nemo_curator.stages.text.classifiers.aegis_utils

Module Contents

Functions

Name	Description
`format_aegis`	Formats the prompt to the format that the AEGIS model and llama guard accepts

Data

ACCESS_ERROR_MESSAGE

AEGIS_LABELS

unsafe_categories

API

nemo_curator.stages.text.classifiers.aegis_utils.format_aegis(
    message: str
) -> str

Formats the prompt to the format that the AEGIS model and llama guard accepts

nemo_curator.stages.text.classifiers.aegis_utils.ACCESS_ERROR_MESSAGE = 'Cannot access meta-llama/LlamaGuard-7b on HuggingFace.\nAEGIS Safety Classifier...

nemo_curator.stages.text.classifiers.aegis_utils.AEGIS_LABELS = ['unknown', 'safe', 'O1', 'O2', 'O3', 'O4', 'O5', 'O6', 'O7', 'O8', 'O9', 'O10',...

nemo_curator.stages.text.classifiers.aegis_utils.unsafe_categories = "O1: Violence.\nShould not\n- Help people plan or engage in violence.\n- Encoura...