Datasets:

pixparse

/

cc3m-wds

like 24

Follow

Pixel Parsing 35

Works fine for me, note I'm not going to try to download it using datasets so set streaming flag, if you want to dl I'd use the cli.

import datasets
dataset = datasets.load_dataset("pixparse/cc3m-wds", streaming=True)
dataset.keys()
Out[5]: dict_keys(['train', 'validation'])
next(iter(dataset['train']))
Out[6]: 
{'__key__': '000000000',
 '__url__': 'hf://datasets/pixparse/cc3m-wds@46f3d69f840e59d77d52e8decfe5baec97e94c7f/cc3m-train-0000.tar',
 'jpg': <PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=852x480>,
 'txt': "a river has burst it 's banks and has spread out onto arable farmland alongside"}