Can't merge MultiDataSets containing 1-dimensional inputs?

I have a List on which I’m trying to call MultiDataSet.merge. Each MDS in the list contains one example with 2 feature arrays and 3 label arrays; the feature arrays are respectively 3-dimensional (9x9x19) and 1-dimensional (11 “global” variables); the label arrays are 3-d, scalar, and scalar.

I get an exception at org.nd4j.linalg.dataset.api.DataSetUtil.mergeFeatures(, which expects arrays of 2 to 4 dimensions only.

Why not 1-dimensional arrays? Am I missing a concept?

(So far I’ve just been preparing my data, and I haven’t started specifying my ComputationGraph yet.)

It turns out that the first dimension is across examples, so if each example has a 1-dimensional input, the total number of dimensions should be 2.