Update app.py

app.py

@@ -22,12 +22,12 @@ def generate_text(input_text, history):
         input_text_with_history = f"{history[-1][1]}"+ "\n"
         input_text_with_history += f"Q: {input_text}" + "\n" +" A:"
     print("new input", input_text_with_history)
-    output = llm(input_text_with_history, max_tokens=1024, stop=["Q:", "\n"], echo=True)
-
+    output = llm(input_text_with_history, max_tokens=1024, stop=["Q:", "\n"], stream=True)
     for out in output:
      stream = copy.deepcopy(out)
      print(stream["choices"][0]["text"])
-    yield stream["choices"][0]["text"]
+    yield output.text
+    
     
     history =["init",str(input_text_with_history)]