datainsight1/Medical_Prescriptions

1 2import re3import gradio as gr4 5import torch6from transformers import DonutProcessor, VisionEncoderDecoderModel7 8#processor = DonutProcessor.from_pretrained("naver-clova-ix/donut-base-finetuned-cord-v2")9#model = VisionEncoderDecoderModel.from_pretrained("naver-clova-ix/donut-base-finetuned-cord-v2")10#processor = DonutProcessor.from_pretrained("Iqra56/ENGLISHDONUT")11#model = VisionEncoderDecoderModel.from_pretrained("Iqra56/ENGLISHDONUT")12processor = DonutProcessor.from_pretrained("Iqra56/DONUTWOKEYS")13model = VisionEncoderDecoderModel.from_pretrained("Iqra56/DONUTWOKEYS")14device = "cuda" if torch.cuda.is_available() else "cpu"15model.to(device)16 17def process_document(image):18    # prepare encoder inputs19    pixel_values = processor(image, return_tensors="pt").pixel_values20    21    # prepare decoder inputs22    task_prompt = "<s>"23    decoder_input_ids = processor.tokenizer(task_prompt, add_special_tokens=False, return_tensors="pt").input_ids24          25    # generate answer26    outputs = model.generate(27        pixel_values.to(device),28        decoder_input_ids=decoder_input_ids.to(device),29        max_length=model.decoder.config.max_position_embeddings,30        early_stopping=True,31        pad_token_id=processor.tokenizer.pad_token_id,32        eos_token_id=processor.tokenizer.eos_token_id,33        use_cache=True,34        num_beams=1,35        bad_words_ids=[[processor.tokenizer.unk_token_id]],36        return_dict_in_generate=True,37    )38    39    # postprocess40    sequence = processor.batch_decode(outputs.sequences)[0]41    sequence = sequence.replace(processor.tokenizer.eos_token, "").replace(processor.tokenizer.pad_token, "")42    sequence = re.sub(r"<.*?>", "", sequence, count=1).strip()  # remove first task start token43    44    return processor.token2json(sequence)45 46description = "Gradio Demo for Donut, an instance of `VisionEncoderDecoderModel` fine-tuned on CORD (document parsing). To use it, simply upload your image and click 'submit', or click one of the examples to load them. Read more at the links below."47article = "<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>"48 49demo = gr.Interface(50    fn=process_document,51    inputs="image",52    outputs="json",53    title="Demo: Donut 🍩 for Document Parsing",54    description=description,55    article=article,56    enable_queue=True,57    examples=[["Binder1_Page_48_Image_0001.png"], ["SKMBT_75122072616550_Page_50_Image_0001.png"]],58    cache_examples=False)59 60demo.launch()61