Open missflash opened 1 year ago
I agree with you. The test dataset should never be used as a validation set during the training process. This practice can lead to data leakage and result in overly optimistic performance evaluations. Any reported test performance metrics would be artificially high and not reflective of the model’s actual ability to generalize to unseen data.
I think you should use
validation_loader
instead oftest_loader
in the followingself.vali
function.https://github.com/thuml/Anomaly-Transformer/blob/e11c317974b729d330b8f84de0619902c3b5cee9/solver.py#L196