diff --git a/src/llamafactory/train/ppo/trainer.py b/src/llamafactory/train/ppo/trainer.py
index 58ea83d8..3b792061 100644
--- a/src/llamafactory/train/ppo/trainer.py
+++ b/src/llamafactory/train/ppo/trainer.py
@@ -135,6 +135,7 @@ class CustomPPOTrainer(PPOTrainer, Trainer):
             dataset=train_dataset,
             data_collator=data_collator,
             lr_scheduler=scheduler,
+            optimizer=optimizer,
         )
 
         self.args = training_args