Clarification of model fold data splits

[adamyhe]

For model f0, sequences labeled fold0 form the test set and fold1 form validation.
For model f1, sequences labeled fold1 form the test set and fold2 form validation. Etc

Originally posted by @davek44 in #1 (comment)

Hi, I just wanted to clarify the exact splits that were used for each of the model folds. My reading is that the test/val/train splits for each of the models is set up as:

f0: test=fold0, val=fold1, train=rest
f1: test=fold1, val=fold2, train=rest
f2: test=fold2, val=fold3, train=rest
f3: test=fold3, val=fold4, train=rest

Thanks!

[davek44]

Yes, this is correct. Here's the code segment that performs that https://github.com/calico/basenji/blob/master/bin/basenji_train_folds.py#L397

[adamyhe]

Awesome. Thanks!