Question about the matrix format of vgg19

[ysj1173886760]

I've read the doc, it seems like that the format of the weight of conv layer in vgg19 pretrained model is [W, H, Cin, Cout]
I've had a discussion about this topic before issue.
I want to ask about the format of weight of conv layer in vgg19 pretrained model. Is that [W, H, Cin, Cout] or [H, W, Cin, Cout]
Btw, the pretrained model is this one MD5(imagenet-vgg-verydeep-19.mat) = 8ee3263992981a1d26e73b3ca028a123