app.py

import streamlit as st
import torch
import pandas as pd
from transformers import AutoTokenizer,  AutoModelForSeq2SeqLM

class preProcess:
    def __init__(self, filename, titlename):
      self.filename = filename
      self.title = titlename + '\n'

    def read_data(self):
      df = pd.read_csv(self.filename)
      return df


    def check_columns(self, df):
      if (len(df.columns) > 3):
        st.error('File has more than 3 coloumns.')
        return False
      if (len(df.columns) == 0):
        st.error('File has no column.')
        return False
      else:
        return True

    def format_data(self, df):
        headers = [[] for i in range(0, len(df.columns))]
        for i in range(len(df.columns)):
            headers[i] = list(df[df.columns[i]])
        zipped = list(zip(*headers))
        res = [' '.join(map(str,tups)) for tups in zipped]
        input_format = ' labels ' + ' - '.join(list(df.columns)) +  ' values '  + ' , '.join(res)

        return input_format


    def combine_title_data(self,df):
      data = self.format_data(df)
      title_data = ' '.join([self.title,data])
      
      return title_data
      
class Model:
    def __init__(self,text,mode):
      self.padding = 'max_length'
      self.truncation = True
      self.prefix = 'C2T: '
      self.device = device = "cuda:0" if torch.cuda.is_available() else "cpu"
      self.text = text
      if mode.lower() == 'simple':
        self.tokenizer = AutoTokenizer.from_pretrained('saadob12/t5_C2T_big')
        self.model = AutoModelForSeq2SeqLM.from_pretrained('saadob12/t5_C2T_big').to(self.device)
      elif mode.lower() == 'analytical':
        self.tokenizer = AutoTokenizer.from_pretrained('saadob12/t5_C2T_autochart')
        self.model = AutoModelForSeq2SeqLM.from_pretrained('saadob12/t5_C2T_autochart').to(self.device)

    def generate(self):
      tokens = self.tokenizer.encode(self.prefix + self.text,  truncation=self.truncation, padding=self.padding, return_tensors='pt').to(self.device)
      generated = self.model.generate(tokens, num_beams=4, max_length=256)
      tgt_text = self.tokenizer.decode(generated[0], skip_special_tokens=True, clean_up_tokenization_spaces=True)
      summary = str(tgt_text).strip('[]""')
      return summary
      
      
      
def main():
  '''
  pre = preProcess('test.csv', 'Comparison between two models')
  contents = pre.read_data()
  check = pre.check_columns(contents)
  if check:
    title_data = pre.combine_title_data(contents)
    print(title_data)
    model = Model(title_data, 'simple')
    summary = model.generate()'''
  uploaded_file = st.file_uploader("Choose a file")
  
  
  
if __name__ == "__main__":
    main()