app.py

import streamlit as st
import transformers
from transformers import AutoTokenizer, AutoModelWithLMHead
import torch
torch.manual_seed(0)

model_name = "orzhan/rut5-base-detox-v2"
tokenizer = AutoTokenizer.from_pretrained(model_name)
@st.cache
def load_model(model_name):
	    model = AutoModelWithLMHead.from_pretrained(model_name)
	    return model
	
model = load_model(model_name)


def infer(input_ids):
    
    output_sequences = model.generate(
        input_ids=input_ids,
        max_length=40,
        do_sample=False,
        num_return_sequences=1,
        num_beams=8,
        length_penalty=1.0,
        no_repeat_ngram_size=3,
      #  do_sample=True, top_p=0.9, temperature=0.9
    )

    return output_sequences
default_value = "всегда ненавидел этих ублюдочных тварей"
examples = [default_value,
"убила бы этих выродков и их родителей.",
"пошли вы сука все на хуй со своим коронавирусом...",
"Перед барином выслуживаешься холоп?",
"просто , зарплату шакалов отрабатывают , а больше на крысенышей похожи...",
"--- свой текст ---"]

#prompts
st.title("Демо детоксификации на ruT5")
sent = st.selectbox("Пример", examples)
if sent == "--- свой текст ---":
    sent = st.text_area("Исходный текст", default_value)
#if custom_sent == default_value:
#    custom_sent = sent

st.button('Сделать нетоксичным')

encoded_prompt = tokenizer.encode(sent, add_special_tokens=True, return_tensors="pt")
if encoded_prompt.size()[-1] == 0:
    input_ids = None
else:
    input_ids = encoded_prompt


output_sequences = infer(input_ids)



for generated_sequence_idx, generated_sequence in enumerate(output_sequences):
    print(f"=== GENERATED SEQUENCE {generated_sequence_idx + 1} ===")
    generated_sequences = generated_sequence.tolist()

    # Decode text
    text = tokenizer.decode(generated_sequence, clean_up_tokenization_spaces=True, skip_special_tokens=True)

    # Remove all text after the stop token
    #text = text[: text.find(args.stop_token) if args.stop_token else None]

    # Add the prompt at the beginning of the sequence. Remove the excess text that was used for pre-processing
    total_sequence = (
        text
    )

    generated_sequences.append(total_sequence)
    print(total_sequence)

st.write("Преобразованный текст: ")
st.write(generated_sequences[-1])