DeCo
DeCo copied to clipboard
Retraining
@yaolinli Q-Former needs to be retrained after changing to 2D average pooling, right?