Qwen3-4B Sinhala Marriage Law Fine-Tuned Model

Developed by: Final Year Project Group, Faculty of Engineering, University of Ruhuna /n License: Apache-2.0 Based on: unsloth/Qwen3-4B-unsloth-bnb-4bit

This is a fine-tuned Qwen3 model trained specifically on Sri Lankan law text, including marriage and related legislation. The model supports text generation tasks and has been optimized for efficiency with Unsloth’s fast training utilities.

Highlights:

Fine-tuned on legal domain text for improved accuracy in Sinhala law queries.

Trained with Unsloth for 2× faster training and memory-efficient 4-bit quantization.

Compatible with GGUF format for local inference or deployment in frameworks like Ollama.

Intended Use:

Generating legal guidance and summaries in Sinhala based on the trained domain corpus.

Educational and research purposes in law-focused NLP applications.