The generalization of sparse compressed layouts to N-dimensionaltensors can lead to some confusion regarding the count of specifiedelements. When a sparse compressed tensor contains batch dimensionsthe number of specified elements will correspond to the number of suchelements per-batch. When a sparse compressed tensor has dense dimensionsthe element considered is now the K-dimensional array. Also for blocksparse compressed layouts the 2-D block is considered as the elementbeing specified. Take as an example a 3-dimensional block sparsetensor, with one batch dimension of length b, and a blockshape of p, q. If this tensor has n specified elements, thenin fact we have n blocks specified per batch. This tensor wouldhave values with shape (b, n, p, q). This interpretation of thenumber of specified elements comes from all sparse compressed layoutsbeing derived from the compression of a 2-dimensional matrix. Batchdimensions are treated as stacking of sparse matrices, dense dimensionschange the meaning of the element from a simple scalar value to anarray with its own dimensions.