zerofata/G4-MeroMero-v2-31B
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
8 derivatives checked
- zerofata/G4-MeroMero-v2-31B-GGUF: Same weights, converted (fewer layers)
- mradermacher/G4-MeroMero-v2-31B-i1-GGUF: Quantization (fewer layers)
- mradermacher/G4-MeroMero-v2-31B-GGUF: Quantization (fewer layers)
- rk35784/G4-MeroMero-v2-31B-GGUF: Same weights, converted (fewer layers)
- DeusImperator/G4-MeroMero-v2-31B-GGUF-long-ctx: Quantization (fewer layers)
- MetaphoricalCode/G4-MeroMero-v2-31B-Q6_K_L-GGUF: Quantization (fewer layers)
- McG-221/G4-MeroMero-v2-31B-mlx-8Bit: Quantization
- kabachuha/G4-MeroMero-v2-31B-Heretic-ARA-LoRA: Adapter
2 other repos hold byte-for-byte identical weight files.