def __init__(self, vector_dim, action_dim, hidden_units):
assert len(
hidden_units
) > 1, "if you want to use this architecture of critic network, the number of layers must greater than 1"
super().__init__()
self.feature_net = mlp(hidden_units[0:1])
self.net = mlp(hidden_units[1:], output_shape=1, out_activation=None)
self(I(shape=vector_dim), I(shape=action_dim))
def __init__(self, vector_dim, output_shape, hidden_units):
super().__init__()
self.share = mlp(hidden_units['share'], out_layer=False)
self.logits = mlp(hidden_units['logits'],
output_shape=output_shape,
out_activation=None)
self.v = mlp(hidden_units['v'], output_shape=1, out_activation=None)
self(I(shape=vector_dim))
def __init__(self, vector_dim, output_shape, hidden_units):
super().__init__()
self.share = mlp(hidden_units['share'], out_layer=False)
self.mu = mlp(hidden_units['mu'],
output_shape=output_shape,
out_activation=None)
self.log_std = mlp(hidden_units['log_std'],
output_shape=output_shape,
out_activation='tanh')
self(I(shape=vector_dim))
def __init__(self, vector_dim, action_dim, quantiles_idx, hidden_units):
super().__init__()
self.action_dim = action_dim
self.q_net_head = mlp(hidden_units['q_net'],
out_layer=False) # [B, vector_dim]
self.quantile_net = mlp(hidden_units['quantile'],
out_layer=False) # [N*B, quantiles_idx]
self.q_net_tile = mlp(
hidden_units['tile'], output_shape=action_dim,
out_activation=None) # [N*B, hidden_units['quantile'][-1]]
self(I(shape=vector_dim), I(shape=quantiles_idx))
def __init__(self, vector_dim, action_dim, atoms, hidden_units):
super().__init__()
self.action_dim = action_dim
self.atoms = atoms
self.share = mlp(hidden_units['share'], layer=Noisy, out_layer=False)
self.v = mlp(hidden_units['v'],
layer=Noisy,
output_shape=atoms,
out_activation=None)
self.adv = mlp(hidden_units['adv'],
layer=Noisy,
output_shape=action_dim * atoms,
out_activation=None)
self(I(shape=vector_dim))
def __init__(self, vector_dim, output_shape, head_num, hidden_units):
super().__init__()
self.nets = [
mlp(hidden_units, output_shape=output_shape, out_activation=None)
for _ in range(head_num)
]
self(I(shape=vector_dim))
def __init__(self, vector_dim, action_dim, nums, hidden_units):
super().__init__()
self.action_dim = action_dim
self.nums = nums
self.net = mlp(hidden_units,
output_shape=nums * action_dim,
out_activation=None)
self(I(shape=vector_dim))
def __init__(self, vector_dim, action_dim, atoms, hidden_units):
super().__init__()
self.action_dim = action_dim
self.atoms = atoms
self.net = mlp(hidden_units,
output_shape=atoms * action_dim,
out_activation='softmax')
self(I(shape=vector_dim))
def __init__(self, vector_dim, output_shape, hidden_units, is_continuous):
super().__init__()
self.is_continuous = is_continuous
out_activation = 'tanh' if self.is_continuous else None
self.net = mlp(hidden_units, act_fn='tanh', output_shape=output_shape, out_activation=out_activation, out_layer=True)
self.weights_2dim = [[i, j] for i, j in zip([vector_dim]+hidden_units, hidden_units+[output_shape])]
self.weights_nums = np.asarray(self.weights_2dim).prod(axis=-1).tolist()
self.weights_total_nums = np.asarray(self.weights_2dim).prod(axis=-1).sum() + np.asarray(hidden_units).sum() + output_shape
self(tf.keras.Input(shape=vector_dim)) # 初始化网络权重
def __init__(self,
vector_dim,
output_shape,
hidden_units,
out_activation='tanh'):
super().__init__()
self.net = mlp(hidden_units,
output_shape=output_shape,
out_activation=out_activation)
self(I(shape=vector_dim))
def __init__(self,
vector_dim,
action_dim,
options_num,
hidden_units,
is_continuous=True):
super().__init__()
self.actions_num = action_dim
self.options_num = options_num
self.share = mlp(hidden_units['share'], out_layer=False)
self.q = mlp(hidden_units['q'],
output_shape=options_num,
out_activation=None)
self.pi = mlp(hidden_units['intra_option'],
output_shape=options_num * action_dim,
out_activation='tanh' if is_continuous else None)
self.beta = mlp(hidden_units['termination'],
output_shape=options_num,
out_activation='sigmoid')
self(I(shape=vector_dim))
def __init__(self,
vector_dim,
output_shape,
options_num,
hidden_units,
out_activation=None):
super().__init__()
self.actions_num = output_shape
self.options_num = options_num
self.pi = mlp(hidden_units,
output_shape=options_num * output_shape,
out_activation=out_activation)
self(I(shape=vector_dim))
def __init__(self, vector_dim, output_shape, hidden_units):
super().__init__()
self.logits = mlp(hidden_units,
output_shape=output_shape,
out_activation=None)
self(I(shape=vector_dim))
def __init__(self, vector_dim, action_dim, hidden_units):
super().__init__()
self.net = mlp(hidden_units, output_shape=1, out_activation=None)
self(I(shape=vector_dim), I(shape=action_dim))
