---
license: mit
language:
- en
base_model:
- mlx-community/Llama-3.2-11B-Vision-Instruct-8bit
library_name: mlx
---

# CaseCraft - Fine-Tuned Legal AI Model for IBC

🚀 **Model Overview**
This fine-tuned **LLM** is built specifically for **Indian legal research**, focusing on **Insolvency and Bankruptcy Code (IBC)** cases, amendments, and legal terminology. The model enables **accurate retrieval, summarization, and Q&A** for legal professionals.

🔗 **GitHub Repository**: [CaseCraft](https://github.com/manisunder246/CaseCraft)


## 📜 **Training Data**
- **IBC Act & Amendments**: Fine-tuned on **official laws, regulatory changes, and case-specific rulings.**
- **Legal Glossary**: An exhaustive set of legal terms and definitions for improved domain understanding.


## 📊 **Fine-Tuning Details**
- **Base Model**: LLaMA 3.2 (8-bit quantized)
- **Fine-Tuning Method**: LoRA (Low-Rank Adaptation)
- **Hardware Used**: Apple Silicon (M3 Max, 40-core GPU)
- **Optimizations**: Gradient checkpointing, adaptive chunking for long legal texts


## 🔧 **Usage**
To download the fine-tuned model from Hugging Face, use the following commands:

```bash
pip install huggingface_hub hf_transfer

export HF_HUB_ENABLE_HF_TRANSFER=1
huggingface-cli download --local-dir chanakya-ai-ibc-finetuned manisunder246/chanakya-ai-ibc-finetuned
```

## 📌 **Future Enhancements**
- Expand dataset with **High Court judgments**
- Improve **retrieval mechanisms** with custom legal embeddings
- Deploy as an **interactive legal assistant** for practitioners

📢 **Contributions & feedback welcome!** Feel free to explore the repo and test the model. 🚀