diff --git a/scripts/finetune.py b/scripts/finetune.py
index 1da332f40..915ba1de1 100644
--- a/scripts/finetune.py
+++ b/scripts/finetune.py
@@ -228,8 +228,11 @@ def train(
     logging.info(
         f"Training Completed!!! Saving pre-trained model to {cfg.output_dir}"
     )
+    
+
     # TODO do we need this fix? https://huggingface.co/docs/accelerate/usage_guides/fsdp#saving-and-loading
-    trainer.save_model(cfg.output_dir)
+    trainer.save_pretrained(cfg.output_dir)
+    # trainer.save_model(cfg.output_dir)  # TODO this may be needed for deepspeed to work? need to review another time
 
 
 if __name__ == "__main__":