Question about the compensate_beta in pruning depthwise separable layer?

[imyangs]

Dear author, thank u for your excellent work and code. Recently when I read the convert code, I cannot completely understand the logic behand the beta compensate, can you explain it or point out the reference?
BTW, why we only need to compensate beta for pointwise layer rather than both the depthwise and pointwise layers?

https://github.com/DingXiaoH/ResRep/blob/aac6ed3c9347bed2628f120920c132cbea16065c/rr/resrep_convert.py#L201-L203