--data-target ${ParsedLabelJsonPath} # is not stable. do i have to change this in order to run and also is it possible to run your vqa model on colab .
Yes. You need to adjust the --data-target for conducting evaluation. And the evaluation grondtruth we used is in json format (Actually, we parse the original label into jason format for better evaluation).
I think it is possible for running on colab, and we will try to implement this. We will do our best to implement this feature, but the timing may be uncertain. We also welcome third party implementations.
--data-target ${ParsedLabelJsonPath} # is not stable. do i have to change this in order to run and also is it possible to run your vqa model on colab .