import ctranslate2
import sentencepiece as spm

# Load the model
translator = ctranslate2.Translator("path/to/model")

# Load tokenizers
sp_source = spm.SentencePieceProcessor(model_file="source.spm")
sp_target = smp.SentencePieceProcessor(model_file="target.spm")

# Translate
source_tokens = sp_source.encode("Your text here", out_type=str)
results = translator.translate_batch([source_tokens])
translation = sp_target.decode(results[0].hypotheses[0])

Performance

This INT8 quantized version provides:

~75% reduction in model size
Faster inference speed
Maintained translation quality
Mobile-friendly deployment

Original Model

Based on the OPUS-MT project: https://github.com/Helsinki-NLP/Opus-MT

manancode/opus-mt-gmq-gmq-ctranslate2-android

opus-mt-gmq-gmq-ctranslate2-android

Model Details

Usage

With CTranslate2

Performance

Original Model

manancode/
opus-mt-gmq-gmq-ctranslate2-android