app.py

import gradio as gr
import torch
import numpy as np
import diffusers
import os
from PIL import Image
hf_token = os.environ.get("HF_TOKEN")
from diffusers import AutoPipelineForText2Image


device = "cuda" if torch.cuda.is_available() else "cpu"
pipe = AutoPipelineForText2Image.from_pretrained("briaai/BRIA-2.3", torch_dtype=torch.float16, force_zeros_for_empty_prompt=False).to(device)
pipe.load_ip_adapter("briaai/DEV-Image-Prompt", subfolder='models', weight_name="ip_adapter_bria.bin")

# default_negative_prompt= "" #"Logo,Watermark,Text,Ugly,Morbid,Extra fingers,Poorly drawn hands,Mutation,Blurry,Extra limbs,Gross proportions,Missing arms,Mutated hands,Long neck,Duplicate,Mutilated,Mutilated hands,Poorly drawn face,Deformed,Bad anatomy,Cloned face,Malformed limbs,Missing legs,Too many fingers"

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

@spaces.GPU
def predict(prompt, ip_adapter_image, ip_adapter_scale=0.5, negative_prompt="", seed=100, randomize_seed=False, center_crop=False, width=1024, height=1024, guidance_scale=5.0, num_inference_steps=50, progress=gr.Progress(track_tqdm=True)):
    if randomize_seed:
        seed = random.randint(0, MAX_SEED)

    if not center_crop:
        ip_adapter_image.resize((224,224))
    
    generator = torch.Generator(device="cuda").manual_seed(seed)
    pipe.to("cuda")
    image_encoder.to("cuda")
    pipe.image_encoder = image_encoder
    pipe.set_ip_adapter_scale([ip_adapter_scale])
    
    image = pipe(
        prompt=prompt,
        ip_adapter_image=[ip_adapter_image],
        negative_prompt=negative_prompt,
        height=height,
        width=width,
        num_inference_steps=num_inference_steps,
        guidance_scale=guidance_scale,
        num_images_per_prompt=1,
        generator=generator,
    ).images[0]
    
    return image, seed

examples = [
    ["A dog", "minta.jpeg", 0.4],
    ["A capybara", "king-min.png", 0.5],
    ["A cat", "blue_hair.png", 0.5],
    ["", "meow.jpeg", 1.0],
]

css="""
#col-container {
    margin: 0 auto;
    max-width: 720px;
}
#result img{
    object-position: top;
}
#result .image-container{
    height: 100%
}
"""
with gr.Blocks(css=css) as demo:
    with gr.Column(elem_id="col-container"):
        gr.Markdown(f"""
        # Kolors IP-Adapter - image reference and variations
        """)
        
        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)
        
        with gr.Row():
            with gr.Column():
                ip_adapter_image = gr.Image(label="IP-Adapter Image", type="pil")
                ip_adapter_scale = gr.Slider(
                    label="Image Input Scale",
                    info="Use 1 for creating image variations",
                    minimum=0.0,
                    maximum=1.0,
                    step=0.05,
                    value=0.5,
                )
            result = gr.Image(label="Result", elem_id="result")

        with gr.Accordion("Advanced Settings", open=False):
            negative_prompt = gr.Text(
                label="Negative prompt",
                max_lines=1,
                placeholder="Enter a negative prompt",
            )
            seed = gr.Slider(
                label="Seed",
                minimum=0,
                maximum=MAX_SEED,
                step=1,
                value=0,
            )
            randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
            center_crop = gr.Checkbox(label="Center Crop image", value=False, info="If not checked, the IP-Adapter image input would be resized to a square.")            
            with gr.Row():
                width = gr.Slider(
                    label="Width",
                    minimum=256,
                    maximum=2048,
                    step=32,
                    value=1024,
                )
                height = gr.Slider(
                    label="Height",
                    minimum=256,
                    maximum=2048,
                    step=32,
                    value=1024,
                )
            with gr.Row():
                guidance_scale = gr.Slider(
                    label="Guidance scale",
                    minimum=0.0,
                    maximum=10.0,
                    step=0.1,
                    value=5.0,
                )
                num_inference_steps = gr.Slider(
                    label="Number of inference steps",
                    minimum=1,
                    maximum=100,
                    step=1,
                    value=25,
                )
            
        
        gr.Examples(
            examples=examples,
            fn=predict,
            inputs=[prompt, ip_adapter_image, ip_adapter_scale],
            outputs=[result, seed],
            cache_examples="lazy"
        )

    gr.on(
        triggers=[run_button.click, prompt.submit],
        fn=predict,
        inputs=[prompt, ip_adapter_image, ip_adapter_scale, negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps],
        outputs=[result, seed]
    )

demo.queue(max_size=25,api_open=False).launch(show_api=False)

# image_blocks.queue(max_size=25,api_open=False).launch(show_api=False)