diff --git a/examples/split_learning/llm_split_learning/split_learning_trainer.py b/examples/split_learning/llm_split_learning/split_learning_trainer.py
index 88a73417f..f73a8ac44 100644
--- a/examples/split_learning/llm_split_learning/split_learning_trainer.py
+++ b/examples/split_learning/llm_split_learning/split_learning_trainer.py
@@ -102,8 +102,9 @@ def __init__(self, model=None, callbacks=None):
             self.model.resize_token_embeddings(len(self.tokenizer))
         # self.training args for huggingface training
         parser = HfArgumentParser(TrainingArguments)
+
         (self.training_args,) = parser.parse_args_into_dataclasses(
-            args=["--output_dir=/tmp", "--report_to=none"]
+            args=["--report_to=none"]
         )
 
     # Redesign the evaluation stage.
@@ -132,8 +133,6 @@ def test_model_split_learning(self, batch_size, testset, sampler=None):
 
         # save other metric information such as accuracy
         tester.log_metrics("eval", metrics)
-        tester.save_metrics("eval", metrics)
-
         return metrics["eval_accuracy"]
 
     # Redesign the training stage specific to Split Learning.
diff --git a/examples/split_learning/llm_split_learning/split_learning_wikitext2_gpt2.yml b/examples/split_learning/llm_split_learning/split_learning_wikitext2_gpt2.yml
index 1c12d146b..e10c55496 100644
--- a/examples/split_learning/llm_split_learning/split_learning_wikitext2_gpt2.yml
+++ b/examples/split_learning/llm_split_learning/split_learning_wikitext2_gpt2.yml
@@ -12,7 +12,7 @@ clients:
     do_test: false
 
     # Split learning iterations for each client
-    iteration: 20
+    iteration: 1
 
 server:
     type: split_learning