Overview
yamraj047/nepal-legal-mistral-7b is a 7 billion parameter Mistral-7B-v0.1 model, fine-tuned using LoRA on 6,400 Nepal legal instruction-following examples. Its primary focus is on the National Penal Code 2017, enabling it to explain legal provisions, answer questions about Nepal law, and summarize legal concepts in simple language. The model's robustness and compatibility are highlighted by its rapid community adoption and re-quantization into various GGUF formats for CPU-only inference.
Key Capabilities
- Legal Q&A: Answers questions about Nepal legal provisions.
- Concept Explanation: Explains legal concepts in simple language.
- Code Summarization: Provides summaries of legal texts, particularly the National Penal Code 2017.
- Educational Tool: Suitable for legal education and research assistance.
Training Details
The model was fine-tuned on a dataset of 6,400 examples derived from the National Penal Code 2017 and other Nepal legal provisions. Training was conducted using LoRA, with only 41.9 million (1.11%) of the total parameters being trainable. The training process took approximately 3 hours on a Kaggle T4 GPU.
Limitations and Disclaimers
It is crucial to note that this model is not a substitute for professional legal advice and may not reflect recent legal changes. Its training data primarily covers the National Penal Code 2017, with limited coverage of other legal areas. The model is primarily English-based and may not perform well with Nepali language queries. Users should always verify important legal information with qualified professionals.