nllb_600M_bi_boen_3

NLLB 600m Tibetan

State of the art

import torch
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline

def translate(text, source_lang, target_lang, model_name="TenzinGayche/nllb_600M_bi_boen_3"):
    # Define flores codes
    flores_codes = {
        "Standard Tibetan": "bod_Tibt",
        "English": "eng_Latn"
    }

    # Convert language names to flores codes
    source = flores_codes[source_lang]
    target = flores_codes[target_lang]

    # Load model and tokenizer
    model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
    tokenizer = AutoTokenizer.from_pretrained("facebook/nllb-200-distilled-600M")

    # Check if a GPU is available and set device accordingly
    device = 0 if torch.cuda.is_available() else -1

    # Create translator pipeline
    translator = pipeline('translation', model=model, tokenizer=tokenizer, 
                          src_lang=source, tgt_lang=target, device=device)

    # Perform translation
    output = translator(text, max_length=400)

    # Extract translated text
    translated_text = output[0]['translation_text']

    return translated_text

# Example usage
if __name__ == "__main__":
    input_text = "Hello, how are you?"
    source_language = "English"
    target_language = "Standard Tibetan"
    
    result = translate(input_text, source_language, target_language)
    print(f"Original: {input_text}")
    print(f"Translated: {result}")