app.py

import os
import torch
import gradio as gr
import time
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline
from flores200_codes import flores_codes


def load_models():
    # build model and tokenizer
    model_name_dict = {#'nllb-distilled-600M': 'facebook/nllb-200-distilled-600M',
                  'nllb-1.3B': 'facebook/nllb-200-1.3B',
                  #'nllb-distilled-1.3B': 'facebook/nllb-200-distilled-1.3B',
                  #'nllb-3.3B': 'facebook/nllb-200-3.3B',
                  }

    model_dict = {}

    for call_name, real_name in model_name_dict.items():
        print('\tLoading model: %s' % call_name)
        model = AutoModelForSeq2SeqLM.from_pretrained(real_name)
        tokenizer = AutoTokenizer.from_pretrained(real_name)
        model_dict[call_name+'_model'] = model
        model_dict[call_name+'_tokenizer'] = tokenizer

    return model_dict


def load_models():
    # build model and tokenizer
    model_name_dict = {'nllb-distilled-600M': 'facebook/nllb-200-distilled-600M',
                  #'nllb-1.3B': 'facebook/nllb-200-1.3B',
                  #'nllb-distilled-1.3B': 'facebook/nllb-200-distilled-1.3B',
                  #'nllb-3.3B': 'facebook/nllb-200-3.3B',
                  }

    model_dict = {}

    for call_name, real_name in model_name_dict.items():
        print('\tLoading model: %s' % call_name)
        model = AutoModelForSeq2SeqLM.from_pretrained(real_name)
        tokenizer = AutoTokenizer.from_pretrained(real_name)
        model_dict[call_name+'_model'] = model
        model_dict[call_name+'_tokenizer'] = tokenizer

    return model_dict


def translation(source, target, text):
    if len(model_dict) == 2:
        model_name = 'nllb-distilled-600M'

    start_time = time.time()
    source = flores_codes[source]
    target = flores_codes[target]

    model = model_dict[model_name + '_model']
    tokenizer = model_dict[model_name + '_tokenizer']

    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']
    result = {'inference_time': end_time - start_time,
              'source': source,
              'target': target,
              'result': output}
    return result





global model_dict

model_dict = load_models()
    
    # define gradio demo
lang_codes = list(flores_codes.keys())
    #inputs = [gr.inputs.Radio(['nllb-distilled-600M', 'nllb-1.3B', 'nllb-distilled-1.3B'], label='NLLB Model'),


title = "NLLB distilled 1.3B demo"

demo_status = "Demo is running on CPU"
description = f"Details: https://github.com/facebookresearch/fairseq/tree/nllb. {demo_status}"
examples = [
['English', 'Urdu', 'Hi. nice to meet you']
]

demo = gr.Interface(translation,
            inputs = [gr.inputs.Dropdown(lang_codes, default='English', label='Source'),
                      gr.inputs.Dropdown(lang_codes, default='Korean', label='Target'),
                      gr.inputs.Textbox(lines=5, label="Input text"),
                      ],

            outputs = gr.outputs.JSON(),
             title=title,
             description=description,
             )
demo.launch()