holi-lab/ArcANE-32B-SFT
LoRA-style vs Qwen/Qwen3-32B. 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 64 of 64 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/1, Norms 0/257, MLP 192/192, Attention 256/256, lm_head 0/1
3 derivatives checked
- mradermacher/ArcANE-32B-SFT-i1-GGUF: Quantization
- mradermacher/ArcANE-32B-SFT-GGUF: Quantization
- holi-lab/ArcANE-32B-DPO: LoRA-style