Naphula/Goetia-26B-A4B-v1.7
Full fine-tune (fewer layers) vs ConicCat/Gemma4-Writer-26BA4B. Naphula/Goetia-26B-A4B-v1.7 matches part of ConicCat/Gemma4-Writer-26BA4B (662 of its 1013 tensors; the rest of ConicCat/Gemma4-Writer-26BA4B has no counterpart). On the shared part: Fine-tune: 100% of the weights changed, across 30 of 30 layers.
Tensors changed: Embeddings 4/4, MLP 150/150, Norms 84/271, other 30/122, Attention 115/115
4 derivatives checked
- mradermacher/Goetia-26B-A4B-v1.7-i1-GGUF: Quantization
- bartowski/Naphula_Goetia-26B-A4B-v1.7-GGUF: Same weights, converted
- mradermacher/Goetia-26B-A4B-v1.7-GGUF: Quantization
- Koshkasa/Naphula_Goetia-26B-A4B-v1.7-GGUF: Same weights, converted