diff --git a/06_gpu_and_ml/vllm_inference.py b/06_gpu_and_ml/vllm_inference.py
index 1b56b2cef..bbb9818ae 100644
--- a/06_gpu_and_ml/vllm_inference.py
+++ b/06_gpu_and_ml/vllm_inference.py
@@ -96,7 +96,7 @@ def __enter__(self):
 
         # Load the model. Tip: MPT models may require `trust_remote_code=true`.
         self.llm = LLM(MODEL_DIR)
-        self.template = """"<s>[INST] <<SYS>>
+        self.template = """<s>[INST] <<SYS>>
 {system}
 <</SYS>>