Hi, Thanks for sharing your code. Could I ask the reason that why you change the batch size to 1 and make the batch size as the number of groups in your ModulatedConv2d function? Repeating the weights will increase the memory request, can you let me know the advantage of doing that?
Hi, Thanks for sharing your code. Could I ask the reason that why you change the batch size to 1 and make the batch size as the number of groups in your ModulatedConv2d function? Repeating the weights will increase the memory request, can you let me know the advantage of doing that?