Set temperature to elimate random responses

document_qa_engine.py

@@ -76,12 +76,13 @@ def create_inference_pipeline(document_store, model_name, api_key):
         generator = OpenAIChatGenerator(api_key=Secret.from_token("<local LLM doesn't need an API key>"),
                                         model=model_name,
                                         api_base_url="http://localhost:1234/v1",
-                                        generation_kwargs={"max_tokens": MAX_TOKENS}
+                                        generation_kwargs={"max_tokens": MAX_TOKENS},
                                         )
     elif "gpt" in model_name:
         generator = OpenAIChatGenerator(api_key=Secret.from_token(api_key), model=model_name,
-                                        generation_kwargs={"max_tokens": MAX_TOKENS},
+                                        generation_kwargs={"max_tokens": MAX_TOKENS, "temperature": 0},
                                         streaming_callback=lambda chunk: print(chunk.content, end="", flush=True),
+                                        
                                         )
     else:
         generator = HuggingFaceTGIChatGenerator(token=Secret.from_token(api_key), model=model_name,