app.py

import os
import uuid
import gradio as gr
import json
from gradio_client import Client, handle_file
from gradio_imageslider import ImageSlider
from PIL import Image
from huggingface_hub import InferenceClient
from loadimg import load_img

with open('loras.json', 'r') as f:
    loras = json.load(f)

job = None

# Verificar las URLs de los modelos
custom_model_url = "https://fffiloni-sd-xl-custom-model.hf.space"
tile_upscaler_url = "https://gokaygokay-tileupscalerv2.hf.space"

client_custom_model = None
client_tile_upscaler = None

# try:
#     client_custom_model = Client(custom_model_url)
#     print(f"Loaded custom model from {custom_model_url}")
# except ValueError as e:
#     print(f"Failed to load custom model: {e}")

# try:
#     client_tile_upscaler = Client(tile_upscaler_url)
#     print(f"Loaded custom model from {tile_upscaler_url}")
# except ValueError as e:
#     print(f"Failed to load custom model: {e}")


def infer(selected_index, prompt, style_prompt, inf_steps, guidance_scale, width, height, seed, lora_weight, progress=gr.Progress(track_tqdm=True)):    
    try:

        global job
        if selected_index is None:
            raise gr.Error("You must select a LoRA before proceeding.")
        
        selected_lora = loras[selected_index]
        custom_model = selected_lora["repo"]
        trigger_word = selected_lora["trigger_word"]

        global client_custom_model

        if client_custom_model is None:
            try:
                client_custom_model = Client(custom_model_url)
                print(f"Loaded custom model from {custom_model_url}")
            except ValueError as e:
                print(f"Failed to load custom model: {e}")
                client_custom_model = None
                raise gr.Error("Failed to load client for " + custom_model_url)


        try:
            result = client_custom_model.submit(
                custom_model=custom_model,
                api_name="/load_model"
            )
        except ValueError as e:
            raise gr.Error(e)

        weight_name = result.result()[2]['value']

        if trigger_word and prompt.startswith(trigger_word):
            prompt = prompt[len(trigger_word+'. '):].lstrip()

        if style_prompt and prompt.endswith(style_prompt):
            prompt = prompt[:-len('. '+style_prompt)].rstrip()
        
        prompt_arr = [trigger_word, prompt, style_prompt]
        prompt = '. '.join([element.strip() for element in prompt_arr if element.strip() != ''])
        
        try:
            job = client_custom_model.submit(
                custom_model=custom_model,
                weight_name=weight_name,
                prompt=prompt,
                inf_steps=inf_steps,
                guidance_scale=guidance_scale,
                width=width,
                height=height,
                seed=seed,
                lora_weight=lora_weight,
                api_name="/infer"
            )
            result = job.result()
        except ValueError as e:
            raise gr.Error(e)


        generated_image_path = result[0]  # Esto puede necesitar ser ajustado basado en la estructura real de result
        used_seed = result[1]             # Esto puede necesitar ser ajustado basado en la estructura real de result
        used_prompt = prompt              # El prompt usado es simplemente el prompt procesado
        generated_image_path = load_img(generated_image_path, output_type="str")

        return [generated_image_path, generated_image_path], used_seed, used_prompt 

    except Exception as e:
        gr.Warning("Error: " + str(e))

def cancel_infer():
    global job
    if job:
        job.cancel()
        return "Job has been cancelled"
    return "No job to cancel"

def update_selection(evt: gr.SelectData):
    selected_lora = loras[evt.index]
    new_placeholder = f"Type a prompt for {selected_lora['title']}"
    lora_repo = selected_lora["repo"]
    updated_text = f"### Selected: [{lora_repo}](https://huggingface.co/{lora_repo}) ✨"
    return (
        gr.update(placeholder=new_placeholder),
        updated_text,
        evt.index
    )

def resize_image(image_path, reduction_factor):
    image = Image.open(image_path)
    width, height = image.size
    new_size = (width // reduction_factor, height // reduction_factor)
    resized_image = image.resize(new_size)
    return resized_image


def save_image(image):
    unique_filename = f"resized_image_{uuid.uuid4().hex}.png"
    image.save(unique_filename)
    return unique_filename


def upscale_image(image, resolution, num_inference_steps, strength, hdr, guidance_scale, controlnet_strength, scheduler_name, reduce_factor):
    global client_tile_upscaler

    image = image[1]
    try:
        client_tile_upscaler = Client(tile_upscaler_url)
        print(f"Loaded custom model from {tile_upscaler_url}")
    except ValueError as e:
        print(f"Failed to load custom model: {e}")
        client_tile_upscaler = None
        raise gr.Error("Failed to load client for " + tile_upscaler_url)
    
    if (reduce_factor > 1):
        image = resize_image(image, reduce_factor)
        image = save_image(image)


    try:
        job = client_tile_upscaler.submit(
            param_0=handle_file(image),
            param_1=resolution,
            param_2=num_inference_steps,
            param_3=strength,
            param_4=hdr,
            param_5=guidance_scale,
            param_6=controlnet_strength,
            param_7=scheduler_name,
            api_name="/wrapper"
        )
        result = job.result()
    except ValueError as e:
        raise gr.Error(e)

    return result


def refine_image(apply_refiner, image, model ,prompt, negative_prompt, num_inference_steps, guidance_scale, seed, strength):
    if (not apply_refiner):
        return image
    
    client = InferenceClient()
    refined_image = client.image_to_image(
        image[0], 
        prompt=prompt, 
        negative_prompt=negative_prompt, 
        num_inference_steps=num_inference_steps, 
        guidance_scale=guidance_scale, 
        seed=seed, 
        model=model,
        strength=strength
    )
    return [image[0], refined_image]


css="""
"""

with gr.Blocks(css=css) as demo:
    gr.Markdown("# lichorosario LoRA Portfolio")
    gr.Markdown(
        "### This is my portfolio.\n"
        "**Note**: Generation quality may vary. For best results, adjust the parameters.\n"
        "Special thanks to [@artificialguybr](https://huggingface.co/artificialguybr) and [@fffiloni](https://huggingface.co/fffiloni).\n"
        "Based on [https://huggingface.co/spaces/fffiloni/sd-xl-custom-model](https://huggingface.co/spaces/fffiloni/sd-xl-custom-model) and [https://huggingface.co/spaces/gokaygokay/TileUpscalerV2](https://huggingface.co/spaces/gokaygokay/TileUpscalerV2)"
    )

    with gr.Row():
        with gr.Column(scale=2):
            prompt_in = gr.Textbox(
                label="Your Prompt",
                info="Don't forget to include your trigger word if necessary"
            )
            style_prompt_in = gr.Textbox(
                label="Your Style Prompt"
            )
            selected_info = gr.Markdown("")
            used_prompt = gr.Textbox(
                label="Used prompt"
            )
            with gr.Column(elem_id="col-container"):
                with gr.Accordion("Advanced Settings", open=False):
                    with gr.Row():
                        inf_steps = gr.Slider(
                            label="Inference steps",
                            minimum=3,
                            maximum=150,
                            step=1,
                            value=25
                        )
                        guidance_scale = gr.Slider(
                            label="Guidance scale",
                            minimum=0.0,
                            maximum=50.0,
                            step=0.1,
                            value=12
                        )
                    with gr.Row():
                        width = gr.Slider(
                            label="Width",
                            minimum=256,
                            maximum=3072,
                            step=32,
                            value=1024,
                        )
                        height = gr.Slider(
                            label="Height",
                            minimum=256,
                            maximum=3072,
                            step=32,
                            value=512,
                        )
                                              
                        examples = [
                            [1024,512],
                            [2048,512],
                            [3072, 512]
                        ]
                        gr.Examples(
                            label="Presets",
                            examples=examples,
                            inputs=[width, height],
                            outputs=[]
                        )
            
                    with gr.Row():
                        seed = gr.Slider(
                            label="Seed",
                            info="-1 denotes a random seed",
                            minimum=-1,
                            maximum=423538377342,
                            step=1,
                            value=-1
                        )
                        last_used_seed = gr.Number(
                            label="Last used seed",
                            info="the seed used in the last generation",
                        )
                    lora_weight = gr.Slider(
                        label="LoRa weight",
                        minimum=0.0,
                        maximum=1.0,
                        step=0.01,
                        value=1.0
                    )

                with gr.Group():
                    apply_refiner = gr.Checkbox(label="Apply refiner", value=False)
                    with gr.Accordion("Refiner params", open=False) as refiner_params:
                        refiner_prompt = gr.Textbox(lines=3, label="Prompt")
                        refiner_negative_prompt = gr.Textbox(lines=3, label="Negative Prompt")
                        refiner_strength = gr.Slider(
                            label="Strength",
                            minimum=0,
                            maximum=300,
                            step=0.01,
                            value=1
                        )
                        refiner_num_inference_steps = gr.Slider(
                            label="Inference steps",
                            minimum=3,
                            maximum=300,
                            step=1,
                            value=25
                        )
                        refiner_guidance_scale = gr.Slider(
                            label="Guidance scale",
                            minimum=0.0,
                            maximum=50.0,
                            step=0.1,
                            value=12
                        )
                        refiner_seed = gr.Slider(
                            label="Seed",
                            info="-1 denotes a random seed",
                            minimum=-1,
                            maximum=423538377342,
                            step=1,
                            value=-1
                        )
                        refiner_model = gr.Textbox(label="Model", value="stabilityai/stable-diffusion-xl-refiner-1.0")

                    apply_refiner.change(
                        fn=lambda x: gr.update(visible=x),
                        inputs=apply_refiner,
                        outputs=refiner_params,
                        queue=False,
                        api_name=False,
                    )

        with gr.Column(scale=1):
            gallery = gr.Gallery(
                [(item["image"], item["title"]) for item in loras],
                label="LoRA Gallery",
                allow_preview=False,
                columns=2,
                height="100%"
            )

    submit_btn = gr.Button("Submit")
    cancel_btn = gr.Button("Cancel")


    with gr.Row():
        with gr.Column():
            generated_image = ImageSlider(label="Image / Refined Image", type="filepath")
            enhace_button = gr.Button("Enhance Image")

        with gr.Column():
            output_slider = ImageSlider(label="Before / After", type="filepath", show_download_button=False)

    with gr.Accordion("Enhacer params", open=False):
        upscale_reduce_factor = gr.Slider(minimum=1, maximum=10, step=1, label="Reduce Factor", info="1/n")
        upscale_resolution = gr.Slider(minimum=128, maximum=2048, value=1024, step=128, label="Resolution", info="Image width")
        upscale_num_inference_steps = gr.Slider(minimum=1, maximum=150, value=50, step=1, label="Number of Inference Steps")
        upscale_strength = gr.Slider(minimum=0, maximum=1, value=0.2, step=0.01, label="Strength", info="Higher values give more detail")
        upscale_hdr = gr.Slider(minimum=0, maximum=1, value=0, step=0.1, label="HDR Effect")
        upscale_guidance_scale = gr.Slider(minimum=0, maximum=20, value=12, step=0.5, label="Guidance Scale")
        upscale_controlnet_strength = gr.Slider(minimum=0.0, maximum=2.0, value=0.75, step=0.05, label="ControlNet Strength")
        upscale_scheduler_name = gr.Dropdown(
            choices=["DDIM", "DPM++ 3M SDE Karras", "DPM++ 3M Karras"],
            value="DDIM",
            label="Scheduler"
        )


    selected_index = gr.State(None)

    submit_btn.click(
        fn=infer,
        inputs=[selected_index, prompt_in, style_prompt_in, inf_steps, guidance_scale, width, height, seed, lora_weight],
        outputs=[generated_image, last_used_seed, used_prompt]
    ).then(refine_image, 
        [apply_refiner, generated_image, refiner_model, refiner_prompt, refiner_negative_prompt, refiner_num_inference_steps, refiner_guidance_scale, refiner_seed, refiner_strength],
        generated_image
    )

    cancel_btn.click(
        fn=cancel_infer,
        outputs=[]
    )


    def clear_output(image_slider):
        return None

    enhace_button.click(
        fn=clear_output,
        inputs=[output_slider],
        outputs=[output_slider]
    ).then(
        upscale_image, 
        [generated_image, upscale_resolution, upscale_num_inference_steps, upscale_strength, upscale_hdr, upscale_guidance_scale, upscale_controlnet_strength, upscale_scheduler_name, upscale_reduce_factor],
        output_slider
    )


    gallery.select(update_selection, outputs=[prompt_in, selected_info, selected_index])

demo.launch(show_error=True)