how to use word vectors?

it seems the classifier just works on the passed in token (words unless you write your own tokenizer)
how could I best use this with multidimensional tokens such as word vectors?