beita6969/EvoSteer-9b
LoRA-style vs Qwen/Qwen3.5-9B. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of in_proj_qkv, k_proj, o_proj, out_proj, q_proj, v_proj changed, in 32 of 32 layers; norms, biases and embeddings are untouched.
Tensors changed: lm_head 0/1, Embeddings 0/5, MLP 0/211, Attention 80/336, other 0/1, Norms 0/221
1 derivatives checked
- mradermacher/EvoSteer-9b-GGUF: Same weights, converted