Delta-Vector/Nanuq-R1-9B
LoRA-style vs Delta-Vector/Austral-Xgen-9B-Winton. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of k_proj, o_proj, q_proj, v_proj changed, in 45 of 45 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/1, Norms 0/91, MLP 0/135, Attention 179/180, lm_head 0/1
2 derivatives checked
- mradermacher/Nanuq-R1-9B-i1-GGUF: Quantization
- mradermacher/Nanuq-R1-9B-GGUF: Same weights, converted