b208fa6cf8bde0ed8a315eef6656271a77ce9d83,texar/agents/dqn_agent.py,DQNAgent,get_action,#DQNAgent#Any#Any#,153

Before Change

                if action_mask[action_id]:
                    break
        else:
            for i in range(self.actions):
                if not action_mask[i]:
                    qvalue[i] -= 10.0 ** 9
            action_id = np.argmax(qvalue)
        action[action_id] = 1.0

        return action

After Change

        qvalue = self._sess.run(
            self._qnet_outputs["qvalues"],
            feed_dict={self._observ_inputs: np.array([observ]),
                       tx.global_mode(): tf.estimator.ModeKeys.PREDICT})

        action = np.zeros(shape=self._num_actions)
        if random.random() < self._exploration.get_epsilon(self.timestep):

Instances