Is the twentybn data really 773GB?

[ana-GT]

Hi, I was trying to give a shot to this code after reading the Grounding Predicates to Actions paper. I was surprised to see that the data for the 20BN is ~773GB in Google Drive. As I took a look at the 20BN data itself from its original page, I see that it amounts to < 30GB, so I was curious on what additional data could be in the Drive link you guys kindly mention in the package.

My hard drive does not have enough space for all this data 😅 , so I am asking just to check if all this data is really necessary. Thanks!

Ana

[Bailey-24]

yes, could you please release the checkpoint

[ana-GT]

Hi again. Just a comment: I tried several time to download the dataset, but its sheer size makes it impossible (the download eventually fails with a 403 error. I've gotten close to downloading up to ~750 GB only for the process to fail). If it were possible, to release a sub-section of the full data (or just divided in smaller size items). It would very much be appreciated.

Thanks a lot Toki! And thanks for making your work publicly available.