| 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) | |