98f57a0d8781083cbca137fb9d84313b94d9d108,agents/ddpg_agent.py,DDPGAgent,choose_action,#DDPGAgent#Any#Any#,84

Before Change

    def choose_action(self, curr_state, phase=RunPhase.TRAIN):
        assert not self.env.discrete_controls, "DDPG works only for continuous control problems"
        // convert to batch so we can run it through the network
        observation = np.expand_dims(np.array(curr_state["observation"]), 0)
        result = self.actor_network.online_network.predict(observation)
        action_values = result[0].squeeze()

        if phase == RunPhase.TRAIN:

After Change

        if type(action) != np.ndarray:
            action_batch = np.array([[action]])
        inputs = self.tf_input_state(curr_state)
        inputs["action"] = action_batch
        q_value = self.critic_network.online_network.predict(inputs)[0]
        self.q_values.add_sample(q_value)
        action_info = {"action_value": q_value}

Instances