2.4.0

Dataset Features

• Add concatenate_datasets for iterable datasets by @lhoestq in https://github.com/huggingface/datasets/pull/4500
• Support parallelism with PyTorch DataLoader with parquet/json/csv/text/image/etc. files by @mariosasko in https://github.com/huggingface/datasets/pull/4625
• Support using PCM audio files (#4323) by @YooSungHyun in https://github.com/huggingface/datasets/pull/4409
• [data_files] Files disambiguation: match split names in data files if they are between separators by @lhoestq in https://github.com/huggingface/datasets/pull/4633
• Support extract 7-zip compressed data files by @albertvillanova in https://github.com/huggingface/datasets/pull/4672
• Support extract lz4 compressed data files by @albertvillanova in https://github.com/huggingface/datasets/pull/4700
• Support metadata.jsonl from parent directories in imagefolder @mariosasko in https://github.com/huggingface/datasets/pull/4576

Dataset changes

• Update: allocine - Support streaming by @albertvillanova in https://github.com/huggingface/datasets/pull/4563
• Update: multi_news - Host data on the Hub instead of Google Drive by @albertvillanova in https://github.com/huggingface/datasets/pull/4585
• Update: pn_summary - Host data on the Hub instead of Google Drive by @albertvillanova in https://github.com/huggingface/datasets/pull/4586
• Update: financial_phrasebank - Host data on the Hub by @albertvillanova in https://github.com/huggingface/datasets/pull/4598
• Update: cfq - Support streaming by @albertvillanova in https://github.com/huggingface/datasets/pull/4579
• Update: head_qa - Host data on the Hub and fix NonMatchingChecksumError by @albertvillanova in https://github.com/huggingface/datasets/pull/4588
• Update: bookcorpus - Support streaming dataset by @albertvillanova in https://github.com/huggingface/datasets/pull/4564
• Update: fever - Refactor and add metadata by @albertvillanova in https://github.com/huggingface/datasets/pull/4503
• Update: mlsum - Support streaming dataset by @albertvillanova in https://github.com/huggingface/datasets/pull/4574
• Fix: cats_vs_dogs - Update download url and improve card by @mariosasko in https://github.com/huggingface/datasets/pull/4523
• Fix: conll2003 - fix empty example by @lhoestq in https://github.com/huggingface/datasets/pull/4662
• Fix: WMT datasets - fix loading issue when choosing specific subsets and docs update by @khushmeeet in https://github.com/huggingface/datasets/pull/4554
• Fix: xtreme - fix empty examples in dataset for bucc18 config by @lhoestq in https://github.com/huggingface/datasets/pull/4706
• Fix: crd3 - fix splits that were containing the same data by @lhoestq in https://github.com/huggingface/datasets/pull/4705

Dataset Cards

• Add action names in schema_guided_dstc8 dataset card by @lhoestq in https://github.com/huggingface/datasets/pull/4559
• Add evaluation data to acronym_identification by @lewtun in https://github.com/huggingface/datasets/pull/4561
• Update WinoBias README by @sashavor in https://github.com/huggingface/datasets/pull/4631
• Support "tags" yaml tag by @lhoestq in https://github.com/huggingface/datasets/pull/4716
• Fix POS tags by @lhoestq in https://github.com/huggingface/datasets/pull/4715
• AESLC dataset: Add summarization tags by @hobson in https://github.com/huggingface/datasets/pull/4517

Documentation

• Update docs around audio and vision by @stevhliu in https://github.com/huggingface/datasets/pull/4440
• Update Google Cloud Storage documentation and add Azure Blob Storage example by @alvarobartt in https://github.com/huggingface/datasets/pull/4513
• Remove multiple config section by @stevhliu in https://github.com/huggingface/datasets/pull/4600
• Create new sections for audio and vision in guides by @stevhliu in https://github.com/huggingface/datasets/pull/4519
• Document installation of sox OS dependency for audio by @albertvillanova in https://github.com/huggingface/datasets/pull/4713

General improvements and bug fixes

• Add regression test for ArrowWriter.write_batch when batch is empty by @alvarobartt in https://github.com/huggingface/datasets/pull/4510
• Support all negative values in ClassLabel by @lhoestq in https://github.com/huggingface/datasets/pull/4511
• Add uppercased versions of image file extensions for automatic module inference by @mariosasko in https://github.com/huggingface/datasets/pull/4515
• Patch tests for hfh v0.8.0 by @LysandreJik in https://github.com/huggingface/datasets/pull/4518
• Replace deprecated logging.warn with logging.warning by @hugovk in https://github.com/huggingface/datasets/pull/4539
• [CI] Fix upstream hub test url by @lhoestq in https://github.com/huggingface/datasets/pull/4543
• Fix timestamp conversion from Pandas to Python datetime in streaming mode by @lhoestq in https://github.com/huggingface/datasets/pull/4541
• [CI] fixing seqeval install in ci by pinning setuptools-scm by @lhoestq in https://github.com/huggingface/datasets/pull/4546
• Tell users to upload on the hub directly by @lhoestq in https://github.com/huggingface/datasets/pull/4552
• Add batch_size parameter when calling add_faiss_index and add_faiss_index_from_external_arrays by @alvarobartt in https://github.com/huggingface/datasets/pull/4535
• Make DuplicateKeysError more user friendly [For Issue #2556] by @VijayKalmath in https://github.com/huggingface/datasets/pull/4545
• Properly raise FileNotFound even if the dataset is private by @lhoestq in https://github.com/huggingface/datasets/pull/4536
• Fix hashing for python 3.9 by @lhoestq in https://github.com/huggingface/datasets/pull/4516
• [CI] Fix some warnings by @lhoestq in https://github.com/huggingface/datasets/pull/4547
• Validate new_fingerprint passed by user by @lhoestq in https://github.com/huggingface/datasets/pull/4587
• Update CI Windows orb by @albertvillanova in https://github.com/huggingface/datasets/pull/4604
• Perform hidden file check on relative data file path by @mariosasko in https://github.com/huggingface/datasets/pull/4551
• Align more metadata with other repo types (models,spaces) by @julien-c in https://github.com/huggingface/datasets/pull/4607
• Align/fix license metadata info by @julien-c in https://github.com/huggingface/datasets/pull/4613
• Preserve member order by MockDownloadManager.iter_archive by @albertvillanova in https://github.com/huggingface/datasets/pull/4611
• Add authentication tip to load_dataset by @mariosasko in https://github.com/huggingface/datasets/pull/4577
• Stop dropping columns in to_tf_dataset() before we load batches by @Rocketknight1 in https://github.com/huggingface/datasets/pull/4553
• fix(dataset_wrappers): Fixes access to fsspec.asyn in torch_iterable_dataset.py. by @gugarosa in https://github.com/huggingface/datasets/pull/4630
• Fix xisfile, xgetsize, xisdir, xlistdir in private repo by @lhoestq in https://github.com/huggingface/datasets/pull/4608
• Rename master to main by @lhoestq in https://github.com/huggingface/datasets/pull/4643
• Set HF_SCRIPTS_VERSION to main by @lhoestq in https://github.com/huggingface/datasets/pull/4645
• [Minor fix] Typo correction by @cakiki in https://github.com/huggingface/datasets/pull/4644
• fixed duplicate calculation of spearmanr function in metrics wrapper. by @benlipkin in https://github.com/huggingface/datasets/pull/4627
• Generalize meta_path json file creation in load.py [#4540] by @VijayKalmath in https://github.com/huggingface/datasets/pull/4590
• Fix time type _arrow_to_datasets_dtype conversion by @mariosasko in https://github.com/huggingface/datasets/pull/4628
• Fix _resolve_single_pattern_locally on Windows with multiple drives by @albertvillanova in https://github.com/huggingface/datasets/pull/4660
• Replace assertEqual with assertTupleEqual in unit tests for verbosity by @alvarobartt in https://github.com/huggingface/datasets/pull/4496
• Fix embed_storage on features inside lists/sequences by @mariosasko in https://github.com/huggingface/datasets/pull/4615
• Add links to vision tasks scripts in ADD_NEW_DATASET template by @mariosasko in https://github.com/huggingface/datasets/pull/4512
• Transfer CI to GitHub Actions by @albertvillanova in https://github.com/huggingface/datasets/pull/4659
• Fix mock fsspec by @albertvillanova in https://github.com/huggingface/datasets/pull/4685
• Trigger CI also on push to main by @albertvillanova in https://github.com/huggingface/datasets/pull/4687
• Fix ImageFolder with parameters drop_metadata=True and drop_labels=False (when metadata.jsonl is present) by @polinaeterna in https://github.com/huggingface/datasets/pull/4622
• Skip test_extractor only for zstd param if zstandard not installed by @albertvillanova in https://github.com/huggingface/datasets/pull/4688
• Test extractors for all compression formats by @albertvillanova in https://github.com/huggingface/datasets/pull/4689
• Refactor base extractors by @albertvillanova in https://github.com/huggingface/datasets/pull/4690
• Update create dataset card docs by @stevhliu in https://github.com/huggingface/datasets/pull/4683
• Add text decorators by @stevhliu in https://github.com/huggingface/datasets/pull/4663
• Skip tests only for lz4/zstd params if not installed by @albertvillanova in https://github.com/huggingface/datasets/pull/4704
• Ensure ConcatenationTable.cast uses target_schema metadata by @dtuit in https://github.com/huggingface/datasets/pull/4614
• Docs: Fix same-page haslinks by @mishig25 in https://github.com/huggingface/datasets/pull/4722
• Fix broken link to the Hub by @stevhliu in https://github.com/huggingface/datasets/pull/4726
• Refactor conftest fixtures by @albertvillanova in https://github.com/huggingface/datasets/pull/4723
• Add object detection processing tutorial by @nateraw in https://github.com/huggingface/datasets/pull/4710
• Fix require torchaudio and refactor test requirements by @albertvillanova in https://github.com/huggingface/datasets/pull/4708
• docs: ✏️ fix TranslationVariableLanguages example by @severo in https://github.com/huggingface/datasets/pull/4731
• Pin rouge_score test dependency by @albertvillanova in https://github.com/huggingface/datasets/pull/4735
• Fix named split sorting and remove unnecessary casting by @albertvillanova in https://github.com/huggingface/datasets/pull/4714
• Make cast in from_pandas more robust by @mariosasko in https://github.com/huggingface/datasets/pull/4703
• Make Extractor accept Path as input by @albertvillanova in https://github.com/huggingface/datasets/pull/4718
• Refactor Hub tests by @albertvillanova in https://github.com/huggingface/datasets/pull/4729
• Fix to dict conversion of DatasetInfo/Features by @mariosasko in https://github.com/huggingface/datasets/pull/4741

New Contributors

• @hugovk made their first contribution in https://github.com/huggingface/datasets/pull/4539
• @VijayKalmath made their first contribution in https://github.com/huggingface/datasets/pull/4545
• @gugarosa made their first contribution in https://github.com/huggingface/datasets/pull/4630
• @benlipkin made their first contribution in https://github.com/huggingface/datasets/pull/4627
• @YooSungHyun made their first contribution in https://github.com/huggingface/datasets/pull/4409
• @hobson made their first contribution in https://github.com/huggingface/datasets/pull/4517
• @khushmeeet made their first contribution in https://github.com/huggingface/datasets/pull/4554
• @dtuit made their first contribution in https://github.com/huggingface/datasets/pull/4614

Full Changelog: https://github.com/huggingface/datasets/compare/2.3.2...2.4.0