InvertingGANs_with_ConsecutiveImgs
InvertingGANs_with_ConsecutiveImgs copied to clipboard
Question about the w_direc tensor shape
HI, I have a question about the codes in main.py file
I am wondering why the shape of the tensor w_direc in main.py line 110 is (1,1,512) while the size of dlatent is (1,18,512). As far as I understood in the paper I think the shape of the two tensors should be same where w_direc corresponds to \vec{n}_{k} and dlatent corresponds to w_b