Pretraining on Val?

Hi,

I noticed that the validation set is used during Sonata pretrain, even though labels are not accessed. I’m concerned this might introduce some data leakage and influence generalization assessment. It might be better to avoid using the validation set at this stage.
https://github.com/Pointcept/Pointcept/blob/5b10bc05b0a3cab8cd5df5ebd142b56bb791ad15/configs/sonata/pretrain-sonata-v1m1-0-base.py#L197

Start at configs/sonata/pretrain-sonata-v1m1-0-base.py around line 197 and trace how the validation set enters Sonata pretraining. Verify whether validation samples are included without labels, then confirm that the pretraining data no longer uses that split and that the configuration still runs as expected.

Written by the indexing model from the issue text.