Kaoeiri/Qwenwify2.5-32B-v4.71-ED

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

Tensors changed: Norms 128/129, MLP 192/192, Attention 448/448

1 derivatives checked