app.py

from __future__ import annotations
import math
import random
# import spaces
import gradio as gr
import numpy as np
import torch
from PIL import Image
from diffusers import StableDiffusionXLImg2ImgPipeline, StableDiffusionXLPipeline, EDMEulerScheduler, StableDiffusionXLInstructPix2PixPipeline, AutoencoderKL, DPMSolverMultistepScheduler
from huggingface_hub import hf_hub_download, InferenceClient

vae = AutoencoderKL.from_pretrained("madebyollin/sdxl-vae-fp16-fix", torch_dtype=torch.float16)
pipe = StableDiffusionXLPipeline.from_pretrained("SG161222/RealVisXL_V4.0", torch_dtype=torch.float16, vae=vae)
pipe.scheduler = DPMSolverMultistepScheduler.from_config(pipe.scheduler.config, use_karras_sigmas=True, algorithm_type="sde-dpmsolver++")
pipe.to("cuda")

refiner = StableDiffusionXLImg2ImgPipeline.from_pretrained("stabilityai/stable-diffusion-xl-refiner-1.0", vae=vae, torch_dtype=torch.float16, use_safetensors=True, variant="fp16")
refiner.to("cuda")

pipe_fast = StableDiffusionXLPipeline.from_pretrained("SG161222/RealVisXL_V4.0_Lightning", torch_dtype=torch.float16, vae=vae, use_safetensors=True)
pipe_fast.to("cuda")

help_text = """
To optimize image results:
- Adjust the **Image CFG weight** if the image isn't changing enough or is changing too much. Lower it to allow bigger changes, or raise it to preserve original details.
- Modify the **Text CFG weight** to influence how closely the edit follows text instructions. Increase it to adhere more to the text, or decrease it for subtler changes.
- Experiment with different **random seeds** and **CFG values** for varied outcomes.
- **Rephrase your instructions** for potentially better results.
- **Increase the number of steps** for enhanced edits.
"""

def set_timesteps_patched(self, num_inference_steps: int, device = None):
    self.num_inference_steps = num_inference_steps
    
    ramp = np.linspace(0, 1, self.num_inference_steps)
    sigmas = torch.linspace(math.log(self.config.sigma_min), math.log(self.config.sigma_max), len(ramp)).exp().flip(0)
    
    sigmas = (sigmas).to(dtype=torch.float32, device=device)
    self.timesteps = self.precondition_noise(sigmas)
    
    self.sigmas = torch.cat([sigmas, torch.zeros(1, device=sigmas.device)])
    self._step_index = None
    self._begin_index = None
    self.sigmas = self.sigmas.to("cpu") 

# Image Editor
edit_file = hf_hub_download(repo_id="stabilityai/cosxl", filename="cosxl_edit.safetensors")
EDMEulerScheduler.set_timesteps = set_timesteps_patched
pipe_edit = StableDiffusionXLInstructPix2PixPipeline.from_single_file( edit_file, num_in_channels=8, is_cosxl_edit=True, vae=vae, torch_dtype=torch.float16 )
pipe_edit.scheduler = EDMEulerScheduler(sigma_min=0.002, sigma_max=120.0, sigma_data=1.0, prediction_type="v_prediction")
pipe_edit.to("cuda")

client1 = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
system_instructions1 = "<|system|>\nAct as Image Prompt Generation expert, Your task is to modify prompt by USER to more better prompt for Image Generation in Stable Diffusion XL. \n Modify the user's prompt to generate a high-quality image by incorporating essential keywords and styles according to prompt if none style is mentioned than assume realistic. The optimized prompt may include keywords according to prompt for resolution (4K, HD, 16:9 aspect ratio, , etc.), image quality (cute, masterpiece, high-quality, vivid colors, intricate details, etc.), and desired art styles (realistic, anime, 3D, logo, futuristic, fantasy, etc.). Ensure the prompt is concise, yet comprehensive and choose keywords wisely, to generate an exceptional image that meets the user's expectations. \n Your task is to reply with final optimized prompt only. If you get big prompt make it concise. and Apply all keyword at last of prompt. Reply with optimized prompt only.\n<|user|>\n"

def promptifier(prompt):
    formatted_prompt = f"{system_instructions1}{prompt}\n<|assistant|>\n"
    stream = client1.text_generation(formatted_prompt, max_new_tokens=100)
    return stream

# Generator
# @spaces.GPU(duration=60, queue=False)
def king(type ,
        input_image ,
        instruction: str ,
        negative_prompt: str ="",
        enhance_prompt: bool = True,
        steps: int = 25,
        randomize_seed: bool = True,
        seed: int = 2404,
        width: int = 1024,
        height: int = 1024,
        guidance_scale: float = 6,
        fast=True,
        progress=gr.Progress(track_tqdm=True)
    ):
    if type=="Image Editing" :
        input_image = Image.open(input_image).convert('RGB')
        if randomize_seed:
            seed = random.randint(0, 999999)
        generator = torch.manual_seed(seed)
        output_image = pipe_edit(
            instruction, negative_prompt=negative_prompt, image=input_image,
            guidance_scale=guidance_scale, image_guidance_scale=1.5,
            width = input_image.width, height = input_image.height,
            num_inference_steps=steps, generator=generator, output_type="latent",
        ).images
        refine = refiner(
            prompt=f"{instruction}, 4k, hd, high quality, masterpiece",
            negative_prompt = negative_prompt,
            guidance_scale=7.5,
            num_inference_steps=steps,
            image=output_image,
            generator=generator,
        ).images[0]  
        return seed, refine
    else :
        if randomize_seed:
            seed = random.randint(0, 999999)
        generator = torch.Generator().manual_seed(seed)
        if enhance_prompt:
            print(f"BEFORE: {instruction} ")
            instruction = promptifier(instruction)
            print(f"AFTER: {instruction} ")
        guidance_scale2=(guidance_scale/2)
        if fast:
            refine = pipe_fast(prompt = instruction,
            guidance_scale = guidance_scale2, 
            num_inference_steps = int(steps/2.5),
            width = width, height = height,
            generator = generator,
            ).images[0]
        else:            
            image = pipe_fast( prompt = instruction,
            negative_prompt=negative_prompt,
            guidance_scale = guidance_scale, 
            num_inference_steps = steps, 
            width = width, height = height,
            generator = generator, output_type="latent",
            ).images 

            refine = refiner( prompt=instruction,
                    negative_prompt = negative_prompt,
                    guidance_scale = 7.5,
                    num_inference_steps=  steps,
                    image=image, generator=generator,
            ).images[0]        
        return seed, refine

client = InferenceClient()
# Prompt classifier
def response(instruction, input_image=None ):
    if input_image is None:
        output="Image Generation"
    else:
        try:
            text = instruction
            labels = ["Image Editing", "Image Generation"]
            classification = client.zero_shot_classification(text, labels, multi_label=True)
            output = classification[0]
            output = str(output)
            if "Editing" in output:
                output = "Image Editing"
            else:
                output = "Image Generation"
        except:
            if input_image is None:
                output="Image Generation"
            else:
                output="Image Editing"
    return output

css = '''
.gradio-container{max-width: 700px !important}
h1{text-align:center}
footer {
    visibility: hidden
}
'''

examples=[
        [
            "Image Generation",
            None,
            "A luxurious supercar with a unique design. The car should have a pearl white finish, and gold accents. 4k, realistic.",

        ],
        [
            "Image Editing",
            "./supercar.png",
            "make it red",

        ],
        [
            "Image Editing",
            "./red_car.png",
            "add some snow",

        ],
        [
            "Image Generation",
            None,
            "An alien grasping a sign board contain word 'ALIEN' with Neon Glow, neon, futuristic, neonpunk, neon lights",
        ],
        [
            "Image Generation",
            None,
            "Beautiful Eiffel Tower at Night",
        ],
        [
            "Image Generation",
            None,
            "Beautiful Eiffel Tower at Night",
        ],
    ]

with gr.Blocks(css=css, theme="Nymbo/Nymbo_Theme") as demo:
    gr.HTML("<center><h1>Image Gen & Auto Edit</h1></center>")
    with gr.Row():
        instruction = gr.Textbox(lines=1, label="Instruction", interactive=True)
        generate_button = gr.Button("Run", scale=0)
    with gr.Row():
        type = gr.Dropdown(["Image Generation","Image Editing"], label="Task", value="Image Generation",interactive=True)
        enhance_prompt = gr.Checkbox(label="Enhance prompt", value=False, scale=0)
        fast = gr.Checkbox(label="FAST Generation", value=True, scale=0)
        
    with gr.Row():
        input_image = gr.Image(label="Image", type='filepath', interactive=True)

    with gr.Row():
        guidance_scale = gr.Number(value=6.0, step=0.1, label="Guidance Scale", interactive=True)
        steps = gr.Number(value=25, step=1, label="Steps", interactive=True)

    with gr.Accordion("Advanced options", open=False):
        with gr.Row():
            negative_prompt = gr.Text(
                    label="Negative prompt",
                    max_lines=1,
                    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, ugly, disgusting, blurry, amputation,(face asymmetry, eyes asymmetry, deformed eyes, open mouth)",
                    visible=True)
        with gr.Row():
            width =  gr.Slider( label="Width", minimum=256, maximum=2048, step=64, value=1024)
            height =  gr.Slider( label="Height", minimum=256, maximum=2048, step=64, value=1024)
        with gr.Row():
            randomize_seed = gr.Checkbox(label="Randomize Seed", value = True, interactive=True )
            seed = gr.Number(value=2404, step=1, label="Seed", interactive=True)

    gr.Examples(
        examples=examples,
        inputs=[type,input_image, instruction],
        fn=king,
        outputs=[input_image],
        cache_examples=False,
    )

    # gr.Markdown(help_text)

    instruction.change(fn=response, inputs=[instruction,input_image], outputs=type, queue=False)

    input_image.upload(fn=response, inputs=[instruction,input_image], outputs=type, queue=False)
    
    gr.on(triggers=[
            generate_button.click,
            instruction.submit
        ],
            fn=king,
            inputs=[type,
                input_image,
                instruction,
                negative_prompt,
                enhance_prompt,
                steps,
                randomize_seed,
                seed,
                width,
                height,
                guidance_scale,
                fast,
            ],
            outputs=[seed, input_image],
          api_name = "image_gen_pro",
          queue=False
        )

demo.queue(max_size=500).launch()