app.py

from transformers import NllbTokenizer, AutoModelForSeq2SeqLM
import gradio as gr

model = AutoModelForSeq2SeqLM.from_pretrained('alimboff/nllb-200-kbd')#.cpu()
tokenizer = NllbTokenizer.from_pretrained('alimboff/nllb-200-kbd')

def fix_tokenizer(tokenizer, new_lang='kbd_Cyrl'):
    """
    Add a new language token to the tokenizer vocabulary
    (this should be done each time after its initialization)
    """
    old_len = len(tokenizer) - int(new_lang in tokenizer.added_tokens_encoder)
    tokenizer.lang_code_to_id[new_lang] = old_len-1
    tokenizer.id_to_lang_code[old_len-1] = new_lang
    # always move "mask" to the last position
    tokenizer.fairseq_tokens_to_ids["<mask>"] = len(tokenizer.sp_model) + len(tokenizer.lang_code_to_id) + tokenizer.fairseq_offset

    tokenizer.fairseq_tokens_to_ids.update(tokenizer.lang_code_to_id)
    tokenizer.fairseq_ids_to_tokens = {v: k for k, v in tokenizer.fairseq_tokens_to_ids.items()}
    if new_lang not in tokenizer._additional_special_tokens:
        tokenizer._additional_special_tokens.append(new_lang)
    # clear the added token encoder; otherwise a new token may end up there by mistake
    tokenizer.added_tokens_encoder = {}
    tokenizer.added_tokens_decoder = {}

fix_tokenizer(tokenizer)

language_codes = {
    "Кабардино-Черкесский": "kbd_Cyrl",
    "Русский": "rus_Cyrl"
}

def translate(
    text, input_language, output_language,
    a=32, b=3, max_input_length=1024, num_beams=8, **kwargs
):
    src_lang = language_codes[input_language]
    tgt_lang = language_codes[output_language]
    """Turn a text or a list of texts into a list of translations"""
    tokenizer.src_lang = src_lang
    tokenizer.tgt_lang = tgt_lang
    inputs = tokenizer(
        text, return_tensors='pt', padding=True, truncation=True,
        max_length=max_input_length
    )
    model.eval() # turn off training mode
    result = model.generate(
        **inputs.to(model.device),
        forced_bos_token_id=tokenizer.convert_tokens_to_ids(tgt_lang),
        max_new_tokens=int(a + b * inputs.input_ids.shape[1]),
        num_beams=num_beams, **kwargs
    )
    return tokenizer.batch_decode(result, skip_special_tokens=True)[0] #без [0]

with gr.Blocks() as demo:
    gr.Markdown("### Переводчик через ИИ")
    
    with gr.Row():
        input_language = gr.Radio(choices=["Кабардино-Черкесский", "Русский"], label="Выберите язык исходного текста", value="Кабардино-Черкесский")
        output_language = gr.Radio(choices=["Кабардино-Черкесский", "Русский"], label="Выберите язык для перевода", value="Русский")
    
    with gr.Row():
        text_input = gr.Textbox(label="Введите текст для перевода")
        text_output = gr.Textbox(label="Перевод", interactive=False)
    
    with gr.Row():
        translate_button = gr.Button("Перевести")
    
    translate_button.click(
        fn=translate,
        inputs=[text_input, input_language, output_language],
        outputs=text_output
    )

demo.launch()

# # Example usage:
# # Ӏ
# t = 'пэшым лӀы зыбжанэ щӀэсщ'

# kbdru = translate(t, 'kbd_Cyrl', 'rus_Cyrl')
# rukbd = translate(kbdru, 'rus_Cyrl', 'kbd_Cyrl')

# print(kbdru)
# print(rukbd)