Update evaluator.py

lm_eval/evaluator.py

@@ -127,7 +127,7 @@ def simple_evaluate(
                     f"Overwriting default num_fewshot of {task_name} from {default_num_fewshot} to {num_fewshot}"
                 )
 
-            task_dict[task_name]._config["num_fewshot"] = num_fewshot
+            task_obj._config["num_fewshot"] = num_fewshot
 
     if check_integrity:
         run_task_tests(task_list=tasks)