First of all, I am very sorry to ask a question about ml-upscale under this project, as the issue tab was not found in ml upscale. I am a student, and your article "UPC ALE: Unconstrained Channel Pruning" was very helpful to me, but I didn't fully understand it. I don't quite understand some of the content in the residual connection structure in Figure 4 of the article. Although A and C are both producers, why is the channel of C after pruning the same as D, 2, 4, and the previous convolution of C is B? So the reserved channels should be the same as or less than the channels in B? Or is it true that C also uses memory replication? I really hope to receive your reply. Thank you very much.