{0}: number of channels in the first argument,
{1}: number of channels per group in the wights,
{2}: channel dimension,
{3}: number of groups,
{4}: group dimension



Number of channels per group ({1}, specified by the size of weights dimension number {2}) times the number of groups ({3}, specified by the size of weights dimension number {4}) must be equal to the size of the 'C' dimension of the input data