Datasets:

usvsnsp

/

semantic-filters

like 1

The default config contains 32 while it should generally contain 3 splits maximum (train/validation/test). If the splits pile_duped_6.9b, memories_duped_1b, pile_deduped_1.4b, memories_duped_6.9b, pile_duped_2.8b... are not used to differentiate between training and evaluation, please consider defining configs of this dataset instead. You can find how to define configs instead of splits here: https://huggingface.co./docs/hub/datasets-data-files-configuration