Datasets:

Modalities:
Tabular
Text
Formats:
parquet
Languages:
code
ArXiv:
Libraries:
Datasets
Dask
License:

Getting NonMatchingSplitsSizesError when loading Python subset

#4
by cassanof - opened
BigCode org

Hello! I am trying to load the Python subset of the dataset. It raises a NonMatchingSplitsSizesError 15% into loading the dataset.

>>> import datasets
>>> ds = datasets.load_dataset("bigcode/the-stack-v2", data_dir="data/Python", split="train")
Generating train split:  15%|_____________                                                                          | 96448523/664910862 [02:09<12:41, 746839.37 examples/s]
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/scratch/federicoc/.env/lib/python3.10/site-packages/datasets/load.py", line 2582, in load_dataset
    builder_instance.download_and_prepare(
  File "/scratch/federicoc/.env/lib/python3.10/site-packages/datasets/builder.py", line 1005, in download_and_prepare
    self._download_and_prepare(
  File "/scratch/federicoc/.env/lib/python3.10/site-packages/datasets/builder.py", line 1118, in _download_and_prepare
    verify_splits(self.info.splits, split_dict)
  File "/scratch/federicoc/.env/lib/python3.10/site-packages/datasets/utils/info_utils.py", line 101, in verify_splits
    raise NonMatchingSplitsSizesError(str(bad_splits))
datasets.utils.info_utils.NonMatchingSplitsSizesError: [{'expected': SplitInfo(name='train', num_bytes=303603153658, num_examples=664910862, shard_lengths=None, dataset_name=None), 'recorded': SplitInfo(name='train', num_bytes=42044267633, num_examples=96448523, shard_lengths=[1148000, 1148000, 1147000, 1148000, 1147000, 1148000, 1147000, 1147048, 1148000, 1147000, 1148000, 1148000, 1147000, 1147000, 1148000, 1147048, 1148000, 1148000, 1147000, 1148000, 1147000, 1147000, 1148048, 1148000, 1147000, 1148000, 1148000, 1147000, 1147000, 1148000, 1147048, 1147000, 1147000, 1147000, 1148000, 1147000, 1147000, 1148000, 1148048, 1148000, 1147000, 1148000, 1147000, 1147000, 1148000, 1147048, 1147000, 1147000, 1147000, 1148000, 1147000, 1148000, 1148000, 1147047, 1148000, 1148000, 1147000, 1147000, 1148000, 1148000, 1147000, 1147047, 1148000, 1147000, 1148000, 1148000, 1147000, 1147000, 1147047, 1147000, 1147000, 1148000, 1148000, 1148000, 1147000, 1147000, 1148047, 1148000, 1147000, 1148000, 1147000, 1147000, 1147000, 1147000, 62047], dataset_name='the-stack-v2')}]

Is this a known issue? I'm on the latest datasets version: datasets==2.18.0

BigCode org

Hi @cassanof ! This should be fixed now, could you please retry with load_dataset(..., download_mode="force_redownload")?

BigCode org

works! thanks!

cassanof changed discussion status to closed

Sign up or log in to comment