ParentFull threaddragandj·convo is smart enough to cover 1D, 2D, 3D, and any other convolution layer that the backend can support. You only need to specify the data that it can't figure out, but you can specify more if you want.View on HN