app.py

import gradio as gr
import os
import argparse
import torch
from PIL import Image
from pdf2image import convert_from_path
from donut import DonutModel

def demo_process(pdf_path):
    global model, task_prompt, task_name
    input_img = convert_from_path(pdf_path.name)
    input_img = Image.fromarray(input_img)
    output = model.inference(image=input_img, prompt=task_prompt)["predictions"][0]
    return output
parser = argparse.ArgumentParser()
parser.add_argument("--task", type=str, default="SGSInvoice")
parser.add_argument("--pretrained_path", type=str, default="uartimcs/donut-invoice-extract")
args, left_argv = parser.parse_known_args()
task_name = args.task
task_prompt = f"<s_{task_name}>"

model = DonutModel.from_pretrained("uartimcs/donut-invoice-extract")
model.eval()
demo = gr.Interface(fn=demo_process,inputs=gr.File(label="Upload PDF"),outputs="json", title=f"Donut 🍩 demonstration for `{task_name}` task",)
demo.launch()