policy_representation.py

class AtariNet(object):
    # ...
    # ...
    def _build(self):
        # ...
        # ...
        # action function identifier policy
        self.function_policy = tf.squeeze(tf.layers.dense(
            inputs=self.state_representation,
            units=NUM_ACTIONS,
            activation=tf.nn.softmax),
            name="function_policy")

        # action function argument policies (nonspatial)
        # action function argument placeholders (for optimization)
        self.argument_policy = dict()
        self.arguments = dict()
        for arg_type in actions.TYPES:

            # for spatial actions, represent each dimension independently
            if len(arg_type.sizes) > 1:
                if arg_type in SCREEN_TYPES:
                    units = self.screen_dimensions
                elif arg_type in MINIMAP_TYPES:
                    units = self.minimap_dimensions

                arg_policy_x = tf.layers.dense(
                    inputs=self.state_representation,
                    units=units[0],
                    activation=tf.nn.softmax)

                arg_policy_y = tf.layers.dense(
                    inputs=self.state_representation,
                    units=units[1],
                    activation=tf.nn.softmax)

                self.argument_policy[str(arg_type) + "x"] = arg_policy_x
                self.argument_policy[str(arg_type) + "y"] = arg_policy_y

                arg_placeholder_x = tf.placeholder(
                    tf.float32,
                    shape=[None, units[0]])

                arg_placeholder_y = tf.placeholder(
                    tf.float32,
                    shape=[None, units[1]])

                self.arguments[str(arg_type) + "x"] = arg_placeholder_x
                self.arguments[str(arg_type) + "y"] = arg_placeholder_y

            else:
                arg_policy = tf.layers.dense(
                    inputs=self.state_representation,
                    units=arg_type.sizes[0],
                    activation=tf.nn.softmax)

                self.argument_policy[str(arg_type)] = arg_policy

                arg_placeholder = tf.placeholder(
                    tf.float32,
                    shape=[None, arg_type.sizes[0]])

                self.arguments[str(arg_type)] = arg_placeholder
        # ...
        # ...