Use main branch auto-round to quant GLM-4.5 and add layer_config with lm head name: <img width="560" height="971" alt="Image" src="https://github.com/user-attachments/assets/6f54fa35-7f10-47e7-b66e-4310ecd8ad35" /> Found the lm head weight is not quanted: <img width="744" height="256" alt="Image" src="https://github.com/user-attachments/assets/89767223-fefb-4d11-a0b3-5d30c9005d48" />