JuliusFx/dyula-french-nmt-8

JuliusFx/dyu-fr-ctranslate2

This is a machine translation model that translates Dyula to French using the ctranslate2 framework.

The modelbased on this google colab notebook that was kindly created by Kartheek Akella.

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation train_df

More information needed

Usage

Load and use for inference

import ctranslate2
import sentencepiece as spm

# Download model
snapshot_download(
    repo_id="JuliusFx/dyula-french-ctranslate2_v2_25epochs",
    local_dir="/path/to/save/locally"
)

# Define model interface
class CTranslate2Model:
    def __init__(self, model_dir: str, sp_model_path: str) -> None:
        self.translator = ctranslate2.Translator(model_dir)
        self.sp = spm.SentencePieceProcessor()
        self.sp.Load(sp_model_path)

    def translate(self, sentence: str) -> str:
        '''
        Translate the given sentence.

        :param sentence: Sentence to be translated
        :return: Translated sentence.
        '''
        prepared_train_df = clean_text(sentence)
        tokens = self.sp.Encode(prepared_train_df, out_type=str)
        translation = self.translator.translate_batch([tokens], beam_size=5)
        translated_sentence = self.sp.Decode(translation[0].hypotheses[0])
        return translated_sentence

# Initialize the model once
model_dir = "/path/to/your/model_dir"
sp_model_path = "/path/to/your/sp_model_path"
translator = CTranslate2Model(model_dir=model_dir, sp_model_path=sp_model_path)

def translate_sentence(sentence: str) -> str:
    return translator.translate(sentence)

# Example usage
dyula_sentence = "i tɔgɔ bi cogodɔ"
translated_sentence = translate_sentence(dyula_sentence)
print("Dyula sentence:", dyula_sentence)
print("Translated French sentence:", translated_sentence)

Training procedure

Training hyperparameters

More information needed

Training results

More information needed

Framework versions