reproducing for 5-datasets

[jcy132]

I applied the implemented code for 5-datasets.
The result was acc@1 75.5033 | forgetting 14.9538,
which is much lower than the result in the paper, acc@1 88.08 | forgetting 2.21

any missed point?

[JH-LEE-KR]

Hi,
thanks for your comment.

I didn't implement and experiment,
because the dualprompt 5-datasets config of the official code was not opened.
Did you use L2P's 5-datasets config as it is?

In my experience, the dataset order (fixed order in official code) and use_prompt_mask have a significant impact on performance in 5-datasets.

Please check it and leave a comment.
Discuss it together.

Best,
Jaeho Lee.