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

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

DESCRIPTION = "Animagine XL 3.0"
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/cagliostrolab/animagine-xl-3.0/blob/main/animagine-xl-3.0.safetensors",
)

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


@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 = 28,
    sampler: str = "Euler a",
    aspect_ratio_selector: str = "896 x 1152",
    style_selector: str = "(None)",
    quality_selector: str = "Standard",
    use_upscaler: bool = False,
    upscaler_strength: float = 0.55,
    upscale_by: float = 1.5,
    add_quality_tags: bool = True,
    progress=gr.Progress(track_tqdm=True),
) -> Image:
    generator = utils.seed_everything(seed)

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

    prompt = utils.add_wildcard(prompt, wildcard_files)

    prompt, negative_prompt = utils.preprocess_prompt(
        quality_prompt, quality_selector, prompt, negative_prompt, add_quality_tags
    )
    prompt, negative_prompt = utils.preprocess_prompt(
        styles, style_selector, prompt, negative_prompt
    )

    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,
        "sdxl_style": style_selector,
        "add_quality_tags": add_quality_tags,
        "quality_tags": quality_selector,
    }

    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()


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

styles = {k["name"]: (k["prompt"], k["negative_prompt"]) for k in config.style_list}
quality_prompt = {
    k["name"]: (k["prompt"], k["negative_prompt"]) for k in config.quality_prompt_list
}

wildcard_files = utils.load_wildcard_files("wildcard")

with gr.Blocks(css="style.css") as demo:
    title = gr.HTML(
        f"""<h1><span>{DESCRIPTION}</span></h1>""",
        elem_id="title",
    )
    gr.Markdown(
        f"""Gradio demo for [cagliostrolab/animagine-xl-3.0](https://huggingface.co/cagliostrolab/animagine-xl-3.0)""",
        elem_id="subtitle",
    )
    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
            )
        result = gr.Gallery(
            label="Result", 
            columns=1, 
            preview=True, 
            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",
        )
        with gr.Row():
            add_quality_tags = gr.Checkbox(
                label="Add Quality Tags", 
                value=True
            )
            quality_selector = gr.Dropdown(
                label="Quality Tags Presets",
                interactive=True,
                choices=list(quality_prompt.keys()),
                value="Standard",
            )
        style_selector = gr.Radio(
            label="Style Preset",
            container=True,
            interactive=True,
            choices=list(styles.keys()),
            value="(None)",
        )
        aspect_ratio_selector = gr.Radio(
            label="Aspect Ratio",
            choices=config.aspect_ratios,
            value="896 x 1152",
            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="Euler a",
        )
        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)
    gr.Examples(
        examples=config.examples,
        inputs=prompt,
        outputs=[result, gr_metadata],
        fn=lambda *args, **kwargs: generate(*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,
        style_selector,
        quality_selector,
        use_upscaler,
        upscaler_strength,
        upscale_by,
        add_quality_tags,
    ]

    prompt.submit(
        fn=utils.randomize_seed_fn,
        inputs=[seed, randomize_seed],
        outputs=seed,
        queue=False,
        api_name=False,
    ).then(
        fn=generate,
        inputs=inputs,
        outputs=result,
        api_name="run",
    )
    negative_prompt.submit(
        fn=utils.randomize_seed_fn,
        inputs=[seed, randomize_seed],
        outputs=seed,
        queue=False,
        api_name=False,
    ).then(
        fn=generate,
        inputs=inputs,
        outputs=result,
        api_name=False,
    )
    run_button.click(
        fn=utils.randomize_seed_fn,
        inputs=[seed, randomize_seed],
        outputs=seed,
        queue=False,
        api_name=False,
    ).then(
        fn=generate,
        inputs=inputs,
        outputs=[result, gr_metadata],
        api_name=False,
    )
demo.queue(max_size=20).launch(debug=IS_COLAB, share=IS_COLAB)