app.py

import gradio as gr
import pandas as pd
from retrieval import *
import os
from transformers import AutoModelForQuestionAnswering, AutoTokenizer, pipeline

model_name = "deepset/deberta-v3-large-squad2"

# a) Get predictions
nlp = pipeline('question-answering', model=model_name, tokenizer=model_name)



hf_writer = gr.HuggingFaceDatasetSaver('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', "Pavankalyan/chitti_data")

df = pd.read_csv("Responses.csv")
text = list(df["text"].values)


def chitti(query):
    re_table = search(query, text)
    answers_re_table = [re_table[i][0] for i in range(0,5)]
    sorted_indices = sorted(range(len(answers_re_table)), key=lambda k: len(answers_re_table[k]))
    repeated_answers_indices =list()
    for i in range(4):
        if answers_re_table[sorted_indices[i]] in answers_re_table[sorted_indices[i+1]]:
            repeated_answers_indices.append(sorted_indices[i])
    for idx in repeated_answers_indices:
        answers_re_table.pop(idx)
        
    QA_input = {'question': query,'context': answers_re_table[0]}
    res1 = nlp(QA_input)['answer']
    QA_input = {'question': query,'context': answers_re_table[1]}
    res2 = nlp(QA_input)['answer']
    return [res1,answers_re_table[0],res2,answers_re_table[1]]

demo = gr.Interface(
    fn=chitti,
    inputs=["text"],
    outputs=["text","text","text","text"],
    allow_flagging = "manual",
    flagging_options = ["0","1","None"],
    flagging_callback=hf_writer
)
demo.launch()