Datasets:

Languages:
English
ArXiv:
License:

google/wit error: NonMatchingSplitsSizesError

#5
by deanflaviano - opened

Thanks for providing such a great resource.

I'm trying to download the google/wit dataset, but keep running into the NonMatchingSplitsSizesError

I can't get it to download. It gets stuck at around 3M images.

I'm using:
dataset = load_dataset("facebook/pmd", 'wit', features=['image_url', 'text', 'source', 'meta'], token=True, trust_remote_code=True, download_mode="force_redownload", verification_mode="no_checks")

pip version datasets==2.18.0

Let me know if anyone has found a workaround.

Context:

verify_splits(expected_splits, recorded_splits)
100 if len(bad_splits) > 0:
--> 101 raise NonMatchingSplitsSizesError(str(bad_splits))
102 logger.info("All the splits matched successfully.")

NonMatchingSplitsSizesError: [{'expected': SplitInfo(name='train', num_bytes=69619778881, num_examples=37046386, shard_lengths=None, dataset_name='wit'), 'recorded': SplitInfo(name='train', num_bytes=61623600519, num_examples=32795105, shard_lengths=[266000, 267000, 267000, 268000, 266000, 266000, 267000, 266000, 267000, 266000, 266000, 266000, 266000, 266000, 266000, 267000, 267000, 266000, 266000, 267000, 267000, 266000, 266000, 267000, 266000, 265000, 267000, 268000, 266000, 267000, 266000, 266000, 266000, 266000, 266000, 267000, 267000, 266000, 267000, 267000, 267000, 267000, 267000, 266000, 267000, 266000, 267000, 267000, 267000, 267000, 267000, 267000, 267000, 267000, 266000, 267000, 267000, 266000, 267000, 266000, 266000, 266000, 267000, 267000, 266000, 266000, 266000, 267000, 267000, 266000, 267000, 267000, 266000, 267000, 267000, 266000, 266000, 266000, 267000, 267000, 267000, 267000, 267000, 266000, 267000, 267000, 267000, 266000, 267000, 267000, 266000, 266000, 266000, 266000, 267000, 266000, 267000, 266000, 267000, 267000, 267000, 266000, 267000, 266000, 266000, 267000, 267000, 267000, 267000, 266000, 267000, 266000, 267000, 267000, 266000, 266000, 267000, 266000, 267000, 266000, 267000, 267000, 267000, 8105], dataset_name='wit')}]

During handling of the above exception, another exception occurred:

Sign up or log in to comment