Update app.py

app.py

@@ -658,7 +658,7 @@ If you want to experience the normal model inference speed, you can run [[Infere
 ## Tips:
 * Check out the conversation examples (at the bottom) for inspiration.
 
-* You can adjust "Max History Rounds" to try a conversation with up to five rounds. For more turns, you can download our checkpoints from GitHub and deploy them locally for inference.
+* You can adjust "Max History Rounds" to try a conversation with up to **three rounds due to insufficient GPU memory**. For more turns, you can download our checkpoints from GitHub and deploy them locally for inference.
 
 * Our demo supports a mix of images and texts as input. You can freely upload an image or enter text, and then click on "Add Image/Text". You can repeat the former step multiple times, and click on "Submit" for model inference at last.
 
@@ -766,7 +766,7 @@ if __name__ == '__main__':
                                                step=64,
                                                interactive=True,
                                                label="Max Output Tokens")
-                    max_turns = gr.Slider(minimum=1, maximum=9, value=3, step=1, interactive=True,
+                    max_turns = gr.Slider(minimum=1, maximum=3, value=3, step=1, interactive=True,
                                           label="Max History Rounds")
                     force_img_gen = gr.Radio(choices=[True, False], value=False, label='Force Image Generation')
                     force_bbox = gr.Radio(choices=[True, False], value=False, label='Force Bounding Box')