anthracite-org/magnum-v2.5-12b-kto
LoRA-style vs anthracite-org/magnum-v2-12b. 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 40 of 40 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/1, Norms 0/81, MLP 120/120, Attention 160/160, lm_head 0/1
12 derivatives checked
- bartowski/magnum-12b-v2.5-kto-GGUF: Quantization
- anthracite-org/magnum-v2.5-12b-kto-gguf: Quantization
- mradermacher/magnum-12b-v2.5-kto-i1-GGUF: Quantization
- mradermacher/magnum-v2.5-12b-kto-i1-GGUF: Quantization
- InferenceIllusionist/magnum-12b-v2.5-kto-iMat-GGUF: Same weights, converted
- featherless-ai-quants/anthracite-org-magnum-v2.5-12b-kto-GGUF: Quantization
- mradermacher/magnum-v2.5-12b-kto-GGUF: Quantization
- mradermacher/magnum-12b-v2.5-kto-GGUF: Quantization
- grimjim/magnum-consolidatum-v1-12b: Full fine-tune
- ockerman0/MN-12B-Starcannon-v4-unofficial: Full fine-tune
- SlerpE/NightyGurps-12b-v1-experimental: LoRA-style
- ockerman0/MN-12B-Starcannon-v5-unofficial: Full fine-tune
1 other repos hold byte-for-byte identical weight files.