tencent/AuK-Flash

Full fine-tune vs tencent/AuK. Fine-tune: 100% of the weights changed, across 0 of 0 layers.

Tensors changed: other 86/86, Embeddings 10/11, Norms 163/163, Attention 160/160