Issues

• Batched map / Dataset.batch with num_proc>1 splits batches at shard boundaries and drop_last_batch loses rows

  huggingface/datasets#8919 · Filter by author udsy19 opened on Oct 31, 2026

• Json.cast_storage only inspects the first 5 values, so plain strings past that point are stored unencoded and fail to decode

  huggingface/datasets#8914 · Filter by author Steve-20-ai opened on Oct 31, 2026
  1

  

• Empty typed Dataset default save cannot reload because it writes zero shards

  huggingface/datasets#8912 · Filter by author alberttwong opened on Oct 31, 2026

• Docs: missing import re in "Use with Polars" data extraction example

  huggingface/datasets#8909 · Filter by author docsalive-team opened on Oct 30, 2026
  1

• xnumpy_load, xpandas_read_excel, xpyarrow_parquet_read_table and xsio_loadmat never close their xopen handle

  huggingface/datasets#8907 · Filter by author michaelkillgta opened on Oct 29, 2026

• JSON loading retries indefinitely when a mixed-type field name contains a slash

  huggingface/datasets#8898 · Filter by author bbbbbbbbbbb114 opened on Oct 27, 2026

• Fixed-shape ArrayXD pandas round trips lose row nulls and integer precision

  huggingface/datasets#8888 · Filter by author sjdoane opened on Oct 23, 2026
  2

• Dataset.from_sql caches different SQLAlchemy queries under the same key when bound values change

  huggingface/datasets#8877 · Filter by author gorkadabo1 opened on Oct 22, 2026
  1

  

• IterableDataset.decode(False) has no effect after shard() on a streamed media dataset

  huggingface/datasets#8863 · Filter by author RaphaelFakhri opened on Oct 20, 2026

• cast_column / cast are ignored on streaming datasets decoded in threads (audio sampling_rate, decode=False)

  huggingface/datasets#8849 · Filter by author sebxt opened on Oct 18, 2026

• to_sql with num_proc > 1 fails with 'table already exists' when a later batch reaches the database first

  huggingface/datasets#8838 · Filter by author mertkayacs opened on Oct 17, 2026

• Indexing a Dataset with a list or array of bools silently returns rows 0 and 1

  huggingface/datasets#8836 · Filter by author mertkayacs opened on Oct 17, 2026
  1

  

• to_csv ignores the encoding argument and always writes UTF-8

  huggingface/datasets#8832 · Filter by author mertkayacs opened on Oct 17, 2026

• to_csv and to_sql with index=True restart the index at 0 in every batch

  huggingface/datasets#8831 · Filter by author mertkayacs opened on Oct 17, 2026

• ClassLabel.cast_storage fails on a pandas Categorical column (dictionary-encoded Arrow storage)

  huggingface/datasets#8830 · Filter by author mertkayacs opened on Oct 17, 2026
  1

• Fatal PyGILState_Release at interpreter exit when streaming Parquet with pyarrow 25 (arrow#45214 workaround is gated to pyarrow <= 24)

  huggingface/datasets#8793 · Filter by author javi22020 opened on Oct 13, 2026
  1

  

• re-pushing all dataset partitions corrupts it

  huggingface/datasets#8789 · Filter by author belambert opened on Oct 12, 2026
  1

• Dataset.add_item fails when an existing struct-backed feature meets an inferred dictionary

  huggingface/datasets#8766 · Filter by author LiamLin1130 opened on Oct 7, 2026
  1

  

• Dataset.shard raises IndexError for the empty shards when num_shards is larger than the dataset

  huggingface/datasets#8763 · Filter by author Mudassiruddin7 opened on Oct 7, 2026
  1

  

• Audio array remains NumPy after Torch formatting

  huggingface/datasets#8755 · Filter by author abellion opened on Oct 6, 2026
  2

• Parquet export does not close its writer after a batch or metadata failure

  huggingface/datasets#8743 · Filter by author nawaaaaaAaar opened on Oct 5, 2026
  3

  

• Parquet file with an empty row group fails to load: ArrowInvalid: BatchSize must be greater than 0

  huggingface/datasets#8740 · Filter by author Sriram-PR opened on Oct 5, 2026
  1

  

• concatenate_datasets(axis=1) crashes on arrow-backed IterableDatasets with different lengths

  huggingface/datasets#8719 · Filter by author Zhuoxi2000 opened on Oct 1, 2026
  1

  

• Casting mixed JSON and plain strings to Json depends on row order and batch size

  huggingface/datasets#8714 · Filter by author Marcuswang0824 opened on Sep 30, 2026
  1

• JSON loader turns strings like "123" or "null" into other types in inferred Json columns

  huggingface/datasets#8713 · Filter by author u7k4rs6 opened on Sep 30, 2026
  1