anjayasu/animagine-xl-3.0

Full fine-tune vs Linaqruf/animagine-xl-2.0. Fine-tune: 94% of the weights changed, across 11 of 12 layers.

Tensors changed: Embeddings 569/591, other 534/662, Attention 700/716, Norms 442/546

6 other repos hold byte-for-byte identical weight files.