search

huggingface / text-embeddings-inference

A blazing fast inference solution for text embeddings models
https://huggingface.co/docs/text-embeddings-inference/quick_tour
Apache License 2.0
2.59k stars 161 forks source link

Request support for Llama Prompt Guard #354

Open bluenevus opened 1 month ago

bluenevus commented 1 month ago

Feature request

This is a Bert based model however when trying to run, the message says model not supported. https://huggingface.co/meta-llama/Prompt-Guard-86M/tree/main

Motivation

LLM-powered applications are susceptible to prompt attacks, which are prompts intentionally designed to subvert the developer’s intended behavior of the LLM. Categories of prompt attacks include prompt injection and jailbreaking:

Prompt Injections are inputs that exploit the concatenation of untrusted data from third parties and users into the context window of a model to get a model to execute unintended instructions. Jailbreaks are malicious instructions designed to override the safety and security features built into a model. Prompt Guard is a classifier model trained on a large corpus of attacks, capable of detecting both explicitly malicious prompts as well as data that contains injected inputs.

Your contribution

testing. Ultimately, I'm a system administrator that loads the models in an inference engine for developers.

vrdn-23 commented 1 month ago

This falls under the Deberta V2 class of models which already has issues here (#281 #199).

@OlivierDehaene Is this something that can hopefully be prioritized by the maintainer team?

gabrielchua commented 1 month ago

+1

blazing fast guardrails would be great.

Stealthwriter commented 3 weeks ago

Deberta V3 is widely used it should be supported