Hi, I was wondering if it was possible to get the output logits ( the model output : model(input)) as an output along with the multi-class label dict in zero-shot classification pipeline?
I want to use the logit probabilities to make some other inferences.