Validation set not being used in fine-tuning / linear probing

[Daniel-Lin-S]

I’ve been reviewing the downstream module in your repository and noticed that in some of the fine-tuning and linear probing scripts, all three sets—training, validation, and test—are extracted. However, it appears that only two of them (typically training and validation or training and test) are used during training.

In particular, I observed that in some cases, the test set is used as a validation set—for instance, in line 224 of downstream/linear_probe_EEGPT_BCIC2A.py. I was wondering if this was an intentional design choice or a placeholder setup for future development. Any clarification you could provide would be greatly appreciated.