Spaces:
Sleeping
Sleeping
| from transformers import AutoTokenizer, AutoModelForSeq2SeqLM | |
| from pathlib import Path | |
| import pdfplumber | |
| import gradio as gr | |
| def respond(pdf_file, upper_page=0): | |
| pdf_file = Path(pdf_file.name) | |
| all_text = "" | |
| with pdfplumber.open(pdf_file) as pdf: | |
| total_pages = len(pdf.pages) | |
| for idx, pdf_page in enumerate(pdf.pages): | |
| single_page_text = pdf_page.extract_text() | |
| all_text = all_text + "\n" + single_page_text | |
| #print(idx / total_pages) | |
| tokenizer=AutoTokenizer.from_pretrained('Einmalumdiewelt/T5-Base_GNAD') | |
| model=AutoModelForSeq2SeqLM.from_pretrained('Einmalumdiewelt/T5-Base_GNAD', return_dict=True) | |
| inputs=tokenizer.encode("summarize: " +all_text, return_tensors='pt', max_length=512, truncation=True) | |
| output = model.generate(inputs, min_length=70, max_length=80) | |
| summary=tokenizer.decode(output[0]) | |
| return summary, all_text | |
| with gr.Blocks() as demo: | |
| title = """<p><h1 align="center" style="font-size: 36px;">Talk with your document</h1></p>""" | |
| gr.HTML(title) | |
| with gr.Row(): | |
| with gr.Column(): | |
| file_input = gr.File(label="PDF File", type="file") | |
| page_input = gr.Textbox(label="Page Limit") | |
| summarize_button = gr.Button(label="Summarize") | |
| with gr.Column(): | |
| summary_output = gr.Textbox(label="Summarized Text") | |
| with gr.Column(): | |
| text_output =gr.Textbox(label="Extracted Text") | |
| summarize_button.click(respond, inputs=[file_input, page_input], outputs=[summary_output, text_output]) | |
| demo.launch(debug=True) |