How to use CLEVR datasets?

[JunyongPark97]

Thank you for your amazing research. It has been incredibly helpful for my work.

I have a question regarding the code. When I use a command like:

python -m ocl.cli.train +experiment=slot_attention/clevr10.yaml

Do I need to download the CLEVR dataset manually as a tar file and specify the path, or is it automatically downloaded from S3 or some other source?

Thank you.

[kfan21]

I apologize for my delayed response; I did not receive the notification. I used the CLEVR dataset processed by the object-centric learning framework, which you can find at this link. You should be able to download the CLEVR dataset and process the data as a tar file by specifying the appropriate path. The datasets script is a helpful reference for data processing.