Merge pull request #2 from MachineLearningReply/fix-streaming-error

document_qa_engine.py

@@ -80,7 +80,8 @@ def create_inference_pipeline(document_store, model_name, api_key):
                                         )
     elif "gpt" in model_name:
         generator = OpenAIChatGenerator(api_key=Secret.from_token(api_key), model=model_name,
-                                        generation_kwargs={"max_tokens": MAX_TOKENS, "stream": False}
+                                        generation_kwargs={"max_tokens": MAX_TOKENS},
+                                        streaming_callback=lambda chunk: print(chunk.content, end="", flush=True),
                                         )
     else:
         generator = HuggingFaceTGIChatGenerator(token=Secret.from_token(api_key), model=model_name,