NexesMess/Llama_3.x_70b_Triads_V4

Full fine-tune vs NexesMess/Llama_3.x_70b_TessEssence_0.10c. Fine-tune: 100% of the weights changed, across 80 of 80 layers.

Tensors changed: lm_head 1/1, Embeddings 1/1, Norms 161/161, MLP 240/240, Attention 320/320

2 derivatives checked