rifkat/Uz-Text-Summarization

1import gradio as gr2import re3from transformers import AutoTokenizer, AutoModelForSeq2SeqLM4 5WHITESPACE_HANDLER = lambda k: re.sub('\s+', ' ', re.sub('\n+', ' ', k.strip()))6 7model_name = "csebuetnlp/mT5_multilingual_XLSum"8tokenizer = AutoTokenizer.from_pretrained(model_name,use_fast=False)9model = AutoModelForSeq2SeqLM.from_pretrained(model_name)10 11def generate_summary(text):12 13   input_ids = tokenizer(14    [WHITESPACE_HANDLER(text)],15    return_tensors="pt",16    padding="max_length",17    truncation=True,18    max_length=512)["input_ids"]19 20   output_ids = model.generate(21       input_ids=input_ids,22       max_length=84,23       no_repeat_ngram_size=2,24       num_beams=425   )[0]26 27   summary = tokenizer.decode(28       output_ids,29       skip_special_tokens=True,30       clean_up_tokenization_spaces=False31   )32 33   return summary34 35demo = gr.Interface(fn=generate_summary,36                    inputs=gr.Textbox(lines=10, placeholder="Matinni kiriting!"),37                    outputs=gr.Textbox(lines=4)38                    )39 40demo.launch()