Log probabilities for discriminative models
Lynx1820 opened this issue · comments
Hi, it says on the paper that you compute the log probability for BERT/RoBERTa, but here it seems like you're just taking the probabilities. Is this intentional?
@Lynx1820 please raise this issue in the new repo (if the problem still exists). We made several changes https://github.com/McGill-NLP/bias-bench
We are no longer maintaining this repo. See you in the other repo!
Closing this since it was resolved here.