thuml/Transfer-Learning-Library

The performance of the model decreases as the number of training times increases

Closed this issue · 1 comments

hi ,i try to run “a-dapt” on my own datasets. However, the performance decrease when i run “a-dapt” multiple times. Could you give me some advices on what problem may happen?

I think it may be because the pseudo labels are not accurate and lead to error accumulation. Can you provide the evaluation result (such as mAP) of each round?