DeepBeepMeep/krea-2
Full fine-tune vs krea/Krea-2-Turbo. Fine-tune: 100% of the weights changed, across 28 of 28 layers.
Tensors changed: Norms 129/129, other 36/36, MLP 105/105, Attention 160/160
Full fine-tune vs krea/Krea-2-Turbo. Fine-tune: 100% of the weights changed, across 28 of 28 layers.
Tensors changed: Norms 129/129, other 36/36, MLP 105/105, Attention 160/160