app.py

#!/usr/bin/env python

import os
import random
import uuid
import base64
import gradio as gr
import numpy as np
from PIL import Image
import spaces
import torch

from diffusers import StableDiffusionXLPipeline, EulerAncestralDiscreteScheduler

DESCRIPTION = """# DALL•E 3 XL v2 High Fi"""

def create_download_link(filename):
    with open(filename, "rb") as file:
        encoded_string = base64.b64encode(file.read()).decode('utf-8')
        download_link = f'<a href="data:image/png;base64,{encoded_string}" download="{filename}">Download Image</a>'
        return download_link
        
def save_image(img, prompt):
    unique_name = str(uuid.uuid4()) + ".png"
    img.save(unique_name)

    # save with prompt to save prompt as image file name
    filename = f"{prompt}.png"
    img.save(filename)
    return filename

def randomize_seed_fn(seed: int, randomize_seed: bool) -> int:
    if randomize_seed:
        seed = random.randint(0, MAX_SEED)
    return seed

MAX_SEED = np.iinfo(np.int32).max

if not torch.cuda.is_available():
    DESCRIPTION += "\n<p>Running on CPU 🥶 This demo may not work on CPU.</p>"

MAX_SEED = np.iinfo(np.int32).max

USE_TORCH_COMPILE = 0
ENABLE_CPU_OFFLOAD = 0

if torch.cuda.is_available():
    pipe = StableDiffusionXLPipeline.from_pretrained(
        "fluently/Fluently-XL-v4",
        torch_dtype=torch.float16,
        use_safetensors=True,
    )
    pipe.scheduler = EulerAncestralDiscreteScheduler.from_config(pipe.scheduler.config)
    
    pipe.load_lora_weights("ehristoforu/dalle-3-xl-v2", weight_name="dalle-3-xl-lora-v2.safetensors", adapter_name="dalle")
    pipe.set_adapters("dalle")

    pipe.to("cuda")

@spaces.GPU(enable_queue=True)
def generate(
    prompt: str,
    negative_prompt: str = "",
    use_negative_prompt: bool = False,
    seed: int = 0,
    width: int = 1024,
    height: int = 1024,
    guidance_scale: float = 3,
    randomize_seed: bool = False,
    progress=gr.Progress(track_tqdm=True),
):
    seed = int(randomize_seed_fn(seed, randomize_seed))

    if not use_negative_prompt:
        negative_prompt = ""  # type: ignore

    images = pipe(
        prompt=prompt,
        negative_prompt=negative_prompt,
        width=width,
        height=height,
        guidance_scale=guidance_scale,
        num_inference_steps=20,
        num_images_per_prompt=1,
        cross_attention_kwargs={"scale": 0.65},
        output_type="pil",
    ).images
    image_paths = [save_image(img, prompt) for img in images]
    download_links = [create_download_link(path) for path in image_paths]

    print(image_paths)
    return image_paths, seed, download_links

examples = [
    "An elderly man engages in a virtual reality physical therapy session, guided by a compassionate AI therapist that adapts the exercises to his abilities and provides encouragement, all from the comfort of his own home.",
    "In a bright, welcoming dental office, a young patient watches in awe as a dental robot efficiently and painlessly repairs a cavity using a laser system, while the dentist explains the procedure using interactive 3D images.",
    "A team of biomedical engineers collaborate in a state-of-the-art research facility, designing and testing advanced prosthetic limbs that seamlessly integrate with the patient's nervous system for natural, intuitive control.",
    "A pregnant woman undergoes a routine check-up, as a gentle robotic ultrasound system captures high-resolution 3D images of her developing baby, while the obstetrician provides reassurance and guidance via a holographic display.",
    "In a cutting-edge cancer treatment center, a patient undergoes a precision radiation therapy session, where an AI-guided system delivers highly targeted doses to destroy cancer cells while preserving healthy tissue.",
    "A group of medical students attend a virtual reality lecture, where they can interact with detailed, 3D anatomical models and simulate complex surgical procedures under the guidance of renowned experts from around the world.",
    "In a remote village, a local healthcare worker uses a portable, AI-powered diagnostic device to quickly and accurately assess a patient's symptoms, while seamlessly connecting with specialists in distant cities for expert advice and treatment planning.",
    "At an advanced fertility clinic, a couple watches in wonder as an AI-assisted system carefully selects the most viable embryos for implantation, while providing personalized guidance and emotional support throughout the process."
]

css = '''
.gradio-container{max-width: 1024px !important}
h1{text-align:center}
footer {
visibility: hidden
}
'''

with gr.Blocks(css=css, theme="pseudolab/huggingface-korea-theme") as demo:
    gr.Markdown(DESCRIPTION)

    with gr.Group():
        with gr.Row():
            prompt = gr.Text(
                label="Prompt",
                show_label=False,
                max_lines=1,
                placeholder="Enter your prompt",
                container=False,
            )
            run_button = gr.Button("Run", scale=0)
        result = gr.Gallery(label="Result", columns=1, preview=True, show_label=False)
        gallery = gr.Gallery(label="Gallery", show_label=False).style(columns=[2], rows=[2], object_fit="cover", height="auto")
    with gr.Accordion("Advanced options", open=False):
        use_negative_prompt = gr.Checkbox(label="Use negative prompt", value=True)
        negative_prompt = gr.Text(
            label="Negative prompt",
            lines=4,
            max_lines=6,
            value="""(deformed, distorted, disfigured:1.3), poorly drawn, bad anatomy, wrong anatomy, extra limb, missing limb, floating limbs, (mutated hands and fingers:1.4), disconnected limbs, mutation, mutated, ugly, disgusting, blurry, amputation, (NSFW:1.25)""",
            placeholder="Enter a negative prompt",
            visible=True,
        )
        seed = gr.Slider(
            label="Seed",
            minimum=0,
            maximum=MAX_SEED,
            step=1,
            value=0,
            visible=True
        )
        randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
        with gr.Row(visible=True):
            width = gr.Slider(
                label="Width",
                minimum=512,
                maximum=2048,
                step=8,
                value=1920,
            )
            height = gr.Slider(
                label="Height",
                minimum=512,
                maximum=2048,
                step=8,
                value=1080,
            )
        with gr.Row():
            guidance_scale = gr.Slider(
                label="Guidance Scale",
                minimum=0.1,
                maximum=20.0,
                step=0.1,
                value=20.0,
            )

    gr.Examples(
        examples=examples,
        inputs=prompt,
        outputs=[result, seed],
        fn=generate,
        cache_examples=False,
    )

    use_negative_prompt.change(
        fn=lambda x: gr.update(visible=x),
        inputs=use_negative_prompt,
        outputs=negative_prompt,
        api_name=False,
    )
    
    gr.on(
        triggers=[
            prompt.submit,
            negative_prompt.submit,
            run_button.click,
        ],
        fn=generate,
        inputs=[
            prompt,
            negative_prompt,
            use_negative_prompt,
            seed,
            width,
            height,
            guidance_scale,
            randomize_seed,
        ],
        outputs=[result, seed, gallery],
        api_name="run",
    )
    
if __name__ == "__main__":
    demo.queue(max_size=20).launch(show_api=False, debug=False)