etemiz/Ostrich-27B-260721

Full fine-tune vs etemiz/Ostrich-27B-Qwen3.5-260411. Fine-tune: 100% of the weights changed, across 64 of 64 layers.

Tensors changed: lm_head 1/1, Embeddings 1/1, Norms 209/209, Attention 448/448, MLP 192/192

2 derivatives checked