milvlg / rosita Goto Github PK

谢谢！

problem solved.

thanks a lot for your great work!

Hi, Your article has a great idea, thank you for sharing it, and thank you for your open source code. It's really great and helpful.

I ran the finetuning code with "scripts/train-vqa-vqav2.sh", for "train+trainvalsplit of VQAv2" and I achieved the accuracy successfully.

But after that when I ran the code with "scripts/test-vqa-vqav2.sh" for "evaluation on the test split" it will run ok but doesn't show any result of accuracy as it was in "train-vqa".

Could you please help me with how can I see the result of accuracy for the test split too? I'm a beginner in python and don't know how to do it.

Thanks.

There is a naming error in the provided vqa2.0 files in datasets.md.
The file names of the minival2014 files v2_mscoco_minival2014_annotations.json and v2_mscoco_minival2014_questions.json should be interchanged.