CrucibleLab/Gemma-4-31B-Loki-Scotoma-V2.0-Swa-4096
LoRA-style vs ReadyArt/gemma-4-31B-it-scotoma-2. 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 50 of 60 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/4, Norms 0/523, other 0/62, MLP 150/261, Attention 200/338
4 derivatives checked
- mradermacher/Gemma-4-31B-Loki-Scotoma-V2.0-Swa-4096-i1-GGUF: Quantization (fewer layers)
- CrucibleLab/Gemma-4-31B-Loki-Scotoma-V2.0-Swa-4096-GGUF: Quantization (fewer layers)
- mradermacher/Gemma-4-31B-Loki-Scotoma-V2.0-Swa-4096-GGUF: Quantization (fewer layers)
- McG-221/Gemma-4-31B-Loki-Scotoma-V2.0-Swa-4096-mlx-8Bit: Quantization