It is pretty common for us to have the data split into multiple folds.
(This issue is not about the semantics or splitting data, especially because PSL data is often graph-based and hard to split directly.)
It would be great if we could have some system where the CLI understands that the data is in folds and it should run validation on each fold and then average across all folds.
This will probably need to be solved in conjunction with #166.
It is pretty common for us to have the data split into multiple folds. (This issue is not about the semantics or splitting data, especially because PSL data is often graph-based and hard to split directly.) It would be great if we could have some system where the CLI understands that the data is in folds and it should run validation on each fold and then average across all folds.
This will probably need to be solved in conjunction with #166.