tiny fix

src/llamafactory/train/ppo/trainer.py

@@ -133,9 +133,9 @@ class CustomPPOTrainer(PPOTrainer, Trainer):
             ref_model=ref_model,
             tokenizer=tokenizer,
             dataset=train_dataset,
+            optimizer=optimizer,
             data_collator=data_collator,
             lr_scheduler=scheduler,
-            optimizer=optimizer,
         )
 
         self.args = training_args