app.py

import gradio as gr
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
import io
from PIL import Image
import requests
import random
import dom
import os

NUM_IMAGES = 2

# Configuración del dispositivo
device = "cpu"
if torch.cuda.is_available():
    device = "cuda"
elif torch.backends.mps.is_available():
    device = "mps"

# Configuración de modelos
API_URL = "https://api-inference.huggingface.co/models/black-forest-labs/FLUX.1-dev"
headers = {"Authorization": f"Bearer {os.getenv('api_token')}"}
model_id_image_description = "vikhyatk/moondream2"
revision = "2024-08-26"

torch_dtype = torch.float32
if torch.cuda.is_available():
    torch_dtype = torch.bfloat16  # Optimización en GPU

# Carga de modelos persistente
print("Cargando modelo de descripción de imágenes...")
model_description = AutoModelForCausalLM.from_pretrained(model_id_image_description, trust_remote_code=True, revision=revision)
tokenizer_description = AutoTokenizer.from_pretrained(model_id_image_description, revision=revision)

def generate_description(image_path):
    image_test = Image.open(image_path)
    enc_image = model_description.encode_image(image_test)
    description = model_description.answer_question(enc_image, "Describe this image to create an avatar", tokenizer_description)
    return description

def query(payload):
	response = requests.post(API_URL, headers=headers, json=payload)
	return response.content

def generate_image_by_description(description, avatar_style=None):
    images = []
    for _ in range(NUM_IMAGES):
        prompt = f"Create a pigeon profile avatar. Use the following description: {description}."
        if avatar_style:
            prompt += f" Use {avatar_style} style."

        image_bytes = query({"inputs": prompt, "parameters": {"seed": random.randint(0, 1000)}})
        image = Image.open(io.BytesIO(image_bytes))
        images.append(image)
    print(images)
    return images

def process_and_generate(image, avatar_style):
    description = generate_description(image)
    return generate_image_by_description(description, avatar_style)

with gr.Blocks(js=dom.generate_title) as demo:
    with gr.Row(): 
        gr.Markdown(dom.generate_markdown)
    with gr.Row():
        with gr.Column(scale=2, min_width=300):
            selected_image = gr.Image(type="filepath", label="Upload an Image of the Pigeon", height=300)
            example_image = gr.Examples(["./examples/pigeon.webp"], label="Example Images", inputs=[selected_image])
            avatar_style = gr.Radio(
                ["Realistic", "Pixel Art", "Imaginative", "Cartoon"], 
                label="(optional) Select the avatar style:"
            )
            generate_button = gr.Button("Generate Avatar", variant="primary")
        with gr.Column(scale=2, min_width=300):
            generated_image = gr.Gallery(type="pil", label="Generated Avatar", height=300)
            
        generate_button.click(process_and_generate, inputs=[selected_image, avatar_style], outputs=generated_image)

demo.launch()