marian-id-en-lg

MarianMT Indonesian-English Translation (Fine-tuned)

This model is a fine-tuned version of Helsinki-NLP/opus-mt-id-en for Indonesian to English translation.

Model Details

• Base Model: Helsinki-NLP/opus-mt-id-en
• Fine-tuned on: TED Talks parallel corpus (Indonesian-English)
• Training Date: 2025-05-25
• Languages: Indonesian (id) → English (en)
• License: Apache 2.0

Training Configuration

• Training Framework: PyTorch + Transformers
• Training Data: TED Talks parallel corpus
• Dataset Usage: 100% of full dataset
• Training Parameters:
  • Learning Rate: 3e-5
  • Batch Size: 4/2 (GPU/CPU)
  • Max Length: 128 tokens
  • Epochs: 10

Usage

from transformers import MarianMTModel, MarianTokenizer

# Load model and tokenizer
tokenizer = MarianTokenizer.from_pretrained("dhintech/marian-id-en-lg")
model = MarianMTModel.from_pretrained("dhintech/marian-id-en-lg")

# Translate Indonesian to English
def translate(text):
    inputs = tokenizer(text, return_tensors="pt", padding=True)
    outputs = model.generate(**inputs, max_length=128, num_beams=4)
    return tokenizer.decode(outputs[0], skip_special_tokens=True)

# Example usage
indonesian_text = "Selamat pagi, terima kasih sudah datang."
english_translation = translate(indonesian_text)
print(english_translation)

Example Translations

Performance

• Optimized for conversational and presentation-style text
• Best performance on formal Indonesian text
• Model size: approximately 300MB
• Suitable for mobile deployment

Citation

@misc{marian-id-en-lg,
  title={MarianMT Indonesian-English Translation (Fine-tuned)},
  author={DhinTech},
  year={2025},
  publisher={Hugging Face},
  journal={Hugging Face Model Hub},
  howpublished={\url{https://huggingface.co/dhintech/marian-id-en-lg}}
}