VijayChandra/english-to-telugu-translator-nllb

English-to-Telugu Translator (NLLB Fine-Tuned)

This is a sequence-to-sequence translation model fine-tuned from facebook/nllb-200-distilled-600M to translate from English to Telugu.

This model was developed as part of the SOAI 2025 Internship Program. The primary goal was to create a useful translation tool that also serves as a "Corpus Collection Engine" to gather human-verified translations, thereby helping to improve future AI models for Indian languages.

Model Details

• Base Model: facebook/nllb-200-distilled-600M
• Source Language: English (eng_Latn)
• Target Language: Telugu (tel_Telu)
• Fine-tuning Data: A custom dataset of English-Telugu parallel sentences.

How to Use

You can use this model directly with the transformers library pipeline for easy translation.

from transformers import pipeline

# Use the model ID for your fine-tuned model
model_id = "VijayChandra/english-to-telugu-translator-nllb"

# Create a translation pipeline
translator = pipeline(
    "translation",
    model=model_id,
    src_lang="eng_Latn",
    tgt_lang="tel_Telu"
)

# Translate a sentence
english_text = "Hello, how are you?"
telugu_translation = translator(english_text)

print(f"English: {english_text}")
print(f"Telugu: {telugu_translation[0]['translation_text']}")
# Expected Output: Telugu: హలో, మీరు ఎలా ఉన్నారు?

Training Procedure

The model was fine-tuned using the Seq2SeqTrainer from the Hugging Face transformers library. The training process involved preparing a parallel corpus of English and Telugu sentences and training for 3 epochs with a learning rate of 3e-5.

Framework versions

• Transformers 4.40.0
• Pytorch 2.6.0
• Datasets 2.19.0
• Tokenizers 0.19.1