b53729be002d8abca55a5b213afcffce052e3645,ml/rl/test/gym/open_ai_gym_environment.py,OpenAIGymEnvironment,policy,#OpenAIGymEnvironment#Any#Any#Any#,107

Before Change

        else:
            // Convert next_state to a list[dict[int,float]]
            next_state_dict = [{}]
            for i in range(next_state.shape[1]):
                next_state_dict[0][i] = next_state[0][i]
            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

            // For DiscreteActionPredictors use the output policy directly
            if isinstance(predictor, DiscreteActionPredictor):
                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

Instances