Inference result of Qwen2 quantized by QoQ is wrong,how to fix?

[limertang]

Hi,

1. I quantize Qwen2/Qwen2.5 with the command as below:
   python -m deepcompressor.app.llm.ptq
   configs/qoq-g128.yaml
   --model-name qwen-2-7b --model-path /PATH/TO/QWEN-2-7B
   --smooth-proj-alpha 0 --smooth-proj-beta 1
   --smooth-attn-alpha 0.5 --smooth-attn-beta 0
2. I deploy the quantized qwen2 with tensorrt-llm, but the inference output is wrong.
   But I quantize Qwen1.5 with the same command and deploy it with Tensorrt-llm, it works well, inference resut is right.
   So, is the smooth value not correct for Qwen2? How to set smooth value for different model?