0b47129713e8cd63c49a0c53202f2b3deac941cc,softlearning/policies/latent_space_policy.py,LatentSpacePolicy,get_action,#LatentSpacePolicy#Any#Any#Any#,175

Before Change

            best_action_index = np.argmax(q_values)

            return action_candidates[best_action_index], {}
        return self.get_actions(observation[None])[0], {}

    def get_actions(self, observations):
        Sample batch of actions based on the observations

After Change

            assert not with_log_pis, "No log pi for deterministic action"
            if with_raw_actions:
                action_candidates, raw_action_candidates = self.get_actions(
                    observations, with_raw_actions=with_raw_actions)

                q_values = self._q_function.eval(observations, action_candidates)
                best_action_index = np.argmax(q_values)

Instances