tiennv

Problem with contrastive loss in pretrain stage

1

Thanks for your great work. I meet the problem when using the same hyperparameters in [NQ example](https://github.com/texttron/tevatron/tree/main/examples/coCondenser-nq#training-with-hard-negatives) pre-train on the second stage like coCondenser (we call uptrain stage with contrastive...