Fine-tuned LLaMA 3.1 8B Model for Spoken Tamil Translation
🌟 Overview
This project fine-tunes the unsloth/Meta-Llama-3.1-8B-bnb-4bit model using LoRA adapters to translate English sentences into Tamil colloquial (spoken) language.
🔥 Features
Uses LoRA-based fine-tuning to efficiently train a large model on limited resources.
Trained on the dataset thaslimthoufica/English_tamil_dataset.
Supports inference with sampling (temperature, top-p, and repetition penalty) for natural responses.
📚 Dataset
The dataset contains parallel English-to-Tamil colloquial translations.
You can find it here: Hugging Face Dataset
