mradermacher/GRMR-V3-G4B-GGUF
Same weights, converted (fewer layers) vs qingy2024/GRMR-V3-G4B. mradermacher/GRMR-V3-G4B-GGUF matches part of qingy2024/GRMR-V3-G4B (444 of its 883 tensors; the rest of qingy2024/GRMR-V3-G4B has no counterpart). On the shared part: Same weights stored in another dtype (BF16 → F16/F32).
Tensors changed: Embeddings 0/1, Norms 0/205, MLP 0/102, Attention 0/136