Closed Florent-Michel closed 2 months ago
The input size of domain_classifier is n_classes * n_features, so I propose to add n_classes as argument of CDAN to create domain_classifier with the right input size when it's not provided.
The input size of domain_classifier is n_classes * n_features, so I propose to add n_classes as argument of CDAN to create domain_classifier with the right input size when it's not provided.