app.py

import torch
import gradio as gr
from googletrans import Translator
from transformers import T5Tokenizer
from transformers import T5ForConditionalGeneration
from transformers import BartForConditionalGeneration
from transformers import BartTokenizer
from transformers import PreTrainedModel
from transformers import PreTrainedTokenizer
from transformers import AutoTokenizer


# Question launcher
class E2EQGPipeline:
    def __init__(
        self,
        model: PreTrainedModel,
        tokenizer: PreTrainedTokenizer
    ):

        self.device = "cuda" if torch.cuda.is_available() else "cpu"

        self.model = model
        self.tokenizer = tokenizer

        self.model_type = "t5"

        self.kwargs = {
            "max_length": 256,
            "num_beams": 4,
            "length_penalty": 1.5,
            "no_repeat_ngram_size": 3,
            "early_stopping": True,
        }

    def generate_questions(self, context: str):
        inputs = self._prepare_inputs_for_e2e_qg(context)

        outs = self.model.generate(
            input_ids=inputs['input_ids'].to(self.device),
            attention_mask=inputs['attention_mask'].to(self.device),
            **self.kwargs
        )

        prediction = self.tokenizer.decode(outs[0], skip_special_tokens=True)

        questions = prediction.split("<sep>")
        questions = [question.strip() for question in questions[:-1]]
        return questions

    def _prepare_inputs_for_e2e_qg(self, context):
        source_text = f"generate questions: {context}"

        inputs = self._tokenize([source_text], padding=False)

        return inputs

    def _tokenize(
        self,
        inputs,
        padding=True,
        truncation=True,
        add_special_tokens=True,
        max_length=512
    ):
        inputs = self.tokenizer.batch_encode_plus(
            inputs,
            max_length=max_length,
            add_special_tokens=add_special_tokens,
            truncation=truncation,
            padding="max_length" if padding else False,
            pad_to_max_length=padding,
            return_tensors="pt"
        )

        return inputs


qg_model = T5ForConditionalGeneration.from_pretrained('valhalla/t5-base-e2e-qg')
qg_tokenizer = T5Tokenizer.from_pretrained('valhalla/t5-base-e2e-qg')


def generate_questions(text):
    qg_final_model = E2EQGPipeline(qg_model, qg_tokenizer)
    questions = qg_final_model.generate_questions(text)
    translator = Translator()
    translated_questions = [translator.translate(question, dest='es').text for question in questions]
    return translated_questions


tokenizer = BartTokenizer.from_pretrained('facebook/bart-large-cnn')
model = BartForConditionalGeneration.from_pretrained('facebook/bart-large-cnn')


def generate_summary(text):
    inputs = tokenizer.encode("summarize: " + text, return_tensors="pt", max_length=1024, truncation=True)
    summary_ids = model.generate(inputs, max_length=150, min_length=50, length_penalty=2.0, num_beams=4, early_stopping=True)
    summary = tokenizer.decode(summary_ids[0], skip_special_tokens=True)
    return summary


# QA
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
ckpt = 'mrm8488/spanish-t5-small-sqac-for-qa'
qa_tokenizer = AutoTokenizer.from_pretrained(ckpt)
qa_model = T5ForConditionalGeneration.from_pretrained(ckpt).to(device)


def generate_question_response(question, context):
    input_text = 'question: %s  context: %s' % (question, context)
    print(input_text)
    features = qa_tokenizer([input_text], padding='max_length', truncation=True, max_length=512, return_tensors='pt')
    output = qa_model.generate(
        input_ids=features['input_ids'].to(device),
        attention_mask=features['attention_mask'].to(device),
        temperature=1.0
    )
    return qa_tokenizer.decode(output[0], skip_special_tokens=True)


class SummarizerAndQA:
    def __init__(self):
        self.input_text = ''
        self.question = ''

        self.summary = ''
        self.study_generated_questions = ''
        self.question_response = ''

    def is_text_loaded(self):
        return self.input_text != ''

    def process_summarizer(self, text):
        self.input_text = text
        return generate_summary(text)

    def process_questions(self):
        return generate_questions(self.input_text)

    def process_question_response(self, question, history):
        return generate_question_response(question, self.input_text)

summarizer_and_qa = SummarizerAndQA()

textbox_input = gr.Textbox(label="Pega el text aca:", placeholder="Texto...", lines=15)
summary_output = gr.Textbox(label="Resumen", lines=15)
questions_output = gr.Textbox(label="Preguntas de guia generadas", lines=5)
questions_generate_button = gr.Button("Generate", variant="primary", interactive=summarizer_and_qa.is_text_loaded())

summarizer_interface = gr.Interface(fn=summarizer_and_qa.process_summarizer, inputs=[textbox_input], outputs=[summary_output], allow_flagging="never")
questions_interface = gr.Interface(fn=summarizer_and_qa.process_questions, inputs=[], outputs=[questions_output], allow_flagging="never", submit_btn=questions_generate_button, live=True)
chatbot_interface = gr.ChatInterface(fn=summarizer_and_qa.process_question_response, type="messages", examples=[], title="Preguntas sobre el texto")
gr.TabbedInterface([summarizer_interface, questions_interface, chatbot_interface], ["Resumidor", "Preguntas de guia", "Chatbot"]).launch()