ApolloRaines/Llama-3.1-405B-Instruct-Jbliterated
LoRA-style vs meta-llama/Llama-3.1-405B-Instruct. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, k_proj, o_proj, q_proj, up_proj, v_proj changed, in 120 of 126 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/1, Attention 480/504, MLP 360/378, Norms 0/253, lm_head 0/1