AliceThirty/G4-Alice-v1.0-31B
Quantization (fewer layers) vs ReadyArt/Serenity-31B-GGUF. ReadyArt/Serenity-31B-GGUF matches part of AliceThirty/G4-Alice-v1.0-31B (772 of its 1188 tensors; the rest of AliceThirty/G4-Alice-v1.0-31B has no counterpart). On the shared part: ReadyArt/Serenity-31B-GGUF is a GGUF quantization (Q8_0) of the same weights: 361 tensors match exactly and 411 within quantization error.
Tensors changed: Norms 0/361, Embeddings 0/1, Attention 0/230, MLP 0/180
3 derivatives checked
- mradermacher/G4-Alice-v1.0-31B-i1-GGUF: Quantization (fewer layers)
- mradermacher/G4-Alice-v1.0-31B-GGUF: Quantization (fewer layers)
- AliceThirty/G4-Alice-v1.0-31B-gguf: Same weights, repacked (fewer layers)