The ordinal encoder is expecting a tensor with at least two elements. Once you fulfill that requirement, it should work. Can you please open up an issue on Scholar? We should check for the shape and “hardcode” the answer if a tensor of size 1 is given. We should probably check the shapes in general.
