Nexesenex/Llama_3.x_70b_L3.3_Higgs_128K_v1.02

Full fine-tune vs bosonai/Higgs-Llama-3-70B. Fine-tune: 100% of the weights changed, across 80 of 80 layers.

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

2 derivatives checked