app.py

import gradio as gr
from diffusers import StableDiffusionXLPipeline, EDMEulerScheduler
from custom_pipeline import CosStableDiffusionXLInstructPix2PixPipeline
from huggingface_hub import hf_hub_download
import numpy as np
import math
import spaces 
import torch 
import sys
import random

from diffusers import StableDiffusionXLPipeline
from gradio_imageslider import ImageSlider

theme = gr.themes.Base(
    font=[gr.themes.GoogleFont('Libre Franklin'), gr.themes.GoogleFont('Public Sans'), 'system-ui', 'sans-serif'],
)

pipe = StableDiffusionXLPipeline.from_pretrained(
    "stabilityai/stable-diffusion-xl-base-1.0",
    custom_pipeline="multimodalart/sdxl_perturbed_attention_guidance",
    torch_dtype=torch.float16
)

device="cuda"
pipe = pipe.to(device)

@spaces.GPU
def run(prompt, negative_prompt="", guidance_scale=7.0, pag_scale=3.0, randomize_seed=True, seed=42, progress=gr.Progress(track_tqdm=True)):
    if(randomize_seed):
        seed = random.randint(0, sys.maxsize)
    if(prompt == ""):
        guidance_scale = 0.0
        
    generator = torch.Generator(device="cuda").manual_seed(seed)
    image_pag = pipe(prompt, guidance_scale=guidance_scale, pag_scale=3.0, pag_applied_layers=['mid'], generator=generator, num_inference_steps=25).images[0]    
    
    generator = torch.Generator(device="cuda").manual_seed(seed)
    image_normal = pipe(prompt, guidance_scale=guidance_scale, generator=generator, num_inference_steps=25).images[0]
    return image_pag, image_normal, seed

css = '''
.gradio-container{
max-width: 768px !important;
margin: 0 auto;
}
'''

with gr.Blocks(css=css, theme=theme) as demo:
    gr.Markdown('''# Perturbed Attention Guidance SDXL
    SDXL 🧨 [diffusers implementation](https://huggingface.co/multimodalart/sdxl_perturbed_attention_guidance) of [Perturbed-Attenton Guidance](https://ku-cvlab.github.io/Perturbed-Attention-Guidance/)
    ''')
  with gr.Group():
      with gr.Row():
        prompt = gr.Textbox(show_label=False, scale=4, placeholder="Your prompt", info="Leave blank to test unconditional generation")
        button = gr.Button("Generate", min_width=120)
      output = ImageSlider(label="Your result image", interactive=False)
      with gr.Accordion("Advanced Settings", open=False):
        guidance_scale = gr.Number(label="Guidance Scale", value=7.0)
        pag_scale = gr.Number(label="Pag Scale", value=3.0)
        randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
        seed = gr.Slider(minimum=1, maximum=18446744073709551615, step=1, randomize=True)
    gr.Examples(["", "an insect robot preparing a delicious meal, anime style", "a photo of a group of friends at an amusement park"])
    gr.on(
        triggers=[
            button.click,
            prompt.submit
        ],
        fn=run,
        inputs=[prompt, guidance_scale, pag_scale, seed],
        outputs=[output, seed],
    )
if __name__ == "__main__":
    demo.launch(share=True)