Move performance metrics below

app.py

@@ -95,15 +95,6 @@ def _launch_demo(args, model, tokenizer, config):
     ## Qwen-14B-Chat: A Large Language Model by Alibaba Cloud
     **Space created by [@artificialguybr](https://twitter.com/artificialguybr) based on QWEN Code. Thanks HF for GPU!**
     **Qwen is currently SOTA in the benchmarks for 14B models.**
-
-    ### Performance Metrics:
-    - **MMLU Accuracy**: 
-      - 0-shot: 64.6
-      - 5-shot: 66.5
-    - **HumanEval Pass@1**: 43.9
-    - **GSM8K Accuracy**: 
-      - 0-shot: 60.1
-      - 8-shot: 59.3
     """)
         chatbot = gr.Chatbot(label='Qwen-Chat', elem_classes="control-height", queue=True)
         query = gr.Textbox(lines=2, label='Input')
@@ -113,11 +104,20 @@ def _launch_demo(args, model, tokenizer, config):
             empty_btn = gr.Button("🧹 Clear History")
             submit_btn = gr.Button("🚀 Submit")
             regen_btn = gr.Button("🤔️ Regenerate")
-
+    
         submit_btn.click(predict, [query, chatbot, task_history], [chatbot], show_progress=True, queue=True)  # Enable queue
         submit_btn.click(reset_user_input, [], [query])
         empty_btn.click(reset_state, [chatbot, task_history], outputs=[chatbot], show_progress=True)
         regen_btn.click(regenerate, [chatbot, task_history], [chatbot], show_progress=True, queue=True)  # Enable queue
+        gr.Markdown("""### Performance Metrics:
+    - **MMLU Accuracy**: 
+      - 0-shot: 64.6
+      - 5-shot: 66.5
+    - **HumanEval Pass@1**: 43.9
+    - **GSM8K Accuracy**: 
+      - 0-shot: 60.1
+      - 8-shot: 59.3
+        """)
     demo.queue(max_size=20)
     demo.launch()