ask for help #4
Description
In a OOD dataset, we cannot get the label of data. and we want to predict the well-trained model's performance on this OOD dataset. but in your experiment, the lables in OOD dataset is obtained. And your code is also do in this way. So, how to predict a unlabeled ood performance with ProjNorm, such as the evaluation of accuracy in a classification problem.
