diff --git a/engines/python/setup/djl_python/huggingface.py b/engines/python/setup/djl_python/huggingface.py
index 930a14643..1153f1556 100644
--- a/engines/python/setup/djl_python/huggingface.py
+++ b/engines/python/setup/djl_python/huggingface.py
@@ -204,16 +204,12 @@ def get_pipeline(self, task: str, model_id_or_path: str, kwargs):
         for element in ["load_in_8bit", "low_cpu_mem_usage"]:
             if element in kwargs:
                 use_pipeline = False
+        device = None if "device_map" in kwargs else self.device_id
         # build pipeline
         if use_pipeline:
-            if "device_map" in kwargs:
                 hf_pipeline = pipeline(task=task,
                                        model=model_id_or_path,
-                                       **kwargs)
-            else:
-                hf_pipeline = pipeline(task=task,
-                                       model=model_id_or_path,
-                                       device=self.device_id,
+                                       device=device
                                        **kwargs)
         else:
             tokenizer = AutoTokenizer.from_pretrained(model_id_or_path)
@@ -223,7 +219,7 @@ def get_pipeline(self, task: str, model_id_or_path: str, kwargs):
             hf_pipeline = pipeline(task=task,
                                    model=model,
                                    tokenizer=tokenizer,
-                                   device=self.device_id)
+                                   device=device)
 
         # wrap specific pipeline to support better ux
         if task == "conversational":