app.py

import gradio as gr
import torch
from transformers import GPT2Tokenizer, T5ForConditionalGeneration


tokenizer = GPT2Tokenizer.from_pretrained('RussianNLP/FRED-T5-Summarizer', eos_token='</s>')
model = T5ForConditionalGeneration.from_pretrained('RussianNLP/FRED-T5-Summarizer')
device = 'cpu'
model.to(device)

input_text = "<LM> Сократи текст.\n "

def make_summarization(user_text):
    processing_text = input_text + user_text
    input_ids = torch.tensor([tokenizer.encode(processing_text)]).to(device)
    outputs = model.generate(input_ids, eos_token_id=tokenizer.eos_token_id,
                             num_beams=3,
                             min_new_tokens=17,
                             max_new_tokens=200,
                             do_sample=True,
                             no_repeat_ngram_size=4,
                             top_p=0.9)
    return tokenizer.decode(outputs[0][1:])

demo = gr.Interface(fn=make_summarization, inputs="text", outputs="text")
demo.launch(share=True)