Closed kiranchari closed 1 year ago
Describe the bug
The code imports tllib.alignment.dann.DomainAdversarialLoss instead of tllib.alignment.adda.DomainAdversarialLoss in adda.py (see https://github.com/thuml/Transfer-Learning-Library/blob/master/examples/domain_adaptation/image_classification/adda.py#L25)
Is this correct?
Yes, and this is because we implemented several modified versions of ADDA but only kept the better performing one. You can refer to issue172 for details.
ok thanks
Describe the bug
The code imports tllib.alignment.dann.DomainAdversarialLoss instead of tllib.alignment.adda.DomainAdversarialLoss in adda.py (see https://github.com/thuml/Transfer-Learning-Library/blob/master/examples/domain_adaptation/image_classification/adda.py#L25)
Is this correct?