import gradio as gr

gr.load(
   "models/Qwen/Qwen2.5-72B-Instruct",
   provider="hf-inference",
).launch()