try queuing for inference only.

app.py

@@ -207,6 +207,7 @@ with gr.Blocks(
                         rep_penalty,
                     ],
                     [caption_output],
+                    queue=True,
                 )
 
             gr.Markdown("""Trying prompting your input for chat; e.g. example prompt for QA, \"Question: {} Answer:\"""")
@@ -233,6 +234,7 @@ with gr.Blocks(
                             state,
                         ],
                         [chatbot, state],
+                        queue=True,
                     )
 
                     with gr.Row():
@@ -272,4 +274,4 @@ with gr.Blocks(
     )
 
 iface.queue(concurrency_count=1, api_open=False, max_size=10)
-iface.launch(enable_queue=True)
+iface.launch(enable_queue=False)