---

language:

• en

license: apache-2.0

base_model: unsloth/llama-3.2-3b-instruct-unsloth-bnb-4bit

tags:

• legal
• indian-law
• BNS
• BNSS
• BSA
• criminal-law
• llama
• llama3
• gguf
• quantized
• ollama
• llama-cpp
• unsloth
• domain-adaptation
• instruction-tuning
• question-answering
• law
• india

datasets:

• GSMS-B/Indian-Legal-QA-BNS-BNSS-BSA

pipeline_tag: text-generation

---

⚖️🦙 Indian Legal Llama 3.2 — 3B · GGUF

<p align="center">

<img src="https://img.shields.io/badge/Base%20Model-Llama%203.2%203B-7C3AED?style=for-the-badge&logo=meta" alt="Base Model"/>

<img src="https://img.shields.io/badge/Type-GGUF%20Quantized-F59E0B?style=for-the-badge" alt="Type"/>

<img src="https://img.shields.io/badge/Domain-Indian%20Criminal%20Law-DC2626?style=for-the-badge" alt="Domain"/>

<img src="https://img.shields.io/badge/Runtime-Ollama%20%7C%20llama.cpp-10B981?style=for-the-badge" alt="Runtime"/>

<img src="https://img.shields.io/badge/Acts-BNS%20%7C%20BNSS%20%7C%20BSA-16A34A?style=for-the-badge" alt="Acts"/>

<img src="https://img.shields.io/badge/License-Apache%202.0-F59E0B?style=for-the-badge" alt="License"/>

</p>

> 🟡 This is the GGUF quantized version — optimised for CPU inference via Ollama or llama.cpp. No GPU required. For full-precision PyTorch usage, see the Merged Model.

---

📖 Model Description

Indian Legal Llama 3.2 — 3B (GGUF) is a quantized, CPU-friendly version of the domain-adapted Llama 3.2 3B model, fine-tuned via QLoRA on unsloth/llama-3.2-3b-instruct-unsloth-bnb-4bit and covering all 1,059 sections of India's three 2023 criminal justice reform acts:

| Act | Full Name | Replaces | Sections |

|---|---|---|---|

| 📕 BNS 2023 | Bharatiya Nyaya Sanhita | IPC 1860 | 358 |

| 📗 BNSS 2023 | Bharatiya Nagarik Suraksha Sanhita | CrPC 1973 | 531 |

| 📘 BSA 2023 | Bharatiya Sakshya Adhiniyam | Indian Evidence Act 1872 | 170 |

Trained on 6,354 instruction-format QA pairs — 6 question types per section covering definitions, scenarios, legal elements, exceptions, and consequences.

---

🔗 Model Family — Llama 3.2 3B

| Variant | Repo | Best For |

|---|---|---|

| 🟢 Merged | GSMS-B/Indian-Legal-Llama-3.2-3B | Out-of-the-box inference, Gradio / API deployment |

| 🔵 LoRA Adapter | GSMS-B/Indian-Legal-Llama-3.2-3B-Adapter | Lightweight loading on top of base model |

| 🟡 GGUF (this repo) | GSMS-B/Indian-Legal-Llama-3.2-3B-GGUF | CPU inference via Ollama / llama.cpp |

---

🚀 Quick Start

🖥️ Run with Ollama (easiest)

ollama run hf.co/GSMS-B/Indian-Legal-Llama-3.2-3B-GGUF

🐍 Run with llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
    repo_id="GSMS-B/Indian-Legal-Llama-3.2-3B-GGUF",
    filename="*Q4_K_M.gguf",   # recommended quant
    n_ctx=2048,
    verbose=False
)

response = llm.create_chat_completion(
    messages=[
        {"role": "system",  "content": "You are an expert legal assistant specializing in Indian criminal law — BNS, BNSS, and BSA 2023."},
        {"role": "user",    "content": "What is a Zero FIR under BNSS 2023?"}
    ],
    max_tokens=300,
    temperature=0.1
)
print(response["choices"][0]["message"]["content"])

🦜 LangChain integration

from langchain_community.llms import LlamaCpp

llm = LlamaCpp(
    model_path="path/to/Indian-Legal-Llama-3.2-3B.Q4_K_M.gguf",
    n_ctx=2048,
    temperature=0.1,
    verbose=False
)
print(llm.invoke("Explain the presumption of innocence under BSA 2023."))

---

🎯 Recommended Use Cases

> ⚠️ Important Note: This model has been domain-adapted on structured QA data and works best as a component in a larger pipeline rather than a standalone answer engine. Direct usage without retrieval context may produce incomplete or imprecise answers on complex legal queries.

✅ Where this model excels

| Use Case | 💡 How to Use |

|---|---|

| 🔍 RAG Pipeline (CPU) | Run a full retrieval-augmented legal assistant entirely on CPU — no GPU needed |

| 🖥️ Local / Offline Deployment | Deploy in air-gapped environments, court offices, or law firms without cloud dependency |

| 🤖 Legal Chatbot Backend | Integrate via Ollama's REST API into any chatbot stack |

| 📚 Legal Education Tool | Build Q&A apps for law students and practitioners on standard hardware |

| 🔎 Section Lookup Assistant | Surface the exact BNS / BNSS / BSA provision relevant to a query on-device |

| 🧪 Research & Benchmarking | Evaluate quantization impact on Indian legal domain understanding |

| 📝 Structured Legal Summarization | Summarize specific sections when the section text is supplied as context |

| 🏛️ Edge / Low-Resource Deployment | Run on laptops, edge servers, or Raspberry Pi-class devices with sufficient RAM |

❌ Not recommended for

• Standalone legal advice without a retrieval component
• High-stakes legal decisions without qualified human review
• Jurisdictions or acts outside BNS / BNSS / BSA 2023

---

🏋️ Training Details

| Property | Value |

|---|---|

| 🤖 Base model | unsloth/llama-3.2-3b-instruct-unsloth-bnb-4bit |

| 🔧 Fine-tuning method | QLoRA |

| 🎛️ LoRA rank | 64 |

| 🎛️ LoRA alpha | 128 |

| 🧩 Target modules | q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj |

| 📊 Training data | 6,354 QA pairs — 1,059 sections × 6 question types |

| 🔁 Epochs | 3 |

| 📦 Batch size (effective) | 4 |

| 📈 Learning rate | 2e-4 |

| ⚙️ Optimizer | adamw_8bit |

| 💻 Hardware | Google Colab T4 GPU |

| 🛠️ Framework | Unsloth + TRL SFTTrainer |

| 💬 Prompt format | ChatML |

---

📊 Training Dataset

| 📂 Dataset | 🔗 Link |

|---|---|

| Indian Legal QA — BNS + BNSS + BSA 2023 | GSMS-B/Indian-Legal-QA-BNS-BNSS-BSA |

6 question types per section:

definitional_topic · definitional_section · scenario · elements · exceptions · consequence

---

👤 Author

GSMS-B — Bugatha Ganasyam Mani Sankar

🤗 Hugging Face Profile

---

⚠️ Disclaimer

This model is intended for research and educational purposes only. It does not constitute legal advice. Outputs should not be relied upon for any legal decision without review by a qualified legal professional. The model's responses reflect patterns in training data and may contain errors or omissions.

---

⚡ Fine-tuned using Unsloth for training efficiency. Quantized to GGUF for broad CPU compatibility.