PurpleLlama
PurpleLlama copied to clipboard
meta-llama
Set of tools to assess and improve LLM security.
I was following this work. It would be greatly appreciated if you could release the evaluation code to help us reproduce your results!
The huggingface [meta-llama/LlamaGuard-7b](https://huggingface.co/meta-llama/LlamaGuard-7b) model seems to be super fast at inference ~0.08-0.10 seconds single sample on an A100 80GB GPU for approx ~300 input tokens and max token generation length...
Appreciate your nice work. Is there any plan to release the dataset or just test set for comparison? Thanks!
When downloading LICENSE and Acceptable Usage Policy, it echos "./download.sh: 13: Bad substitution"
I have tried adding some custom taxonomy to unsafe_categories in llama_guard example [(safety_utils.py)](https://github.com/facebookresearch/llama-recipes/blob/main/src/llama_recipes/inference/safety_utils.py) like this: ```python unsafe_categories = """07: Politics. Should not - Discuss or engage in conversations about political...
Do you plan to release the eval scripts for the results in the paper and here (https://github.com/facebookresearch/PurpleLlama/blob/main/Llama-Guard/MODEL_CARD.md) for openai moderation and toxicchat?
I am using wsl in windows machine, I can use ```wget``` to download the ```download.sh``` file, but when I use ```bash download.sh``` to run the file, I got below error:...
Can you please share some details about fine-tuning LlamaGuard for additional categories? Specifically, every time we need to add one additional category to the existing model, do we need to...
"""The guidelines that Llama Guard is trained on may not be the same as the desired guidelines for the target domain. For such cases, we can leverage the zero-shot or...
Metadata
Owner
Metadata
Set of tools to assess and improve LLM security.