[jborichevskiy]

This is pretty cool to see. Glad to see the weights published. Currently it seems the best we have is OpenAI's `omni-moderation` model series or private models by companies like ActiveFence etc.