Пример #1
0
    def __init__(self, obs_num, n_actions):
        self.head= A3C_HEAD(obs_num, n_actions)
        self.pi= policy.FCSoftmaxPolicy(
            n_actions, n_actions)
        self.v= v_function.FCVFunction(n_actions)

        super().__init__(self.head, self.pi, self.v)
Пример #2
0
 def __init__(self, n_actions):
     self.head = links.NIPSDQNHead()
     self.pi = policy.FCSoftmaxPolicy(self.head.n_output_channels,
                                      n_actions)
     self.v = v_function.FCVFunction(self.head.n_output_channels)
     self.lstm = L.LSTM(self.head.n_output_channels,
                        self.head.n_output_channels)
     super().__init__(self.head, self.lstm, self.pi, self.v)
Пример #3
0
    def __init__(self, n_input, n_actions, n_hidden):

        self.head = QFunction(n_input, n_hidden)

        self.pi = policy.FCSoftmaxPolicy(
            self.head.n_output_channels, n_actions)
        self.v = v_function.FCVFunction(self.head.n_output_channels)
        super().__init__(self.head, self.pi, self.v)
 def __init__(self, n_actions):
     self.head = ICLRACERHead(activation=guided_relu)
     self.pi = policy.FCSoftmaxPolicy(self.head.n_output_channels,
                                      n_actions)
     self.v = v_function.FCVFunction(self.head.n_output_channels)
     super().__init__(self.head, self.pi, self.v)
 def __init__(self, trial, n_actions, width=None, height=None):
     self.head = MyHead(trial, width=width, height=height)
     self.pi = policy.FCSoftmaxPolicy(self.head.n_output_channels,
                                      n_actions)
     self.v = v_function.FCVFunction(self.head.n_output_channels)
     super().__init__(self.head, self.pi, self.v)