An question about the architecture of DQNAgent

[tazhu2023]

Regarding the arch of DQNAgent, the last layer is "softmax" is to output the probability distribution of the actions?
And the layer before is "sigmoid", why the architecure is like this? Noramlly "sigmoid" is not required before a "softmax"? 
    
 def _build_policy_network(self):
    network = tf.keras.Sequential([
        tf.keras.layers.InputLayer(input_shape=self.observation_shape),
        tf.keras.layers.Conv1D(filters=64, kernel_size=6, padding="same", activation="tanh"),
        tf.keras.layers.MaxPooling1D(pool_size=2),
        tf.keras.layers.Conv1D(filters=32, kernel_size=3, padding="same", activation="tanh"),
        tf.keras.layers.MaxPooling1D(pool_size=2),
        tf.keras.layers.Flatten(),
        tf.keras.layers.Dense(self.n_actions, activation="sigmoid"),
        tf.keras.layers.Dense(self.n_actions, activation="softmax")
    ])

    return network