app.py

import spaces, torch
from diffusers import DiffusionPipeline
import gradio as gr

pipe = DiffusionPipeline.from_pretrained('ptx0/terminus-xl-velocity-v2', torch_dtype=torch.float16)
pipe.unet = torch.compile(pipe.unet, mode='reduce-overhead', fullgraph=True)
pipe.to('cuda')

@spaces.GPU
def generate(prompt):
    return pipe(
        prompt,
        guidance_rescale=0.7,
        guidance_scale=7.5,
        num_inference_steps=25,
    ).images

gr.Interface(
    fn=generate,
    inputs=gr.Text(),
    outputs=gr.Gallery(),
).launch()