KAN accepts lower input dimension than dataset features

[mmghorbani]

In the classification example using make_moons (shape: (1000, 2)), the model works even with KAN(width=[1, 2]), despite the input dimension being less than the number of features. I expected an error unless the input width matched the feature count (e.g., width=[2, 2]). Interestingly, increasing the input dimension to 3 does raise an error as expected.
Why does the model accept a lower input dimension without complaint? Is this intended?