## Model Overview

### Description:

Llama-Guard is a 7B parameter Llama 2-based input-output safeguard model.This model guards against safety risks in human-AI conversations.
Llama-Guard classifies both user prompts and LLM responses for potential harm.

Key Features:
- LLM-based: Leverages the power of a large language model (LLM) for robust analysis.
- Safety Risk Taxonomy: Employs a structured system for categorizing different types of safety concerns.
- Prompt and Response Classification: Identifies potentially harmful content in both user inputs and AI outputs.
- High-Quality Dataset: Trained on carefully curated data for accurate risk assessment.
- Benchmark Performance: Matches or exceeds existing content moderation tools on established benchmarks.
- Multi-class Classification: Provides detailed categorization of risks, not just binary safe/unsafe flags.
- Instruction Fine-tuning: Can be customized for specific use cases and output formats.
- Adjustable Taxonomy: Adaptable to different risk categories and priorities.
- Zero-shot/Few-shot Prompting: Works with diverse taxonomies without extensive re-training.
- Openly Available: Model weights accessible for further research and development.

### Terms of use

By using this software or model, you are agreeing to the terms and conditions of the license, acceptable use policy and Meta’s privacy policy.

### References(s):

- [Llama Guard: LLM-based Input-Output Safeguard for Human-AI Conversations | Research - AI at Meta](https://ai.meta.com/research/publications/llama-guard-llm-based-input-output-safeguard-for-human-ai-conversations/)
- [meta-llama/LlamaGuard-7b · Hugging Face](https://huggingface.co/meta-llama/LlamaGuard-7b)

### Model Architecture:

**Architecture Type:** Transformer 
**Network Architecture:** Llama 2 
**Model Version:** N/A 

### Input:

**Input Format:** Text 
**Input Parameters:** None 
**Other Properties Related to Output:** None 

### Output:

**Output Format:** Text 
**Output Parameters:** None 
**Other Properties Related to Output:** None 

### Software Integration:

**Runtime(s):** N/A 
**Supported Hardware Platform(s):** Hopper, Ampere, Ada 
**Supported Operating System(s):** Linux 

## Inference:

**Engine:** [Triton](https://developer.nvidia.com/triton-inference-server) 
**Test Hardware:** Other

llama-guard

Llama Guard is a model for classifying the safety of LLM prompts and responses, using a taxonomy of safety risks.