Add warning message for run_qa.py (#29867)

* improve: error message for best model metric

* update: raise warning instead of error

examples/pytorch/question-answering/run_qa.py

@@ -627,6 +627,14 @@ def main():
         references = [{"id": str(ex["id"]), "answers": ex[answer_column_name]} for ex in examples]
         return EvalPrediction(predictions=formatted_predictions, label_ids=references)
 
+    if data_args.version_2_with_negative:
+        accepted_best_metrics = ("exact", "f1", "HasAns_exact", "HasAns_f1")
+    else:
+        accepted_best_metrics = ("exact_match", "f1")
+
+    if training_args.load_best_model_at_end and training_args.metric_for_best_model not in accepted_best_metrics:
+        warnings.warn(f"--metric_for_best_model should be set to one of {accepted_best_metrics}")
+
     metric = evaluate.load(
         "squad_v2" if data_args.version_2_with_negative else "squad", cache_dir=model_args.cache_dir
     )