GitHub - ubergarm/llama.cpp at ug/port-sweep-bench
LLM inference in C/C++. Contribute to ubergarm/llama.cpp development by creating an account on GitHub.
LLM Safety Tool
Summary
Official Meta Llama Cookbook guide for Prompt Guard, a lightweight 86M parameter classifier model that provides input guardrails against prompt injection attacks. Includes inference scripts, a tutorial notebook, and fine-tuning examples runnable on local CPUs.
Highlights
- Prompt Guard is a small 86M parameter classifier model designed to detect prompt injection attacks before they reach the LLM
- Inference and fine-tuning are feasible on local CPUs, making it accessible without GPU resources
- Includes both a Python inference script and an interactive Jupyter notebook tutorial with scenario demos
- Requires gated access to model weights on Hugging Face Hub
- Part of Meta's responsible AI toolkit alongside Llama Guard for output safety
auto-generated
Context
Audience
AI developers and ML engineers building LLM applications who need input safety guardrails against prompt injection attacks
DomainLLM Safety and Security
Formatcode repository with tutorial notebook
Accessopen source code; model weights require gated Hugging Face access
Source Overview
Welcome to the Llama Cookbook! This is your go to guide for Building with Llama: Getting started with Inference, Fine-Tuning, RAG. We also show you how to solve end to end problems using Llama mode…
Topics
Related
Meta Prompt Guard model cardLlama GuardHugging Face HubPrompt injection attacksLLM guardrails
