I’m trying to download just the test and validation splits of the following dataset: yuvalkirstain/pickapic_v2 at main.
I can tell that the files are named according to the bottommost scheme here: File names and splits. But, when I use load_dataset
with split="validation"
, it still downloads the entire dataset. Why?
Is load_dataset
not smart enough to only download files with which have validation
in the name?