djuna/G2-Noranum-27B
Full fine-tune vs anthracite-org/magnum-v3-27b-kto. Fine-tune: 100% of the weights changed, across 46 of 46 layers.
Tensors changed: Embeddings 1/1, Norms 183/185, MLP 138/138, Attention 184/184
1 derivatives checked
- mradermacher/G2-Noranum-27B-GGUF: Quantization