app.py

import os
import torch
import gradio as gr
import time
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline


def load_models():
    model = AutoModelForSeq2SeqLM.from_pretrained("facebook/nllb-200-distilled-1.3B")
    tokenizer = AutoTokenizer.from_pretrained("facebook/nllb-200-distilled-1.3B")
    
    return model, tokenizer


def translation(source, target, text):

    translator = pipeline('translation', model=model, tokenizer=tokenizer, src_lang=source, tgt_lang=target)
    output = translator(text, max_length=400)

    end_time = time.time()

    output = output[0]['translation_text']
    return output


if __name__ == '__main__':
    print('\tinit models')
 
    model, tokenizer = load_models()
    global model
    global tokenizer
    
    # define gradio demo
    lang_codes = ["eng_Latn", "fuv_Latn", "fra_Latn", "arb_Arab"]
    #inputs = [gr.inputs.Radio(['nllb-distilled-600M', 'nllb-1.3B', 'nllb-distilled-1.3B'], label='NLLB Model'),
    inputs = [gr.inputs.Dropdown(lang_codes, default='fra_Latn', label='Source'),
              gr.inputs.Dropdown(lang_codes, default='fuv_Latn', label='Target'),
              gr.inputs.Textbox(lines=5, label="Input text"),
              ]

    title = "Fulfulde translator"

    demo_status = "Demo is running on CPU"
    description = "Fulfulde to French, English or Arabic and vice-versa translation demo using NLLB."
    examples = [
    ['fra_Latn', 'fuv_Latn', 'La traduction est une tâche facile.']
    ]

    gr.Interface(translation,
                 inputs,
                 outputs,
                 title=title,
                 description=description,
                 examples=examples,
                 examples_per_page=50,
                 ).launch()