import re
import gradio as gr

import torch
from functools import partial
from PIL import Image
from transformers import Pix2StructForConditionalGeneration, Pix2StructProcessor


model = Pix2StructForConditionalGeneration.from_pretrained("google/pix2struct-docvqa-large")
processor = Pix2StructProcessor.from_pretrained("google/pix2struct-docvqa-large")


device = "cuda" if torch.cuda.is_available() else "cpu"
model.to(device)

def generate(img, questions):
    global model, processor
    inputs = processor(images=[img for _ in range(len(questions))], text=questions, return_tensors="pt").to(device)
    predictions = model.generate(**inputs, max_new_tokens=256)
    return zip(questions, processor.batch_decode(predictions, skip_special_tokens=True))
  

def process_document(image, question):
    return generate(image, [question])


description = "Gradio Demo for Pix2Struct, an instance of `VisionEncoderDecoderModel` fine-tuned on DocVQA (document visual question answering). To use it, simply upload your image and type a question and click 'submit', or click one of the examples to load them. Read more at the links below. \n Note: Average Inference time 60s."
article = "<p style='text-align: center'><a href='https://www.linkedin.com/in/khadke-chetan/' target='_blank'>Chetan Khadke</a></p> |<a href='https://arxiv.org/abs/2111.15664' target='_blank'>Pix2Struct for DocVQA</a> | <a href='https://arxiv.org/pdf/2210.03347.pdf' target='_blank'>Paper link</a></p>"

demo = gr.Interface(
    fn=process_document,
    inputs=["image", "text"],
    outputs="json",
    title="Demo: Pix2Struct for DocVQA",
    description=description,
    article=article,
    enable_queue=True,
    examples=[["example_1.png", "When is the coffee break?"], ["example_2.jpeg", "What's the population of Stoddard?"]],
    cache_examples=False)

demo.launch()