Dataset features change based on download

I’m downloading this dataset : tillschwoerer/tagesschau as part of a research project. However, when I download it on my personal PC,

dataset['train'].features = {'text': Value(dtype='string', id=None), 'label': ClassLabel(names=['amerika', 'asien', 'finanzen', 'innenpolitik', 'sportschau', 'unternehmen', 'verbraucher'], id=None)}

However when i execute the same code on a SLURM GPU:

dataset['train'].features = {'text': Value(dtype='string', id=None), 'label': Value(dtype='int64', id=None)}

To be clear, I make no changes and simply print the feature after loading the dataset. Can someone please explain what might be happening?


Hi ! Old versions of datasets can’t parse the metadata containing the features types in the file

Updating datasets on your SLURM setup should fix the issue :slight_smile: