app.py

import torch
from transformers import MBart50TokenizerFast, MBartForConditionalGeneration
import gradio as gr


class MBartTranslator:
    """ 
    MBartTranslator provides a simple user interface for text translation using the MBart Langugae model.

    This class can translate one language to 50 other languages and is based on the model "facebook/mbart-large-50-one-to-many-mmt", a pretrained model from facebook.

    Attributes:
    model (MBartForConditionalGeneration): The MBart language model.
    tokenizer (MBart50TokenizerFast): The MBart tokenizer.
    """

    def __init__(self, model_name="facebook/mbart-large-50-many-to-many-mmt", src_lang=None, tgt_lang=None):

        self.supported_languages = [
            "en_XX",
            "hi_IN",
            "ml_IN",
            "ta_IN",
            "te_IN"
        ]

        print("Building translator")
        print("Loading tokenizer")
        self.tokenizer = MBart50TokenizerFast.from_pretrained(model_name)
        print("Laoding generator")
        self.model = MBartForConditionalGeneration.from_pretrained(model_name)


    def translate(self, text: str, input_language: str, output_language: str) -> str: 
        """
        Translates the given text from input to output langauge.

        Args:
            text (str): The text to translate.
            input_language (str): The input language code (e.g. "en_XX" for English).
            output_language (str): The output language code (e.g. "ta_IN" for Tamil).
        Returns:
            str: The translated text.
        """

        if input_language not in self.supported_languages:
            raise ValueError(f"Input language not supported. Supported languages: {self.supported_languages}")
        if output_language not in self.supported_languages:
            raise ValueError(f"Output language not supported. Supported languages: {self.supported_languages}")
        
        self.tokenizer.src_lang = input_language
        encoded_input = self.tokenizer(text, return_tensors='pt')
        generated_tokens = self.model.generate(
            **encoded_input, forced_bos_token_id=self.tokenizer.lang_code_to_id[output_language]
        )
        translated_text = self.tokenizer.batch_decode(generated_tokens, skip_special_tokens=True)

        return translated_text[0]
    

translator = MBartTranslator()

def translate_text(source_lang, target_lang, text):
    return translator.translate(text, source_lang, target_lang)

    
translation_interface = gr.Interface(
    fn=translate_text,
    inputs=[
    gr.Dropdown(choices=["en_XX", "hi_IN", "ml_IN", "ta_IN", "te_IN"], label = "Source Language"),
    gr.Dropdown(choices=['en_XX', "hi_IN", "ml_IN", "ta_IN", "te_IN"], label = "Target Language"),
    gr.Textbox(label = "Text to translate")
    ],
    outputs=gr.Textbox(label = "Translated text")
    )
                                         
translation_interface.launch()