Update app.py

app.py

@@ -23,8 +23,9 @@ def generate_text(input_text, history):
     output = llm(input_text_with_history, max_tokens=1024, stop=["Q:", "\n"], stream=True)
     for out in output:
      stream = copy.deepcopy(out)
-     yield print(stream["choices"][0]["text"])   
+     print(stream["choices"][0]["text"])   
     history =["init",input_text_with_history]
+    yield stream
     
 
 demo = gr.ChatInterface(generate_text)