grimjim/Gemma2-Nephilim-v3-9B
Full fine-tune vs UCLA-AGI/Gemma-2-9B-It-SPPO-Iter3. Fine-tune: 98% of the weights changed, across 42 of 42 layers.
Tensors changed: Embeddings 1/1, Norms 105/169, MLP 123/126, Attention 164/168
2 derivatives checked
- mradermacher/Gemma2-Nephilim-v3-9B-i1-GGUF: Quantization
- mradermacher/Gemma2-Nephilim-v3-9B-GGUF: Same weights, converted