SubMaroon/Boulesis-26B-A4B
LoRA-style vs Gryphe/Gemma-4-26B-A4B-StyleTune-V2. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of k_proj, o_proj, q_proj, v_proj changed, in 30 of 30 layers; norms, biases and embeddings are untouched.
Tensors changed: lm_head 0/1, Embeddings 0/4, MLP 0/231, Norms 0/433, other 0/122, Attention 110/223
6 derivatives checked
- mradermacher/Boulesis-26B-A4B-i1-GGUF: Quantization
- spiritfather/Boulesis-26B-A4B-i1-GGUF: Quantization
- mradermacher/Boulesis-26B-A4B-GGUF: Quantization
- spiritfather/Boulesis-26B-A4B-GGUF: Quantization
- Wiself/Boulesis-26B-A4B-Voice: Same weights, repacked (fewer layers)
- McG-221/Boulesis-26B-A4B-9bit: Quantization