Bloodviper/deepseek-llama3-ablit-slerp

Full fine-tune vs failspy/Meta-Llama-3-70B-Instruct-abliterated-v3.5. Fine-tune: 100% of the weights changed, across 79 of 79 layers.

Tensors changed: Embeddings 1/1, Norms 159/159, MLP 237/237, Attention 316/316, lm_head 1/1

2 derivatives checked