5dafbde8a22a81827514d41513f32ea9be63a9b2,ml/rl/test/gym/open_ai_gym_environment.py,OpenAIGymEnvironment,policy,#OpenAIGymEnvironment#Any#Any#Any#,108

Before Change

                action_str = predictor.discrete_action_policy(next_state_dict)[1]
                action_idx = self.actions.index(action_str.decode("utf-8"))
            else:
                action_score_dict = predictor.predict(next_state_dict)[0]
                best_action = None
                best_score = None
                for action_name, action_score in action_score_dict.items():

After Change

                action_str = predictor.discrete_action_policy(next_state_dict)[1]
                action_idx = self.actions.index(action_str.decode("utf-8"))
            elif isinstance(predictor, ContinuousActionDQNPredictor):
                normed_action_keys = sorted(self.normalization_action.keys())
                best_action = None
                best_score = None
                for action_key in normed_action_keys:
                    action_score = predictor.predict(
                        next_state_dict, [{action_key: 1}])[0]["Q"]
                    if best_action is None or best_score < action_score:
                        best_action = action_key
                        best_score = action_score
                action_idx = normed_action_keys.index(best_action)
        action[action_idx] = 1.0
        return action

Instances