app.py

import gradio as gr
import spaces
from diffusers import StableDiffusionControlNetPipeline, ControlNetModel, UniPCMultistepScheduler
import torch

# Initialize models outside the GPU function
controlnet = ControlNetModel.from_pretrained(
    "lllyasviel/control_v11p_sd15_openpose", torch_dtype=torch.float16
)

pipe = StableDiffusionControlNetPipeline.from_pretrained(
    "runwayml/stable-diffusion-v1-5",
    controlnet=controlnet,
    torch_dtype=torch.float16,
    safety_checker=None
)
pipe.scheduler = UniPCMultistepScheduler.from_config(pipe.scheduler.config)

# Move model to GPU inside the decorated function
@spaces.GPU(duration=60)  # Request GPU for 60 seconds per call
def generate(image, prompt="a person posing"):
    pipe.to("cuda")
    result = pipe(prompt=prompt, image=image, num_inference_steps=20).images[0]
    return result

demo = gr.Interface(
    fn=generate,
    inputs=[gr.Image(type="pil"), gr.Textbox(label="Prompt", value="a person posing")],
    outputs="image",
    title="Pose Generator",
    description="Upload an image and enter a prompt to generate a ControlNet-based pose output."
)

if __name__ == "__main__":
    demo.launch()