Datasets:

Modalities:
Text
Formats:
parquet
Languages:
Italian
Libraries:
Datasets
pandas
License:

Load doesn't seem to be loading validation and labels

#3
by skorkmaz88 - opened
BigLAM: BigScience Libraries, Archives and Museums org

I don't know why exactly, as I see labels are identified in the data and I am a beginner in HF, however, when I load the dataset with default values I get this:

image.png

BigLAM: BigScience Libraries, Archives and Museums org

If you look at the loader, there doesn't seem to be any splits other than Train

BigLAM: BigScience Libraries, Archives and Museums org

This dataset only includes a training split since the source dataset doesn't define a specific validation dataset (the paper uses k-fold validation).

You can use datasets to generates a validation set using the train_test_split method. If you want more control over the split you can also use the select method and pass in indexes for the splits you desire.

Let me know if I misunderstood the question/issue.

BigLAM: BigScience Libraries, Archives and Museums org

I was able to do that with train_test_split, my worry was about label designation but I realized it may be in vein, as loader automatically handled conversion to numeric labels

skorkmaz88 changed discussion status to closed

Sign up or log in to comment