Zero-shot learning with allennlp

Hi!
I’m super new to allennlp so these questions might be relatively simple to answer. I’m trying to build a model for zero-shot learning. I’m using the allennlp model

So, I want to know to right way to invoke my unseen classes only during testing. Previously in pytorch, I used a flag. But, i’m not sure how to do that with allennlp.

Thanks!