Почетна Преглед Помоћ
Пријавите се
radu
/
LLamaRecipes
огледало од https://github.com/facebookresearch/llama-recipes.git
Прати 1
Волим 0
Креирај огранак 0
Датотеке Дискусије 0 Вики
Дрво: 6a5c0f89de
Гране Ознаке
LLamaRecipes
/
recipes
/
responsible_ai
Beto 6a5c0f89de Updating readme and notebook пре 11 месеци
..
llama_guard 6a5c0f89de Updating readme and notebook пре 11 месеци
prompt_guard d28400efb5 Merge branch 'main' of https://github.com/meta-llama/llama-recipes-alpha into responsible_ai пре 1 година
Purple_Llama_Anyscale.ipynb f53f17138b fix dead links after refactor пре 1 година
Purple_Llama_OctoAI.ipynb f53f17138b fix dead links after refactor пре 1 година
README.md bc56973186 update rai readme пре 1 година
code_shield_usage_demo.ipynb 77156c7324 Minor file renames пре 1 година
input_output_guardrails_with_llama.ipynb a404c9249c Notebook to demonstrate using llama and llama-guard together using OctoAI пре 1 година

README.md

Trust and Safety with Llama

The Purple Llama project provides tools and models to improve LLM security. This folder contains examples to get started with PurpleLlama tools.

Tool/Model Description Get Started
Llama Guard Provide guardrailing on inputs and outputs Inference, Finetuning
Prompt Guard Model to safeguards against jailbreak attempts and embedded prompt injections Notebook
Code Shield Tool to safeguard against insecure code generated by the LLM Notebook

Running on hosted APIs

The notebooks input_output_guardrails.ipynb, Purple_Llama_Anyscale & Purple_Llama_OctoAI contain examples for running Meta Llama Guard on cloud hosted endpoints.