OddTheGreat/Luna_12B_V.3
LoRA-style vs LastRef/gemma-3-12b-it-heretic-x. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, fc2, gate_proj, k_proj, o_proj, q_proj, up_proj, v_proj changed, in 48 of 48 layers; norms, biases and embeddings are untouched.
Tensors changed: Norms 0/400, MLP 145/252, Attention 192/408, other 0/1, Embeddings 0/3
2 derivatives checked
- mradermacher/Luna_12B_V.3-i1-GGUF: Quantization (fewer layers)
- mradermacher/Luna_12B_V.3-GGUF: Quantization (fewer layers)