import numpy as np
import torch
import torchvision.transforms as T
from PIL import Image
from torchvision.transforms.functional import InterpolationMode
from transformers import AutoModel, AutoTokenizer
import gradio as gr

# Dùng CPU thay vì GPU
device = torch.device("cpu")

IMAGENET_MEAN = (0.485, 0.456, 0.406)
IMAGENET_STD = (0.229, 0.224, 0.225)

def build_transform(input_size):
    transform = T.Compose([
        T.Lambda(lambda img: img.convert('RGB') if img.mode != 'RGB' else img),
        T.Resize((input_size, input_size), interpolation=InterpolationMode.BICUBIC),
        T.ToTensor(),
        T.Normalize(mean=IMAGENET_MEAN, std=IMAGENET_STD)
    ])
    return transform

def load_image(image, input_size=448):
    transform = build_transform(input_size=input_size)
    pixel_values = transform(image).unsqueeze(0)  # Thêm batch dimension
    return pixel_values

# Load model trên CPU
model = AutoModel.from_pretrained(
    "5CD-AI/Vintern-1B-v3_5",
    torch_dtype=torch.float32,  # Dùng float32 cho CPU
    low_cpu_mem_usage=True,
    trust_remote_code=True,
).eval().to(device)

tokenizer = AutoTokenizer.from_pretrained("5CD-AI/Vintern-1B-v3_5", trust_remote_code=True, use_fast=False)

def process_image(image):
    pixel_values = load_image(image).to(device)
    generation_config = dict(max_new_tokens=1024, do_sample=False, num_beams=3, repetition_penalty=2.5)

    question = "<image>\nTrích xuất toàn bộ thông tin trong ảnh và trả về dạng text."
    response, _ = model.chat(tokenizer, pixel_values, question, generation_config, history=None, return_history=True)

    return response

iface = gr.Interface(
    fn=process_image,
    inputs=gr.Image(type="pil"),
    outputs="text",
    title="Vietnamese Hand Writing ORC",
    description="Extract all the information from the image and return it in text form."
)

if __name__ == "__main__":
    iface.launch()