app.py

import os
import gc
import gradio as gr
import numpy as np
import torch
import json
import spaces
import config
import utils
import logging
from PIL import Image, PngImagePlugin
from datetime import datetime
from diffusers.models import AutoencoderKL
from diffusers import StableDiffusionXLPipeline, StableDiffusionXLImg2ImgPipeline
import random

logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)

if not torch.cuda.is_available():
    DESCRIPTION += "\n<p>Running on CPU 🥶 This demo does not work on CPU. </p>"
IS_COLAB = utils.is_google_colab() or os.getenv("IS_COLAB") == "1"
HF_TOKEN = os.getenv("HF_TOKEN")
CACHE_EXAMPLES = torch.cuda.is_available() and os.getenv("CACHE_EXAMPLES") == "1"
MIN_IMAGE_SIZE = int(os.getenv("MIN_IMAGE_SIZE", "512"))
MAX_IMAGE_SIZE = int(os.getenv("MAX_IMAGE_SIZE", "2048"))
USE_TORCH_COMPILE = os.getenv("USE_TORCH_COMPILE") == "1"
ENABLE_CPU_OFFLOAD = os.getenv("ENABLE_CPU_OFFLOAD") == "1"
OUTPUT_DIR = os.getenv("OUTPUT_DIR", "./outputs")

MODEL = os.getenv(
    "MODEL",
    "https://huggingface.co/AstraliteHeart/pony-diffusion-v6/blob/main/v6.safetensors",
)

DESCRIPTION = '''
<div>
<h1 style="text-align: center;">High Definition Pony Diffusion</h1>
<p>Gradio demo for PonyDiffusion v6 with image gallery, json prompt support, advanced options and more.</p>
<p>❤️ Thanks for ✨2000 visits! Heart this space if you like it!</p>
<p>🔎 For more details about me, take a look at <a href="https://sergidev.me">My website</a>.</p>
<p>🌚 For dark mode compatibility, click <a href="https://sergidev.me/hdiffusion">here</a>.</p>
</div>
'''

torch.backends.cudnn.deterministic = True
torch.backends.cudnn.benchmark = False

device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")

def load_pipeline(model_name):
    vae = AutoencoderKL.from_pretrained(
        "madebyollin/sdxl-vae-fp16-fix",
        torch_dtype=torch.float16,
    )
    pipeline = (
        StableDiffusionXLPipeline.from_single_file
        if MODEL.endswith(".safetensors")
        else StableDiffusionXLPipeline.from_pretrained
    )

    pipe = pipeline(
        model_name,
        vae=vae,
        torch_dtype=torch.float16,
        custom_pipeline="lpw_stable_diffusion_xl",
        use_safetensors=True,
        add_watermarker=False,
        use_auth_token=HF_TOKEN,
        variant="fp16",
    )

    pipe.to(device)
    return pipe

def parse_json_parameters(json_str):
    try:
        params = json.loads(json_str)
        required_keys = ['prompt', 'negative_prompt', 'resolution', 'guidance_scale', 'num_inference_steps', 'seed', 'sampler']
        for key in required_keys:
            if key not in params:
                raise ValueError(f"Missing required key: {key}")
        
        width, height = map(int, params['resolution'].split(' x '))
        
        return {
            'prompt': params['prompt'],
            'negative_prompt': params['negative_prompt'],
            'seed': params['seed'],
            'width': width,
            'height': height,
            'guidance_scale': params['guidance_scale'],
            'num_inference_steps': params['num_inference_steps'],
            'sampler': params['sampler'],
            'use_upscaler': params.get('use_upscaler', False)
        }
    except json.JSONDecodeError:
        raise ValueError("Invalid JSON format")
    except Exception as e:
        raise ValueError(f"Error parsing JSON: {str(e)}")

@spaces.GPU
def generate(
    prompt: str,
    negative_prompt: str = "",
    seed: int = 0,
    custom_width: int = 1024,
    custom_height: int = 1024,
    guidance_scale: float = 7.0,
    num_inference_steps: int = 30,
    sampler: str = "DPM++ 2M SDE Karras",
    aspect_ratio_selector: str = "1024 x 1024",
    use_upscaler: bool = False,
    upscaler_strength: float = 0.55,
    upscale_by: float = 1.5,
    json_params: str = "",
    progress=gr.Progress(track_tqdm=True),
) -> Image:
    if json_params:
        try:
            params = parse_json_parameters(json_params)
            prompt = params['prompt']
            negative_prompt = params['negative_prompt']
            seed = params['seed']
            custom_width = params['width']
            custom_height = params['height']
            guidance_scale = params['guidance_scale']
            num_inference_steps = params['num_inference_steps']
            sampler = params['sampler']
            use_upscaler = params['use_upscaler']
        except ValueError as e:
            raise gr.Error(str(e))

    generator = utils.seed_everything(seed)

    width, height = utils.aspect_ratio_handler(
        aspect_ratio_selector,
        custom_width,
        custom_height,
    )

    width, height = utils.preprocess_image_dimensions(width, height)

    backup_scheduler = pipe.scheduler
    pipe.scheduler = utils.get_scheduler(pipe.scheduler.config, sampler)

    if use_upscaler:
        upscaler_pipe = StableDiffusionXLImg2ImgPipeline(**pipe.components)
    metadata = {
        "prompt": prompt,
        "negative_prompt": negative_prompt,
        "resolution": f"{width} x {height}",
        "guidance_scale": guidance_scale,
        "num_inference_steps": num_inference_steps,
        "seed": seed,
        "sampler": sampler,
    }

    if use_upscaler:
        new_width = int(width * upscale_by)
        new_height = int(height * upscale_by)
        metadata["use_upscaler"] = {
            "upscale_method": "nearest-exact",
            "upscaler_strength": upscaler_strength,
            "upscale_by": upscale_by,
            "new_resolution": f"{new_width} x {new_height}",
        }
    else:
        metadata["use_upscaler"] = None
    logger.info(json.dumps(metadata, indent=4))

    try:
        if use_upscaler:
            latents = pipe(
                prompt=prompt,
                negative_prompt=negative_prompt,
                width=width,
                height=height,
                guidance_scale=guidance_scale,
                num_inference_steps=num_inference_steps,
                generator=generator,
                output_type="latent",
            ).images
            upscaled_latents = utils.upscale(latents, "nearest-exact", upscale_by)
            images = upscaler_pipe(
                prompt=prompt,
                negative_prompt=negative_prompt,
                image=upscaled_latents,
                guidance_scale=guidance_scale,
                num_inference_steps=num_inference_steps,
                strength=upscaler_strength,
                generator=generator,
                output_type="pil",
            ).images
        else:
            images = pipe(
                prompt=prompt,
                negative_prompt=negative_prompt,
                width=width,
                height=height,
                guidance_scale=guidance_scale,
                num_inference_steps=num_inference_steps,
                generator=generator,
                output_type="pil",
            ).images

        if images and IS_COLAB:
            for image in images:
                filepath = utils.save_image(image, metadata, OUTPUT_DIR)
                logger.info(f"Image saved as {filepath} with metadata")

        return images, metadata
    except Exception as e:
        logger.exception(f"An error occurred: {e}")
        raise
    finally:
        if use_upscaler:
            del upscaler_pipe
        pipe.scheduler = backup_scheduler
        utils.free_memory()

def fetch_recent_images(num_images=20):
    image_files = [f for f in os.listdir(OUTPUT_DIR) if f.endswith('.png')]
    image_files.sort(key=lambda x: os.path.getmtime(os.path.join(OUTPUT_DIR, x)), reverse=True)
    recent_images = []
    
    for img_file in image_files[:num_images]:
        img_path = os.path.join(OUTPUT_DIR, img_file)
        img = Image.open(img_path)
        metadata = utils.get_image_metadata(img_path)
        recent_images.append({
            "image": img_path,
            "prompt": metadata.get("prompt", ""),
            "timestamp": datetime.fromtimestamp(os.path.getmtime(img_path)).strftime("%Y-%m-%d %H:%M:%S"),
            "metadata": metadata
        })
    
    return recent_images

def update_history_list():
    return fetch_recent_images()

def handle_image_click(evt: gr.SelectData):
    selected = fetch_recent_images()[evt.index]
    return selected["image"], json.dumps(selected["metadata"], indent=2)

def generate_and_update_history(*args, **kwargs):
    images, metadata = generate(*args, **kwargs)
    utils.save_image(images[0], metadata, OUTPUT_DIR)
    return images[0], json.dumps(metadata, indent=2), update_history_list()

with open('characterfull.txt', 'r') as f:
    characters = [line.strip() for line in f.readlines()]

def get_random_character():
    return random.choice(characters)

def add_quality_tags(prompt, negative_prompt):
    positive_tags = "score_9, score_8_up, score_7_up, score_6_up, dramatic lighting"
    negative_tags = "worst quality, low quality, text, censored, deformed, bad hand, blurry, (watermark), mutated hands, monochrome"
    
    new_prompt = f"{positive_tags}, {prompt}" if prompt else positive_tags
    new_negative_prompt = f"{negative_tags}, {negative_prompt}" if negative_prompt else negative_tags
    
    return new_prompt, new_negative_prompt

if torch.cuda.is_available():
    pipe = load_pipeline(MODEL)
    logger.info("Loaded on Device!")
else:
    pipe = None

with gr.Blocks(css="style.css") as demo:
    gr.Markdown(DESCRIPTION)

    gr.DuplicateButton(
        value="Duplicate Space for private use",
        elem_id="duplicate-button",
        visible=os.getenv("SHOW_DUPLICATE_BUTTON") == "1",
    )
    with gr.Group():
        with gr.Row():
            prompt = gr.Text(
                label="Prompt",
                show_label=False,
                max_lines=5,
                placeholder="Enter your prompt",
                container=False,
            )
            run_button = gr.Button(
                "Generate", 
                variant="primary", 
                scale=0
            )
        with gr.Row():
            random_character_button = gr.Button("Random Character")
            add_quality_tags_button = gr.Button("Add quality tags")
        result = gr.Image(
            label="Result", 
            show_label=False
        )
    with gr.Accordion(label="Advanced Settings", open=False):
        negative_prompt = gr.Text(
            label="Negative Prompt",
            max_lines=5,
            placeholder="Enter a negative prompt",
            value=""
        )
        aspect_ratio_selector = gr.Radio(
            label="Aspect Ratio",
            choices=config.aspect_ratios,
            value="1024 x 1024",
            container=True,
        )
        with gr.Group(visible=False) as custom_resolution:
            with gr.Row():
                custom_width = gr.Slider(
                    label="Width",
                    minimum=MIN_IMAGE_SIZE,
                    maximum=MAX_IMAGE_SIZE,
                    step=8,
                    value=1024,
                )
                custom_height = gr.Slider(
                    label="Height",
                    minimum=MIN_IMAGE_SIZE,
                    maximum=MAX_IMAGE_SIZE,
                    step=8,
                    value=1024,
                )
        use_upscaler = gr.Checkbox(label="Use Upscaler", value=False)
        with gr.Row() as upscaler_row:
            upscaler_strength = gr.Slider(
                label="Strength",
                minimum=0,
                maximum=1,
                step=0.05,
                value=0.55,
                visible=False,
            )
            upscale_by = gr.Slider(
                label="Upscale by",
                minimum=1,
                maximum=1.5,
                step=0.1,
                value=1.5,
                visible=False,
            )
        sampler = gr.Dropdown(
            label="Sampler",
            choices=config.sampler_list,
            interactive=True,
            value="DPM++ 2M SDE Karras",
        )
        with gr.Row():
            seed = gr.Slider(
                label="Seed", minimum=0, maximum=utils.MAX_SEED, step=1, value=0
            )
            randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
        with gr.Group():
            with gr.Row():
                guidance_scale = gr.Slider(
                    label="Guidance scale",
                    minimum=1,
                    maximum=12,
                    step=0.1,
                    value=7.0,
                )
                num_inference_steps = gr.Slider(
                    label="Number of inference steps",
                    minimum=1,
                    maximum=50,
                    step=1,
                    value=28,
                )
                
    with gr.Accordion(label="Generation Parameters", open=False):
        gr_metadata = gr.JSON(label="Metadata", show_label=False)
        json_input = gr.TextArea(label="Edit/Paste JSON Parameters", placeholder="Paste or edit JSON parameters here")
        generate_from_json = gr.Button("Generate from JSON")

    with gr.Accordion("Generation History", open=False) as history_accordion:
        update_gallery_button = gr.Button("Update Gallery")
        history_gallery = gr.Gallery(
            label="History",
            show_label=False,
            elem_id="history_gallery",
            columns=5,
            rows=4,
            height="auto"
        )
        with gr.Row():
            selected_image = gr.Image(label="Selected Image", interactive=False)
            selected_metadata = gr.JSON(label="Selected Metadata", show_label=False)

    gr.Examples(
        examples=config.examples,
        inputs=prompt,
        outputs=[result, gr_metadata],
        fn=lambda *args, **kwargs: generate_and_update_history(*args, use_upscaler=True, **kwargs),
        cache_examples=CACHE_EXAMPLES,
    )

    use_upscaler.change(
        fn=lambda x: [gr.update(visible=x), gr.update(visible=x)],
        inputs=use_upscaler,
        outputs=[upscaler_strength, upscale_by],
        queue=False,
        api_name=False,
    )
    
    aspect_ratio_selector.change(
        fn=lambda x: gr.update(visible=x == "Custom"),
        inputs=aspect_ratio_selector,
        outputs=custom_resolution,
        queue=False,
        api_name=False,
    )

    inputs = [
        prompt,
        negative_prompt,
        seed,
        custom_width,
        custom_height,
        guidance_scale,
        num_inference_steps,
        sampler,
        aspect_ratio_selector,
        use_upscaler,
        upscaler_strength,
        upscale_by,
        json_input,
    ]

    prompt.submit(
        fn=utils.randomize_seed_fn,
        inputs=[seed, randomize_seed],
        outputs=seed,
        queue=False,
        api_name=False,
    ).then(
        fn=generate_and_update_history,
        inputs=inputs,
        outputs=[result, gr_metadata, history_gallery],
    )
    
    negative_prompt.submit(
        fn=utils.randomize_seed_fn,
        inputs=[seed, randomize_seed],
        outputs=seed,
        queue=False,
        api_name=False,
    ).then(
        fn=generate_and_update_history,
        inputs=inputs,
        outputs=[result, gr_metadata, history_gallery],
    )
    
    run_button.click(
        fn=utils.randomize_seed_fn,
        inputs=[seed, randomize_seed],
        outputs=seed,
        queue=False,
        api_name=False,
    ).then(
        fn=generate_and_update_history,
        inputs=inputs,
        outputs=[result, gr_metadata, history_gallery],
    )
    
    generate_from_json.click(
        fn=generate_and_update_history,
        inputs=inputs,
        outputs=[result, gr_metadata, history_gallery],
    )

    random_character_button.click(
        fn=get_random_character,
        inputs=[],
        outputs=[prompt]
    )

    add_quality_tags_button.click(
        fn=add_quality_tags,
        inputs=[prompt, negative_prompt],
        outputs=[prompt, negative_prompt]
    )

    history_gallery.select(
        fn=handle_image_click,
        inputs=[],
        outputs=[selected_image, selected_metadata]
    )

    update_gallery_button.click(
        fn=update_history_list,
        inputs=[],
        outputs=[history_gallery]
    )

demo.queue(max_size=20).launch(debug=IS_COLAB, share=IS_COLAB)