I'd like to benchmark monoT5 against a custom dataset which has 25 possible labels/targets.
The model generates a seq, I'm still not sure how you can transform that to logits representing specific targets/labels?
Is there a notebook for custom datasets ? thanks
I'd like to benchmark monoT5 against a custom dataset which has 25 possible labels/targets. The model generates a seq, I'm still not sure how you can transform that to logits representing specific targets/labels? Is there a notebook for custom datasets ? thanks