Jackrong/Gemopus-4-31B-it
LoRA-style vs google/gemma-4-31B-it. 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 60 of 60 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/4, Norms 0/523, other 0/62, MLP 180/261, Attention 230/338
4 derivatives checked
- mradermacher/Gemopus-4-31B-it-i1-GGUF: Quantization (fewer layers)
- mradermacher/Gemopus-4-31B-it-GGUF: Quantization (fewer layers)
- Cyclone-Labs/Lucid-Shadow-31B: Full fine-tune
- maldv/Gemma-4-31B-Isometry-Fabled-Persona: LoRA-style
1 other repos hold byte-for-byte identical weight files.