app.py

import gradio as gr
import spaces
import torch
from spaces.zero.client import _get_token

zero = torch.Tensor([0]).cuda()
print(zero.device) # <-- 'cpu' 🤔

@spaces.GPU(duration=4*60)
def inner():
    return "ok"

def greet(request: gr.Request, n):
    assert inner() ==  "ok"
    return _get_token(request)

demo = gr.Interface(fn=greet, inputs=gr.Number(), outputs=gr.Text())
demo.launch()