SubMaroon-exp/Kanimus-26B-A4B-FFT-heretic
LoRA-style vs Vortex5/Gemma-4-26B-A4B-Animus-V14.1-FFT-heretic. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of k_proj, q_proj changed, in 30 of 30 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/4, MLP 0/231, Norms 0/433, other 0/122, Attention 60/223
2 derivatives checked
- mradermacher/Kanimus-26B-A4B-FFT-heretic-i1-GGUF: Quantization
- mradermacher/Kanimus-26B-A4B-FFT-heretic-GGUF: Quantization