modified cpu calling

graph.py

@@ -57,7 +57,7 @@ def generate(state: Dict[str, Any]) -> Dict[str, Any]:
     )
 
     try:
-        response_raw = llm(
+        response_raw = llm.invoke(
             prompt,
             temperature=temperature,
             do_sample=False,
@@ -105,7 +105,7 @@ def build_graph(retriever) -> Any:
         max_length=512,
         device=-1  # CPU only
     )
-    llm = HuggingFacePipeline(pipeline=hf_pipe)
+    llm = HuggingFacePipeline(pipeline=hf_pipe, model_kwargs={"device": -1})
 
     graph = StateGraph(dict)
     graph.add_node("retrieve", retrieve)