gopi30/english-to-tamil-stage2

🚀 English-to-Tamil Translation — Fine-Tuned M2M100 (Stage 2)

Model ID: gopi30/english-to-tamil-stage2
Model Type: M2M100ForConditionalGeneration
Language Pair: English ➡ Tamil
Framework: 🤗 Transformers

---

📌 Model Summary

This model is the Stage 2 fine-tuned version of Facebook's M2M100 (418M) Multilingual Machine Translation model for English-to-Tamil translation. It builds on the improvements made in Stage 1 and incorporates more targeted domain-specific training to enhance translation fluency, contextual accuracy, and Tamil grammar structure.

---

✅ Improvements Over Stage 1

• Trained with cleaner and more diverse parallel English-Tamil sentence pairs
• Better handling of idioms and complex sentence structures
• Enhanced translation consistency and Tamil morphology
• Optimized tokenizer usage for more accurate sentence segmentation

---

📈 Use Cases

• Translating English educational content to Tamil
• Localizing web and mobile apps for Tamil-speaking audiences
• Assisting in communication for native Tamil speakers
• Voice assistants and accessibility tools

---

🧩 Base Model

• Base: facebook/m2m100_418M
• Languages Fine-Tuned: en ➡ ta

---

📦 Installation

Make sure you have the transformers library installed:

pip install transformers

Via Transformers Library

Code

from transformers import M2M100ForConditionalGeneration, M2M100Tokenizer

model_path = "gopi30/english-to-tamil-stage2"
model = M2M100ForConditionalGeneration.from_pretrained(model_path)
tokenizer = M2M100Tokenizer.from_pretrained(model_path)

def translate_en_to_ta(text):
    tokenizer.src_lang = "en"
    encoded = tokenizer(text, return_tensors="pt")
    generated = model.generate(**encoded, forced_bos_token_id=tokenizer.get_lang_id("ta"))
    return tokenizer.decode(generated[0], skip_special_tokens=True)

# Example
print(translate_en_to_ta("Hello!"))