import streamlit as st

from transformers import T5ForConditionalGeneration, T5Tokenizer

model = T5ForConditionalGeneration.from_pretrained("Unbabel/gec-t5_small")
tokenizer = T5Tokenizer.from_pretrained('t5-small')

sentence = st.text_area("enter some text")

tokenized_sentence = tokenizer('gec: ' + sentence, max_length=128, truncation=True, padding='max_length', return_tensors='pt')
corrected_sentence = tokenizer.decode(
    model.generate(
        input_ids = tokenized_sentence.input_ids,
        attention_mask = tokenized_sentence.attention_mask, 
        max_length=128,
        num_beams=5,
        early_stopping=True,
    )[0],
    skip_special_tokens=True, 
    clean_up_tokenization_spaces=True
)
print(corrected_sentence)