app.py

import spaces
from diffusers import AutoPipelineForImage2Image, AutoPipelineForText2Image
import torch
import os
import gradio as gr
import time
import math
from PIL import Image

import numpy as np

try:
    import intel_extension_for_pytorch as ipex
except:
    pass

SAFETY_CHECKER = os.environ.get("SAFETY_CHECKER", None)
TORCH_COMPILE = os.environ.get("TORCH_COMPILE", None)
HF_TOKEN = os.environ.get("HF_TOKEN", None)

# Device management based on available hardware
device = torch.device(
    "cuda" if torch.cuda.is_available() else "mps" if torch.backends.mps.is_available() else "cpu"
)
torch_device = device
torch_dtype = torch.float16 if device == "cuda" else torch.float32

print(f"Device: {device}")
print(f"Safety Checker: {SAFETY_CHECKER}")
print(f"Torch Compile: {TORCH_COMPILE}")

# Loading model pipelines
if SAFETY_CHECKER == "True":
    i2i_pipe = AutoPipelineForImage2Image.from_pretrained(
        "stabilityai/sdxl-turbo",
        torch_dtype=torch_dtype,
        variant="fp16" if torch_dtype == torch.float16 else "fp32",
    )
    t2i_pipe = AutoPipelineForText2Image.from_pretrained(
        "stabilityai/sdxl-turbo",
        torch_dtype=torch_dtype,
        variant="fp16" if torch_dtype == torch.float16 else "fp32",
    )
else:
    i2i_pipe = AutoPipelineForImage2Image.from_pretrained(
        "stabilityai/sdxl-turbo",
        safety_checker=None,
        torch_dtype=torch_dtype,
        variant="fp16" if torch_dtype == torch.float16 else "fp32",
    )
    t2i_pipe = AutoPipelineForText2Image.from_pretrained(
        "stabilityai/sdxl-turbo",
        safety_checker=None,
        torch_dtype=torch_dtype,
        variant="fp16" if torch_dtype == torch.float16 else "fp32",
    )

# Method for Kiwi model handling
@spaces.GPU()
def kiwi_process(prompt, seed=123123, width=512, height=512):
    """
    Custom Kiwi method for image generation using advanced AI techniques.
    """
    print(f"Generating Kiwi-style image for prompt: {prompt}")
    generator = torch.manual_seed(seed)
    result = t2i_pipe(
        prompt=prompt,
        generator=generator,
        num_inference_steps=25,  # Using more steps for finer results
        guidance_scale=7.5,  # More refined guidance
        width=width,
        height=height,
        output_type="pil",
    )
    return result.images[0]

# Resize image helper
def resize_crop(image, size=512):
    image = image.convert("RGB")
    w, h = image.size
    image = image.resize((size, int(size * (h / w))), Image.BICUBIC)
    return image

# Main prediction method
async def predict(init_image, prompt, strength, steps, seed=123123):
    if init_image is not None:
        init_image = resize_crop(init_image)
        generator = torch.manual_seed(seed)
        
        results = i2i_pipe(
            prompt=prompt,
            image=init_image,
            generator=generator,
            num_inference_steps=steps,
            guidance_scale=0.0,
            strength=strength,
            width=512,
            height=512,
            output_type="pil",
        )
    else:
        return kiwi_process(prompt, seed)  # Using the Kiwi method for text-to-image

# Gradio UI with a custom description for Kiwi
css = """
#container{
    margin: 0 auto;
    max-width: 80rem;
}
#intro{
    max-width: 100%;
    text-align: center;
    margin: 0 auto;
}
"""
with gr.Blocks(css=css) as demo:
    init_image_state = gr.State()
    with gr.Column(elem_id="container"):
        gr.Markdown(
            """# Kiwi Image Generator Demo
            ## Harnessing the Power of Kiwi AI
            This demo integrates the Kiwi AI model to generate high-quality images using cutting-edge techniques like quantization and pruning.
            """,
            elem_id="intro",
        )
        with gr.Row():
            prompt = gr.Textbox(
                placeholder="Insert your prompt for Kiwi here:",
                scale=5,
                container=False,
            )
            generate_bt = gr.Button("Generate with Kiwi", scale=1)
        with gr.Row():
            with gr.Column():
                image_input = gr.Image(
                    sources=["upload", "webcam", "clipboard"],
                    label="Upload or Capture Image",
                    type="pil",
                )
            with gr.Column():
                image = gr.Image(type="filepath")
                with gr.Accordion("Advanced options", open=False):
                    strength = gr.Slider(
                        label="Strength",
                        value=0.7,
                        minimum=0.0,
                        maximum=1.0,
                        step=0.001,
                    )
                    steps = gr.Slider(
                        label="Steps", value=25, minimum=1, maximum=50, step=1
                    )
                    seed = gr.Slider(
                        randomize=True,
                        minimum=0,
                        maximum=12013012031030,
                        label="Seed",
                        step=1,
                    )

        inputs = [image_input, prompt, strength, steps, seed]
        generate_bt.click(fn=predict, inputs=inputs, outputs=image, show_progress=False)
        prompt.change(fn=predict, inputs=inputs, outputs=image, show_progress=False)
        steps.change(fn=predict, inputs=inputs, outputs=image, show_progress=False)
        seed.change(fn=predict, inputs=inputs, outputs=image, show_progress=False)
        strength.change(fn=predict, inputs=inputs, outputs=image, show_progress=False)

demo.queue()
demo.launch()