katanemo/Arch-Guard

Partial fine-tune vs meta-llama/Prompt-Guard-86M. Partial change: 31% of the weights changed (attention, embeddings, norms, other), in 12 of 12 layers.

Tensors changed: other 52/52, Embeddings 3/4, Norms 47/50, Attention 96/96

On the Hub · Downloads on Model Pulse