| --- |
| base_model: google/gemma-2-2b-it |
| library_name: peft |
| pipeline_tag: text-generation |
| tags: |
| - base_model:adapter:google/gemma-2-2b-it |
| - lora |
| - transformers |
| - legal |
| - indian-law |
| - bare-acts |
| - bns |
| - bnss |
| - bsa |
| license: gemma |
| language: |
| - en |
| --- |
| |
| # gemma-2-indian-bare-acts |
|
|
| This is a parameter-efficient fine-tuned (PEFT / LoRA) adapter for **Google's Gemma-2-2b-it**, trained specifically on central and statutory Indian Bare Acts. |
|
|
| ## Model Details |
|
|
| ### Model Description |
|
|
| `dotzcode/gemma-2-indian-bare-acts` adapts Google's Gemma 2 2B Instruct model for accurate statutory interpretation, legal querying, and understanding statutory provisions, definitions, and penalties under Indian jurisprudence. |
|
|
| - **Model Name:** `dotzcode/gemma-2-indian-bare-acts` |
| - **Developed by:** dotzcode |
| - **Model Type:** PEFT LoRA Adapter (Causal Language Model) |
| - **Base Model:** `google/gemma-2-2b-it` |
| - **Language(s) (NLP):** English (Statutory text & legal terminology) |
| - **License:** gemma |
| - **Finetuned from model:** google/gemma-2-2b-it |
|
|
| ### Statutory Frameworks Covered |
| - **Reformed Criminal Codes:** Bharatiya Nyaya Sanhita (BNS), Bharatiya Nagarik Suraksha Sanhita (BNSS), and Bharatiya Sakshya Adhiniyam (BSA). |
| - **Historic Codes & Cross-References:** Indian Penal Code (IPC), Code of Criminal Procedure (CrPC), and Indian Evidence Act (IEA). |
| - **Constitutional Law:** The Constitution of India (Articles, Schedules, and Fundamental Rights). |
| - **Civil & Commercial Acts:** Indian Contract Act, Negotiable Instruments Act, Companies Act, and Code of Civil Procedure (CPC). |
|
|
| ## Uses |
|
|
| ### Direct Use |
| - Querying specific legal sections, definitions, statutory exceptions, and punishable offenses under Indian law. |
| - Cross-referencing transitional mappings between historic penal laws (IPC/CrPC/IEA) and reformed codes (BNS/BNSS/BSA). |
| - Educational and assistive legal research. |
|
|
| ### Downstream Use |
| - Integrating into Retrieval-Augmented Generation (RAG) pipelines for legal question answering and statutory document assistants. |
|
|
| ### Out-of-Scope Use |
| - Automated formal legal counsel or unverified courtroom drafting. |
| - High-stakes judicial decisions without qualified legal review. |
|
|
| ## Bias, Risks, and Limitations |
|
|
| - **Hallucinations:** Language models may occasionally miscite subsections, procedural conditions, or penalty durations. |
| - **Statutory Revisions:** Judicial precedents and statutory amendments evolve over time; answers should always be cross-referenced with official Gazette notifications. |
|
|
| ### Recommendations |
| All outputs should be treated as research assistance and verified against authoritative statutory texts and primary sources. |
|
|
|
|
| #Training Details |
|
|
| ### Training Data |
| Trained on instruction-tuned statutory datasets encompassing Indian Bare Acts, cross-code transition tables (IPC to BNS), and section-by-section legal provisions. |
|
|
| ### Training Procedure |
| - **Fine-Tuning Method:** Low-Rank Adaptation (LoRA / PEFT) |
| - **Precision:** Mixed precision (`bfloat16`) |
| - **Objective:** Causal Language Modeling / Instruction Tuning |
|
|
| ## Legal Disclaimer |
|
|
| > **Disclaimer:** This model is designed solely for educational, academic, and assistive research purposes. It does not constitute certified legal counsel or formal legal advice under the Advocates Act, 1961. Always consult a qualified advocate for actionable legal matters. |
|
|
| ## Framework versions |
|
|
| - PEFT 0.16.0 |
| - Transformers |
| - PyTorch |
|
|
| ## How to Get Started with the Model |
|
|
| ```python |
| import torch |
| from peft import PeftModel |
| from transformers import AutoModelForCausalLM, AutoTokenizer |
| |
| base_model_id = "google/gemma-2-2b-it" |
| adapter_id = "dotzcode/gemma-2-indian-bare-acts" |
| |
| tokenizer = AutoTokenizer.from_pretrained(adapter_id) |
| base_model = AutoModelForCausalLM.from_pretrained( |
| base_model_id, |
| torch_dtype=torch.bfloat16, |
| device_map="auto" |
| ) |
| |
| model = PeftModel.from_pretrained(base_model, adapter_id) |
| |
| messages = [ |
| { |
| "role": "user", |
| "content": "Explain the essential elements and punishment for theft under Section 303 of the Bharatiya Nyaya Sanhita (BNS)." |
| } |
| ] |
| |
| prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True) |
| inputs = tokenizer(prompt, return_tensors="pt").to(model.device) |
| |
| outputs = model.generate( |
| **inputs, |
| max_new_tokens=350, |
| temperature=0.2, |
| top_p=0.9, |
| do_sample=True |
| ) |
| |
| response = tokenizer.decode(outputs[0][inputs.input_ids.shape[-1]:], skip_special_tokens=True) |
| print(response) |
| |
| ---------------------------------------------------------------------------------------------- |
| |
