MlpPolicy, self).__init__(state_dim, action_dim, hidden,
discrete, **kwargs)
self.state_dim = state_dim
self.action_dim = action_dim
self.activation = kwargs[
"activation"] if "activation" in kwargs else "relu"
self.model = mlp(
[state_dim] + list(hidden) + [action_dim],
activation=self.activation,
sac=self.sac,
)