Hi, thanks for sharing this great work!
We’re trying to reproduce your results but can only reach about 2.5 FPS on an A100 GPU.

Could you please clarify:
• What GPU/model you used to achieve 16 FPS?
• Whether you have a specific configuration or mode for high-speed inference?

One more thing:
Line 10 in app.py should be updated to
import gradio as gr

Thanks!