sartajbhuvaji/Gemma-4-31B-IT-NVFP4-FP8attn-text
LoRA-style vs nvidia/Gemma-4-31B-IT-NVFP4. 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 60 of 60 layers; norms, biases and embeddings are untouched.
Tensors changed: MLP 0/720, Embeddings 0/1, Norms 0/361, other 0/60, Attention 230/230