app.py

# import firebase_admin
# from firebase_admin import credentials
# from firebase_admin import firestore
import io
from fastapi import FastAPI, File, UploadFile
from werkzeug.utils import secure_filename
# import speech_recognition as sr
import subprocess
import os
import requests
import random
import pandas as pd
from pydub import AudioSegment
from datetime import datetime
from datetime import date
import numpy as np
# from sklearn.ensemble import RandomForestRegressor
import shutil
import json
# from transformers import AutoModelForQuestionAnswering, AutoTokenizer, pipeline
from pydantic import BaseModel
from typing import Annotated
# from transformers import BertTokenizerFast, EncoderDecoderModel
import torch
import re
# from transformers import AutoTokenizer, T5ForConditionalGeneration
from fastapi import Form
# from transformers import AutoModelForSequenceClassification
# from transformers import TFAutoModelForSequenceClassification
# from transformers import AutoTokenizer, AutoConfig
import numpy as np
import threading
import random
import string
import time

from transformers import AutoTokenizer, AutoModelForSeq2SeqLM,pipeline

device = "cpu"

tokenizer = AutoTokenizer.from_pretrained("humarin/chatgpt_paraphraser_on_T5_base")

model = AutoModelForSeq2SeqLM.from_pretrained("humarin/chatgpt_paraphraser_on_T5_base").to(device)

def paraphrase(
    question,
    num_beams=5,
    num_beam_groups=5,
    num_return_sequences=1,
    repetition_penalty=10.0,
    diversity_penalty=3.0,
    no_repeat_ngram_size=2,
    temperature=0.7,
    max_length=10000
):
    input_ids = tokenizer(
        f'paraphrase: {question}',
        return_tensors="pt", padding="longest",
        max_length=max_length,
        truncation=True,
    ).input_ids
    
    outputs = model.generate(
        input_ids, temperature=temperature, repetition_penalty=repetition_penalty,
        num_return_sequences=num_return_sequences, no_repeat_ngram_size=no_repeat_ngram_size,
        num_beams=num_beams, num_beam_groups=num_beam_groups,
        max_length=max_length, diversity_penalty=diversity_penalty
    )

    res = tokenizer.batch_decode(outputs, skip_special_tokens=True)

    return res



class Query(BaseModel):
    text: str

   



class Query2(BaseModel):
    text: str
    host:str

   



from fastapi import FastAPI, Request, Depends, UploadFile, File
from fastapi.exceptions import HTTPException
from fastapi.middleware.cors import CORSMiddleware
from fastapi.responses import JSONResponse


# now = datetime.now()


# UPLOAD_FOLDER = '/files'
# ALLOWED_EXTENSIONS = {'txt', 'pdf', 'png',
#                       'jpg', 'jpeg', 'gif', 'ogg', 'mp3', 'wav'}


app = FastAPI()

app.add_middleware(
    CORSMiddleware,
    allow_origins=['*'],
    allow_credentials=True,
    allow_methods=['*'],
    allow_headers=['*'],
)


# cred = credentials.Certificate('key.json')
# app1 = firebase_admin.initialize_app(cred)
# db = firestore.client()
# data_frame = pd.read_csv('data.csv')



@app.on_event("startup")
async def startup_event():
   print("on startup")

   


@app.post("/")
async def get_answer(q: Query ):
    
    text = q.text
    x= paraphrase(text)
    
    return x[0]



@app.post("/large")
async def get_answer2(q: Query2 ):
    
    text = q.text
    host= q.host

    N = 20
    res = ''.join(random.choices(string.ascii_uppercase +
                             string.digits, k=N))
    res= res+ str(time.time())

    id= res
    
    t = threading.Thread(target=do_ML, args=(id,text,host))  
    t.start()

    return JSONResponse({"id":id})

    



def do_ML(id:str,text:str,host:str):
    try:
        x= paraphrase(text)
        result=x[0]
        data={"id":id,"result":result}
        x=requests.post(host,data= data)
        print(x.text)


    except:
        print("Error occured id="+id)