SicariusSicariiStuff/Oni_Mitsubishi_12B
LoRA-style (fewer layers) vs google/gemma-3-12b-it. SicariusSicariiStuff/Oni_Mitsubishi_12B matches part of google/gemma-3-12b-it (626 of its 1065 tensors; the rest of google/gemma-3-12b-it has no counterpart). On the shared part: 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 48 of 48 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/1, Norms 0/289, MLP 144/144, Attention 192/192
5 derivatives checked
- bartowski/SicariusSicariiStuff_Oni_Mitsubishi_12B-GGUF: Same weights, converted
- mradermacher/Oni_Mitsubishi_12B-i1-GGUF: Quantization
- SicariusSicariiStuff/Oni_Mitsubishi_12B_iMatrix: Quantization
- mradermacher/Oni_Mitsubishi_12B-GGUF: Quantization
- SicariusSicariiStuff/Oni_Mitsubishi_12B_GGUF: Quantization