Naphula/Evilmind-24B-v1

Full fine-tune vs Naphula/BeaverAI_Fallen-Mistral-Small-3.1-24B-v1e_textonly. Fine-tune: 100% of the weights changed, across 40 of 40 layers.

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

2 derivatives checked