fix: gpu is not a valid device

model.py

@@ -21,7 +21,7 @@ inference_engine = os.getenv("INFERENCE_ENGINE", "VLLM")
 logger.debug(f"Inference engine is: '{inference_engine}'")
 
 if inference_engine == "VLLM":
-    device = torch.device("gpu")
+    device = torch.device("cuda")
 
     model_path = os.getenv("MODEL_PATH", "ibm-granite/granite-guardian-3.0-8b")
     logger.debug(f"model_path is {model_path}")