app.py

import re
import gradio as gr

import torch
from transformers import DonutProcessor, VisionEncoderDecoderModel

device = "cuda" if torch.cuda.is_available() else "cpu"
dtype = torch.float16 if torch.cuda.is_available() else torch.float32

processor = DonutProcessor.from_pretrained("sam749/donut-base-finetuned-sroie-v2")
model = VisionEncoderDecoderModel.from_pretrained("sam749/donut-base-finetuned-sroie-v2", dtype=dtype)
model.to(device)

def process_document(image):
    # prepare encoder inputs
    pixel_values = processor(image, return_tensors="pt").pixel_values
    
    # generate answer
    outputs = model.generate(
        pixel_values.to(device),
        use_cache=True,
        num_beams=1,
        max_length=128,
        bad_words_ids=[[processor.tokenizer.unk_token_id]],
        return_dict_in_generate=True,
    )
    
    # postprocess
    sequence = processor.batch_decode(outputs.sequences)[0]
    sequence = sequence.replace(processor.tokenizer.eos_token, "").replace(processor.tokenizer.pad_token, "")
    sequence = re.sub(r"<.*?>", "", sequence, count=1).strip()  # remove first task start token
    
    return processor.token2json(sequence)

description = """Gradio Demo for Donut, an instance of `VisionEncoderDecoderModel` fine-tuned on SROI (document parsing & information extraction).
To use it, simply upload your image and click 'submit', or click one of the examples to load them.
<br>
<em>Output: </em>extracts [date, company, total] from the document.
"""
article = "<p style='text-align: center'><a href='https://arxiv.org/abs/2111.15664' target='_blank'>Donut: OCR-free Document Understanding Transformer</a> | <a href='https://github.com/clovaai/donut' target='_blank'>Github Repo</a></p>"

demo = gr.Interface(
    fn=process_document,
    inputs="image",
    outputs="json",
    title="Demo: Donut 🍩 for Document Parsing",
    description=description,
    article=article,
    examples=[["example_1.png"], ["example_2.png"], ["example_3.png"]],
    cache_examples=False)

demo.launch()