app.py

import gradio as gr
import torch 
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

tokenizer = AutoTokenizer.from_pretrained("Armandoliv/t5-small-summarizer-scitldr")

model = AutoModelForSeq2SeqLM.from_pretrained("Armandoliv/t5-small-summarizer-scitldr")
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
model = model.to(device)

def main_summarizer(text):
  max_input_length = 1024
  preprocess_text = text.strip().replace("\n"," ").replace("’", "'").strip()
  tokenized_text = tokenizer.encode(preprocess_text, return_tensors="pt", truncation=True, max_length=max_input_length,).to(device)

  summary_ids = model.generate(
              tokenized_text,
              max_length=256, 
              num_beams=8,
              repetition_penalty=3.0, 
              length_penalty=2.5, 
              early_stopping=False
          )

  output = tokenizer.decode(summary_ids[0], skip_special_tokens=True)

  return output
  
inputs = [gr.Textbox(lines=10, placeholder="Text Here...", label="Input")]
outputs = gr.Text( label="Summary")
title="Text summarisation app"
description = "This demo uses AI Models to summarize long text.\nIt focus on scientific texts."

io = gr.Interface(fn=main_summarizer, inputs=inputs, outputs=outputs, title=title, description = description,

                  css= """.gr-button-primary { background: -webkit-linear-gradient( 
                    90deg, #355764 0%, #55a8a1 100% ) !important;     background: #355764;
                        background: linear-gradient( 
                    90deg, #355764 0%, #55a8a1 100% ) !important;
                        background: -moz-linear-gradient( 90deg, #355764 0%, #55a8a1 100% ) !important;
                        background: -webkit-linear-gradient( 
                    90deg, #355764 0%, #55a8a1 100% ) !important;
                    color:white !important}"""
                  )
  
io.launch()