ReadyArt/Dark-Scarlett-v2.0-31B
LoRA-style vs nbeerbower/Gemma4-Gutenberg-31B. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_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 120/261, Attention 230/338
7 derivatives checked
- mradermacher/Dark-Scarlett-v2.0-31B-i1-GGUF: Quantization (fewer layers)
- ReadyArt/Dark-Scarlett-v2.0-31B-GGUF: Quantization (fewer layers)
- mradermacher/Dark-Scarlett-v2.0-31B-GGUF: Quantization (fewer layers)
- Ateron/Gemma-4-Dark-Thoughts-V2-31B: LoRA-style
- Ateron/Gemma-4-Dark-Thoughts-31B: LoRA-style
- Vortex5/Scarlet-Shadow-31B: Full fine-tune
- hoborific/Dark-Scarlett-v2.0-31B-W8A16-FP8: Quantization