diff --git a/llama_stack/providers/inline/post_training/torchtune/recipes/lora_finetuning_single_device.py b/llama_stack/providers/inline/post_training/torchtune/recipes/lora_finetuning_single_device.py
index 8e1c65b26..6da5857f4 100644
--- a/llama_stack/providers/inline/post_training/torchtune/recipes/lora_finetuning_single_device.py
+++ b/llama_stack/providers/inline/post_training/torchtune/recipes/lora_finetuning_single_device.py
@@ -448,6 +448,10 @@ class LoraFinetuningSingleDevice:
 
     async def _loss_step(self, batch: Dict[str, torch.Tensor]) -> torch.Tensor:
         # Shape [b, s], needed for the loss not the model
+        # print("tokens", batch["tokens"])
+        torch.save(batch["tokens"], "/home/markchen1015/new_alpaca_tokens.pth")
+        # print("labels", batch["labels"])
+        torch.save(batch["labels"], "/home/markchen1015/new_alpaca_labels.pth")
         labels = batch.pop("labels")
         # run model
         with self.activations_handling_ctx: