Sorawiz/MistralSmall-Creative-24B-Realist

Full fine-tune vs Darkknight535/WinterEngine-24B-Instruct. Fine-tune: 100% of the weights changed, across 40 of 40 layers.

Tensors changed: lm_head 1/1, Embeddings 1/1, Norms 74/81, MLP 120/120, Attention 160/160

2 derivatives checked