xl_sum_gen.py

import re
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
import json

class NLPFactGenerator:
    def __init__(self, model_name="csebuetnlp/mT5_multilingual_XLSum"):
        self.max_length = 1024
        self.tokenizer = AutoTokenizer.from_pretrained(model_name)
        self.model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
        self.WHITESPACE_HANDLER = lambda k: re.sub('\s+', ' ', re.sub('\n+', ' ', k.strip()))
        self.sentences_list = []
        self.justification_list = []
        self.titles_list = []
        self.labels_list = []
        self.claim_list = []
    
    def load_data(self, filename):
        with open(filename, "r") as infile:
            self.data = json.load(infile)

    def preprocess_data(self):
        max_seq_length = 1024
        for entry in self.data:
            if "data" in entry:
                self.titles_list.append(entry["title"])
                justification = ' '.join(entry["paragraphs"])
                for evidence in self.sentences_list:
                    if len(evidence) > max_seq_length:
                        evidence = evidence[:max_seq_length]
                _evidence = ' '.join([item["sentence"] for item in entry["data"]])
                self.justification_list.append(justification)
                self.sentences_list.append(_evidence)
                self.labels_list.append(entry["label"])

    def generate_fact(self):
        max_seq_length = 1024
        generated_facts = []
        count = 0
        for evidence in self.justification_list:
            if len(evidence) > max_seq_length:
                evidence = evidence[:max_seq_length]
            input_ids = self.tokenizer(
                    [self.WHITESPACE_HANDLER(evidence)],
                    return_tensors="pt",
                    padding="max_length",
                    truncation=True,
                    max_length=1024)["input_ids"]
            try:
                output_ids = self.model.generate(
                    input_ids=input_ids,
                    max_length=128,
                    no_repeat_ngram_size=2,
                    num_beams=4)[0]
                summary = self.tokenizer.decode(
                            output_ids,
                            skip_special_tokens=True,
                            clean_up_tokenization_spaces=False)
                count+=1
                print(count)
                generated_facts.append(summary)
            except:
                print('Input ID: ', len(input_ids))
        return generated_facts


if __name__ == "__main__":
    fact_generator = NLPFactGenerator()
    fact_generator.load_data("finfact_old.json")
    fact_generator.preprocess_data()
    generated_facts = fact_generator.generate_fact()
    generated_data = []

    for title, evi, fact in zip(fact_generator.titles_list, fact_generator.sentences_list, generated_facts):
        generated_data.append({"title": title, "evidence":evi, "generated_fact": fact})
    with open("generated_facts_xlsum.json", "w") as outfile:
        json.dump(generated_data, outfile, indent=4)