Kaoeiri/Qwenwify-32B-v2.6

Full fine-tune vs AXCXEPT/EZO-Qwen2.5-32B-Instruct. Fine-tune: 100% of the weights changed, across 64 of 64 layers.

Tensors changed: Norms 126/129, MLP 192/192, Attention 448/448

1 derivatives checked