Nexesenex/Llama_3.x_70b_SmarTricks_v1.70
Full fine-tune vs NexesMess/Llama_3.3_70b_Negative_Wayfarer_fusion. Fine-tune: 100% of the weights changed, across 80 of 80 layers.
Tensors changed: Norms 160/161, MLP 240/240, Attention 320/320
1 derivatives checked
- mradermacher/Llama_3.x_70b_SmarTricks_v1.70-GGUF: Quantization