netcat420/MFANN-abliterated-phi2-merge-unretrained

LoRA-style vs M1LL1X/microsoft-phi-2-trained-abliterated-QA-1960-all-modules-merged. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of k_proj, q_proj, v_proj changed, in 31 of 32 layers; norms, biases and embeddings are untouched.

Tensors changed: Embeddings 0/1, Norms 0/66, MLP 0/128, Attention 93/256, lm_head 0/2

1 derivatives checked