NexesMess/Llama_3.3_70b_FallenMare

Full fine-tune vs EVA-UNIT-01/EVA-LLaMA-3.33-70B-v0.1. 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