Why doesn't the optimization goal include Quantized X

Hi there, I am a rookie in model quantization. I've read your paper and the results are impressing! However, I couldn't help but wonder, in your optimization problem, you use the below formula:

Why not use Q(XA^(-1))Q(AW) as part of the optimization goal. Wouldn't it help if the quantization error on X is also taken into account?

I hope my silly question won't bother you 😊./

Best regards.

Read the paper's optimization formulation and locate the corresponding objective in the Python implementation; no file, test, or entry point is named in the issue. Establish whether the proposed quantization term is intended, then document the rationale or define an agreed implementation and validation target.

Written by the indexing model from the issue text.