How to ensure the number of test set samples?

[terry-r123]

Thanks for your wonderful work!

I would like to use the original training and test sets, keeping the same sequences as in the original paper. I ran the file https://nbviewer.org/github/johli/aparent/blob/master/analysis/evaluate_aparent_random_mpra_legacy.ipynb and found that every library had the same test number. It seems different from the split in Figure 1. C in the paper.

Thanks for your time!
Best regards!

Original paper:

evaluate_aparent_random_mpra_legacy.ipynb: