See if yield works now?

app.py

@@ -86,6 +86,7 @@ def response(state: AppState, audio: tuple):
         )
     else:
         causal_outs = state.model_outs
+    state.model_outs = None
     prev_outs = causal_outs
     start = False
     for resp, outs in diva_audio(
@@ -97,6 +98,7 @@ def response(state: AppState, audio: tuple):
             start = True
         else:
             state.conversation[-1]["content"] = resp
+        print(resp)
         yield state, state.conversation
 
     del outs.logits