OccultAI/Ouroboros-24B-v1.4

Partial fine-tune vs Naphula/Goetia-24B-v1.4. Partial change: 7% of the weights changed (attention, mlp, norms), in 29 of 40 layers.

Tensors changed: lm_head 0/1, Embeddings 0/1, Norms 19/81, MLP 9/120, Attention 16/160

2 derivatives checked