app.py

import os
import spaces
import torch
from diffusers import StableDiffusionXLPipeline, EulerAncestralDiscreteScheduler
import gradio as gr
import random
import tqdm

# Enable TQDM progress tracking
tqdm.monitor_interval = 0

# Load the diffusion pipeline
pipe = StableDiffusionXLPipeline.from_pretrained(
    "kayfahaarukku/UrangDiffusion-1.0", 
    torch_dtype=torch.float16, 
    custom_pipeline="lpw_stable_diffusion_xl",
    use_safetensors=True, 
)
pipe.scheduler = EulerAncestralDiscreteScheduler.from_config(pipe.scheduler.config)

# Function to generate an image
@spaces.GPU()  # Adjust the duration as needed
def generate_image(prompt, negative_prompt, use_defaults, resolution, guidance_scale, num_inference_steps, seed, randomize_seed, progress=gr.Progress()):
    pipe.to('cuda')  # Move the model to GPU when the function is called
    
    if randomize_seed:
        seed = random.randint(0, 99999999)
    if use_defaults:
        prompt = f"{prompt}, masterpiece, best quality"
        negative_prompt = f"lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, artist name, {negative_prompt}"
    generator = torch.manual_seed(seed)
    
    def callback(step, timestep, latents):
        progress(step / num_inference_steps)
        return
    
    width, height = map(int, resolution.split('x'))
    image = pipe(
        prompt, 
        negative_prompt=negative_prompt,
        width=width,
        height=height, 
        guidance_scale=guidance_scale,
        num_inference_steps=num_inference_steps,
        generator=generator,
        callback=callback,
        callback_steps=1
    ).images[0]

    torch.cuda.empty_cache()
    pipe.to('cpu')  # Move the model back to CPU after generation

    return image, seed

# Define Gradio interface
def interface_fn(prompt, negative_prompt, use_defaults, resolution, guidance_scale, num_inference_steps, seed, randomize_seed, progress=gr.Progress()):
    image, seed = generate_image(prompt, negative_prompt, use_defaults, resolution, guidance_scale, num_inference_steps, seed, randomize_seed, progress)
    return image, seed, gr.update(value=seed)

def reset_inputs():
    return gr.update(value=''), gr.update(value=''), gr.update(value=True), gr.update(value='1024x1024'), gr.update(value=7), gr.update(value=28), gr.update(value=0), gr.update(value=False)

with gr.Blocks(title="UrangDiffusion 1.0 Demo", theme="NoCrypt/[email protected]") as demo:
    gr.HTML(
        "<h1>UrangDiffusion 1.0 Demo</h1>"
        )
    with gr.Row():
        with gr.Column():
            prompt_input = gr.Textbox(lines=2, placeholder="Enter prompt here", label="Prompt")
            negative_prompt_input = gr.Textbox(lines=2, placeholder="Enter negative prompt here", label="Negative Prompt")
            use_defaults_input = gr.Checkbox(label="Use Default Quality Tags and Negative Prompt", value=True)
            resolution_input = gr.Radio(
                choices=[
                    "1024x1024", "1152x896", "896x1152", "1216x832", "832x1216",
                    "1344x768", "768x1344", "1536x640", "640x1536"
                ],
                label="Resolution",
                value="896x1152"
            )
            guidance_scale_input = gr.Slider(minimum=1, maximum=20, step=0.5, label="Guidance Scale", value=7)
            num_inference_steps_input = gr.Slider(minimum=1, maximum=100, step=1, label="Number of Inference Steps", value=28)
            seed_input = gr.Slider(minimum=0, maximum=99999999, step=1, label="Seed", value=0, interactive=True)
            randomize_seed_input = gr.Checkbox(label="Randomize Seed", value=True)
            generate_button = gr.Button("Generate")
            reset_button = gr.Button("Reset")

        with gr.Column():
            output_image = gr.Image(type="pil", label="Generated Image")

    generate_button.click(
        interface_fn,
        inputs=[
            prompt_input, negative_prompt_input, use_defaults_input, resolution_input, guidance_scale_input, num_inference_steps_input, seed_input, randomize_seed_input
        ],
        outputs=[output_image, seed_input]
    )
    
    reset_button.click(
        reset_inputs,
        inputs=[],
        outputs=[
            prompt_input, negative_prompt_input, use_defaults_input, resolution_input, guidance_scale_input, num_inference_steps_input, seed_input, randomize_seed_input
        ]
    )

demo.queue(max_size=20).launch(share=False)