NexesMess/Llama_3.x_70b_Triads_V7

Full fine-tune vs NexesMess/Llama_3.1_70b_Luminoblivion_CW_V1.01. Fine-tune: 100% of the weights changed, across 80 of 80 layers.

Tensors changed: Norms 161/161, MLP 240/240, Attention 320/320

1 derivatives checked