mlx-community/NousResearch_Hermes-4-14B-BF16-abliterated-mlx
LoRA-style vs NousResearch/Hermes-4-14B. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, q_proj changed, in 1 of 40 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/1, Norms 0/161, MLP 1/120, Attention 1/160, lm_head 0/1