handle cases where there are mulitple correct refs and use the best score

nl2bash_m.py

@@ -134,8 +134,8 @@ class nl2bash_m(evaluate.Metric):
 
                     # Calculate scores
                     cmd_score = cmd_weight * cmd_corr
-                    opt_score = opt_weight * get_score(pred_option, ref_option)
-                    arg_score = arg_weight * get_score(pred_args, ref_args)
+                    opt_score = opt_weight * self.get_score(pred_option, ref_option)
+                    arg_score = arg_weight * self.get_score(pred_args, ref_args)
 
                     score = cmd_score + opt_score + arg_score 
                     best_score = max(best_score, score)