Datasets:

---

Francesco-A

/

github-issues_huggingface-datasets

like 0

[]

[ "I discussed with Patrick and I think we could have a nice addition: have a parameter `concatenate_texts` that, if `True`, uses the old implementation.\r\n\r\nBy default `concatenate_texts` would be `False`, so that sentences are evaluated independently, and to save resources (the WER computation has a quadratic complexity).\r\n\r\nSome users might still want to use the old implementation.", "@lhoestq @patrickvonplaten are you sure of the parameter name `concatenate_texts`? I was thinking about something like `iter`...", "Not sure about the name, if you can improve it feel free to do so ^^'\r\nThe old implementation computes the WER on the concatenation of all the input texts, while the new one makes WER measures computation independent for each reference/prediction pair.\r\nThat's why I thought of `concatenate_texts`", "@lhoestq yes, but the end user does not necessarily know the details of the implementation of the WER computation.\r\n\r\nFrom the end user perspective I think it might make more sense: how do you want to compute the metric?\r\n- all in once, more RAM memory needed?\r\n- iteratively, less RAM requirements?\r\n\r\nBecause of that I was thinking of something like `iter` or `iterative`...", "Personally like `concatenate_texts` better since I feel like `iter` or `iterate` are quite vague", "Therefore, you can merge... ;)", "Ok ! merging :)" ]

[]

[ "@lhoestq looks like a lot of files were updated... shall I create a new PR?", "Hi @chaitnayabasava ! you can try rebasing and see if that fixes the number of files changed, otherwise please do open a new PR with only the relevant files and close this one :) ", "Created a new PR #1230.\r\nclosing this one :)" ]

[]

[ "IMO it makes more sense to have `ignore_url_params` as an attribute of `DownloadConfig` to avoid defining a new argument in `DownloadManger`'s methods.", "@mariosasko that would make sense to me too, but it seems like `DownloadConfig` wasn't intended to be modified from a dataset loading script. @lhoestq wdyt?", "We can expose `DownloadConfig` as a property of `DownloadManager`, and then in the script before the download call we could do: `dl_manager.download_config.ignore_url_params = True`. But yes, let's hear what Quentin thinks.", "Oh indeed that's a great idea. This parameter is similar to others like `download_config.use_etag` that defines the behavior of the download and caching, so it's better if we have it there, and expose the `download_config`", "Implemented it via `dl_manager.download_config.ignore_url_params` now, and also added a usage example above :) " ]

[]

[ "This code behaves as expected on my local machine or in Colab. Which version of `soundfile` do you have installed? MP3 requires `soundfile>=0.12.1`.", "I upgraded the `soundfile` and it's working now! \r\nThanks @mariosasko for the help!" ]

[]

[]

[]

[]

[]

[]

[]

[]

[ { "color": "d73a4a", "default": true, "description": "Something isn't working", "id": 1935892857, "name": "bug", "node_id": "MDU6TGFiZWwxOTM1ODkyODU3", "url": "https://api.github.com/repos/huggingface/datasets/labels/bug" } ]

[ "Hi @junwang-wish, thanks for reporting.\r\n\r\nUnfortunately, I'm not able to reproduce the bug. Which version of `datasets` are you using? Does the problem persist if you update `datasets`?\r\n```shell\r\npip install -U datasets\r\n``` ", "Thanks @albertvillanova I updated `datasets` from `2.5.1` to `2.5.2` and tested copying the `json.gz` to a different directory and my mind was blown:\r\n\r\n```python\r\nfpath = '/data/junwang/.cache/general/57b6f2314cbe0bc45dda5b78f0871df2/test.json.gz'\r\nds_panda = DatasetDict(\r\n test=Dataset.from_pandas(\r\n pd.read_json(fpath, lines=True)\r\n )\r\n)\r\nds_direct = load_dataset(\r\n 'json', data_files={\r\n 'test': fpath\r\n }, features=Features(\r\n text_input=Value(dtype=\"string\", id=None),\r\n text_output=Value(dtype=\"string\", id=None)\r\n )\r\n)\r\nlen(ds_panda['test']), len(ds_direct['test'])\r\n```\r\nproduces \r\n```python\r\nUsing custom data configuration default-0e6cf24134163e8b\r\nFound cached dataset json (/data/junwang/.cache/huggingface/datasets/json/default-0e6cf24134163e8b/0.0.0/e6070c77f18f01a5ad4551a8b7edfba20b8438b7cad4d94e6ad9378022ce4aab)\r\n(1, 0)\r\n```\r\nbut then I ran below command to see if the same file in a different directory leads to same discrepancy\r\n```shell\r\ncp /data/junwang/.cache/general/57b6f2314cbe0bc45dda5b78f0871df2/test.json.gz tmp_test.json.gz\r\n```\r\nand so I ran\r\n```python\r\nfpath = 'tmp_test.json.gz'\r\nds_panda = DatasetDict(\r\n test=Dataset.from_pandas(\r\n pd.read_json(fpath, lines=True)\r\n )\r\n)\r\nds_direct = load_dataset(\r\n 'json', data_files={\r\n 'test': fpath\r\n }, features=Features(\r\n text_input=Value(dtype=\"string\", id=None),\r\n text_output=Value(dtype=\"string\", id=None)\r\n )\r\n)\r\nlen(ds_panda['test']), len(ds_direct['test'])\r\n```\r\nand behold, I get \r\n```python\r\nUsing custom data configuration default-f679b32ab0008520\r\nDownloading and preparing dataset json/default to /data/junwang/.cache/huggingface/datasets/json/default-f679b32ab0008520/0.0.0/e6070c77f18f01a5ad4551a8b7edfba20b8438b7cad4d94e6ad9378022ce4aab...\r\nDataset json downloaded and prepared to /data/junwang/.cache/huggingface/datasets/json/default-f679b32ab0008520/0.0.0/e6070c77f18f01a5ad4551a8b7edfba20b8438b7cad4d94e6ad9378022ce4aab. Subsequent calls will reuse this data.\r\n(1, 1)\r\n```\r\nThey match now !\r\n\r\nThis problem happens regardless of the shell I use (VScode jupyter extension or plain old Python REPL). \r\n\r\nI attached the `json.gz` here for reference: [test.json.gz](https://github.com/huggingface/datasets/files/9734843/test.json.gz)\r\n\r\n" ]

[ { "color": "d73a4a", "default": true, "description": "Something isn't working", "id": 1935892857, "name": "bug", "node_id": "MDU6TGFiZWwxOTM1ODkyODU3", "url": "https://api.github.com/repos/huggingface/datasets/labels/bug" } ]

[ "Hi! \r\n\r\nThis issue stems from `datasets` having `py.typed` defined (see https://github.com/microsoft/pyright/discussions/3764#discussioncomment-3282142) - to avoid it, we would either have to remove `py.typed` (added to be compliant with PEP-561) or export the names with `__all__`/`from .submodule import name as name`.\r\n\r\nTransformers is fine as it no longer has `py.typed` (removed in https://github.com/huggingface/transformers/pull/18485)\r\n\r\nWDYT @lhoestq @albertvillanova @polinaeterna \r\n\r\n@sgugger's point makes sense - we should either be \"properly typed\" (have py.typed + mypy tests) or drop `py.typed` as Transformers did (I like this option better).\r\n\r\n(cc @Wauplin since `huggingface_hub` has the same issue.)", "I'm fine with dropping it, but autotrain people won't be happy @SBrandeis ", "> (cc @Wauplin since huggingface_hub has the same issue.)\r\n\r\nHmm maybe we have the same issue but I haven't been able to reproduce something similar to `\"load_dataset\" is not exported from module \"datasets\"` message (using VSCode+Pylance -that is powered by Pyright). `huggingface_hub` contains a `py.typed` file but the package itself is actually typed. We are running `mypy` in our CI tests since ~3 months and so far it seems to be ok. But happy to change if it causes some issues with linters.\r\n\r\nAlso the top-level [`__init__.py`](https://github.com/huggingface/huggingface_hub/blob/main/src/huggingface_hub/__init__.py) is quite different in `hfh` than `datasets` (at first glance). We have a section at the bottom to import all high level methods/classes in a `if TYPE_CHECKING` block.", "@Wauplin I only get the error if I use Pyright's CLI tool or the Pyright extension (not sure why, but Pylance also doesn't report this issue on my machine)\r\n\r\n> Also the top-level [`__init__.py`](https://github.com/huggingface/huggingface_hub/blob/main/src/huggingface_hub/__init__.py) is quite different in `hfh` than `datasets` (at first glance). We have a section at the bottom to import all high level methods/classes in a `if TYPE_CHECKING` block.\r\n\r\nI tried to fix the issue with `TYPE_CHECKING`, but it still fails if `py.typed` is present.", "@mariosasko thank for the tip. I have been able to reproduce the issue as well. I would be up for including a (huge) static `__all__` variable in the `__init__.py` (since the file is already generated automatically in `hfh`) but honestly I don't think it's worth the hassle. \r\n\r\nI'll delete the `py.typed` file in `huggingface_hub` to be consistent between HF libraries. I opened a PR here: https://github.com/huggingface/huggingface_hub/pull/1329", "I am getting this error in google colab today:\r\n\r\n\r\n\r\nThe code runs just fine too." ]

[ { "color": "e99695", "default": false, "description": "Requesting to add a new dataset", "id": 2067376369, "name": "dataset request", "node_id": "MDU6TGFiZWwyMDY3Mzc2MzY5", "url": "https://api.github.com/repos/huggingface/datasets/labels/dataset%20request" } ]

[ "Hi @lbourdois, yes please share it with us", "@mariamabarham \r\nI put all the datasets on this drive: https://1drv.ms/u/s!Ao2Rcpiny7RFinDypq7w-LbXcsx9?e=iVsEDh\r\n\r\n\r\nSome information : \r\n• For FLUE, the quote used is\r\n\r\n> @misc{le2019flaubert,\r\n> title={FlauBERT: Unsupervised Language Model Pre-training for French},\r\n> author={Hang Le and Loïc Vial and Jibril Frej and Vincent Segonne and Maximin Coavoux and Benjamin Lecouteux and Alexandre Allauzen and Benoît Crabbé and Laurent Besacier and Didier Schwab},\r\n> year={2019},\r\n> eprint={1912.05372},\r\n> archivePrefix={arXiv},\r\n> primaryClass={cs.CL}\r\n> }\r\n\r\n• The Github repo of FLUE is avaible here : https://github.com/getalp/Flaubert/tree/master/flue\r\n\r\n\r\n\r\nInformation related to the different tasks of FLUE : \r\n\r\n**1. Classification**\r\nThree dataframes are available: \r\n- Book\r\n- DVD\r\n- Music\r\nFor each of these dataframes is available a set of training and test data, and a third one containing unlabelled data.\r\n\r\nCitation : \r\n>@dataset{prettenhofer_peter_2010_3251672,\r\n author = {Prettenhofer, Peter and\r\n Stein, Benno},\r\n title = {{Webis Cross-Lingual Sentiment Dataset 2010 (Webis- \r\n CLS-10)}},\r\n month = jul,\r\n year = 2010,\r\n publisher = {Zenodo},\r\n doi = {10.5281/zenodo.3251672},\r\n url = {https://doi.org/10.5281/zenodo.3251672}\r\n}\r\n\r\n\r\n**2. Paraphrasing** \r\nFrench part of the PAWS-X dataset (https://github.com/google-research-datasets/paws).\r\nThree dataframes are available: \r\n- train\r\n- dev\r\n- test \r\n\r\nCitation : \r\n> @InProceedings{pawsx2019emnlp,\r\n> title = {{PAWS-X: A Cross-lingual Adversarial Dataset for Paraphrase Identification}},\r\n> author = {Yang, Yinfei and Zhang, Yuan and Tar, Chris and Baldridge, Jason},\r\n> booktitle = {Proc. of EMNLP},\r\n> year = {2019}\r\n> }\r\n\r\n\r\n\r\n**3. Natural Language Inference**\r\nFrench part of the XNLI dataset (https://github.com/facebookresearch/XNLI).\r\nThree dataframes are available: \r\n- train\r\n- dev\r\n- test \r\n\r\nFor the dev and test datasets, extra columns compared to the train dataset were available so I left them in the dataframe (I didn't know if these columns could be useful for other tasks or not). \r\nIn the context of the FLUE benchmark, only the columns gold_label, sentence1 and sentence2 are useful.\r\n\r\n\r\nCitation : \r\n\r\n> @InProceedings{conneau2018xnli,\r\n> author = \"Conneau, Alexis\r\n> and Rinott, Ruty\r\n> and Lample, Guillaume\r\n> and Williams, Adina\r\n> and Bowman, Samuel R.\r\n> and Schwenk, Holger\r\n> and Stoyanov, Veselin\",\r\n> title = \"XNLI: Evaluating Cross-lingual Sentence Representations\",\r\n> booktitle = \"Proceedings of the 2018 Conference on Empirical Methods\r\n> in Natural Language Processing\",\r\n> year = \"2018\",\r\n> publisher = \"Association for Computational Linguistics\",\r\n> location = \"Brussels, Belgium\",\r\n\r\n\r\n**4. Parsing**\r\nThe dataset used by the FLUE authors for this task is not freely available.\r\nUsers of your library will therefore not be able to access it.\r\nNevertheless, I think maybe it is useful to add a link to the site where to request this dataframe: http://ftb.linguist.univ-paris-diderot.fr/telecharger.php?langue=en \r\n(personally it was sent to me less than 48 hours after I requested it).\r\n\r\n\r\n**5. Word Sense Disambiguation Tasks**\r\n5.1 Verb Sense Disambiguation\r\n\r\nTwo dataframes are available: train and test\r\nFor both dataframes, 4 columns are available: document, sentence, lemma and word.\r\nI created the document column thinking that there were several documents in the dataset but afterwards it turns out that there were not: several sentences but only one document. It's up to you to keep it or not when importing these two dataframes.\r\n\r\nThe sentence column is used to determine to which sentence the word in the word column belongs. It is in the form of a dictionary {'id': 'd000.s001', 'idx': '1'}. I thought for a while to keep only the idx because the id doesn't matter any more information. Nevertheless for the test dataset, the dictionary has an extra value indicating the source of the sentence. I don't know if it's useful or not, that's why I left the dictionary just in case. The user is free to do what he wants with it.\r\n\r\nCitation : \r\n\r\n> Segonne, V., Candito, M., and Crabb ́e, B. (2019). Usingwiktionary as a resource for wsd: the case of frenchverbs. InProceedings of the 13th International Confer-ence on Computational Semantics-Long Papers, pages259–270\r\n\r\n5.2 Noun Sense Disambiguation\r\nTwo dataframes are available: 2 train and 1 test\r\n\r\nI confess I didn't fully understand the procedure for this task.\r\n\r\nCitation : \r\n\r\n> @dataset{loic_vial_2019_3549806,\r\n> author = {Loïc Vial},\r\n> title = {{French Word Sense Disambiguation with Princeton \r\n> WordNet Identifiers}},\r\n> month = nov,\r\n> year = 2019,\r\n> publisher = {Zenodo},\r\n> version = {1.0},\r\n> doi = {10.5281/zenodo.3549806},\r\n> url = {https://doi.org/10.5281/zenodo.3549806}\r\n> }\r\n\r\nFinally, additional information about FLUE is available in the FlauBERT publication : \r\nhttps://arxiv.org/abs/1912.05372 (p. 4).\r\n\r\n\r\nHoping to have provided you with everything you need to add this benchmark :) \r\n", "https://github.com/huggingface/datasets/pull/943" ]

[]

[ "Hii please follow me", "thank you" ]

[]

[ "_The documentation is not available anymore as the PR was closed or merged._" ]

[]

[ "merging since the CI is fixed on master" ]

[]

[ "Thanks, this was fixed with #135 :)" ]

[]

[ "> The implementation makes heavy use of multiprocessing which this PR does not touch. Is this conflicting with multiprocessing natively integrated in datasets?\r\n\r\nIt should work normally, but feel free to test it.\r\nThere is some documentation about using metrics in a distributed setup that uses multiprocessing [here](https://huggingface.co/docs/datasets/loading.html?highlight=rank#distributed-setup)\r\nYou can test to spawn several processes where each process would load the metric. Then in each process you add some references/predictions to the metric. Finally you call compute() in each process and on process 0 it should return the result on all the references/predictions\r\n\r\nLet me know if you have questions or if I can help", "Is there a good way to debug the Windows tests? I suspect it is an issue with `multiprocessing`, but I can't see the error messages.", "Indeed it has an issue on windows.\r\nIn your example it's supposed to output\r\n```python\r\n{'pass@1': 0.5, 'pass@2': 1.0}\r\n```\r\nbut it gets\r\n```python\r\n{'pass@1': 0.0, 'pass@2': 0.0}\r\n```\r\n\r\nI'm not on my windows machine today so I can't take a look at it. I can dive into it early next week if you want", "> I'm not on my windows machine today so I can't take a look at it. I can dive into it early next week if you want\r\n\r\nThat would be great - unfortunately I have no access to a windows machine at the moment. I am quite sure it is an issue with in exectue.py because of multiprocessing.\r\n" ]

[]

[]

[]

[ "Thanks for reporting ! The issue doesn't come from shuffling, but from `beans` row order not being deterministic:\r\n\r\nhttps://huggingface.co/datasets/beans/blob/main/beans.py uses `dl_manager.iter_files` on ZIP archives and the file order doesn't seen to be deterministic and changes across machines", "Thank you for noticing indeed!", "This is still a bug, so I'd keep this one open if you don't mind ;)", "Besides the linked PR, to make the loading process fully deterministic, I believe we should also sort the data files [here](https://github.com/huggingface/datasets/blob/df4bdd365f2abb695f113cbf8856a925bc70901b/src/datasets/data_files.py#L276) and [here](https://github.com/huggingface/datasets/blob/df4bdd365f2abb695f113cbf8856a925bc70901b/src/datasets/data_files.py#L485) (e.g. fsspec's `LocalFileSystem.glob` relies on `os.scandir`, which yields the contents in arbitrary order). My concern is the overhead of these sorts... Maybe we could introduce a new flag to `load_dataset` similar to TFDS' [`shuffle_files`](https://www.tensorflow.org/datasets/determinism#determinism_when_reading) or sort only if the number of data files is small?", "We already return the result sorted at the end of `_resolve_single_pattern_locally` and `_resolve_single_pattern_in_dataset_repository` if I'm not mistaken", "@lhoestq Oh, you are right. Feel free to ignore my comment.", "I think the corresponding PR is ready to be merged :hugs: ", "@albertvillanova Thanks for the fix!" ]

[ { "color": "d876e3", "default": true, "description": "Further information is requested", "id": 1935892912, "name": "question", "node_id": "MDU6TGFiZWwxOTM1ODkyOTEy", "url": "https://api.github.com/repos/huggingface/datasets/labels/question" } ]

[ "An educated guess: does this refer to the fact that depending on the custom column names in the dataset files (csv in this case), there is a dataset loader being created? and this dataset loader - using the \"custom data configuration\" is used among all jobs running using this particular csv files? (thinking out loud here...)\r\n\r\nIf this is the case, it may be ok for my use case (have to think about it more), still a bit surprising given that datasets caching is disabled (or so I hope) by the lines I pasted above. ", "Hi ! Currently disabling the caching means that all the dataset transform like `map`, `filter` etc. ignore the cache: it doesn't write nor read processed cache files.\r\nHowever `load_dataset` reuses datasets that have already been prepared: it does reload prepared dataset files.\r\n\r\nIndeed from the documentation:\r\n> datasets.set_caching_enabled(boolean: bool)\r\n\r\n> When applying transforms on a dataset, the data are stored in cache files. The caching mechanism allows to reload an existing cache file if it’s already been computed.\r\n> Reloading a dataset is possible since the cache files are named using the dataset fingerprint, which is updated after each transform.\r\n> If disabled, the library will no longer reload cached datasets files when applying transforms to the datasets. More precisely, if the caching is disabled:\r\n> - cache files are always recreated\r\n> - cache files are written to a temporary directory that is deleted when session closes\r\n> - cache files are named using a random hash instead of the dataset fingerprint - use datasets.Dataset.save_to_disk() to save a transformed dataset or it will be deleted when session closes\r\n> - caching doesn’t affect datasets.load_dataset(). If you want to regenerate a dataset from scratch you should use the download_mode parameter in datasets.load_dataset().", "Thank you for the clarification. \r\n\r\nThis is a bit confusing. On one hand, it says that cache files are always recreated and written to a temporary directory that is removed; on the other hand the last bullet point makes me think that since the default according to the docs for `download_mode (Optional datasets.GenerateMode) – select the download/generate mode - Default to REUSE_DATASET_IF_EXISTS` => it almost sounds that it could reload prepared dataset files. Where are these files stored? I guess not in the temporary directory that is removed... \r\n\r\nI find this type of api design error-prone. When I see as a programmer `datasets.set_caching_enabled(False)` I expect no reuse of anything in the cache. ", "It would be nice if the documentation elaborated on all the possible values for `download_mode` and/or a link to `datasets.GenerateMode`. \r\nThis info here:\r\n```\r\n \"\"\"`Enum` for how to treat pre-existing downloads and data.\r\n The default mode is `REUSE_DATASET_IF_EXISTS`, which will reuse both\r\n raw downloads and the prepared dataset if they exist.\r\n The generations modes:\r\n | | Downloads | Dataset |\r\n | -----------------------------------|-----------|---------|\r\n | `REUSE_DATASET_IF_EXISTS` (default)| Reuse | Reuse |\r\n | `REUSE_CACHE_IF_EXISTS` | Reuse | Fresh |\r\n | `FORCE_REDOWNLOAD` | Fresh | Fresh |\r\n```", "I have another question. Assuming that I understood correctly and there is reuse of datasets files when caching is disabled (!), I'm guessing there is a directory that is created based on some information on the dataset file. I'm interested in the situation where I'm loading a (custom) dataset from local disk. What information is used to create the directory/filenames where the files are stored?\r\n\r\nI'm concerned about the following scenario: if I have a file, let's say `train.csv` at path `the_path`, run once, the dataset is prepared, some models are run, etc. Now let's say there is an issue and I recreate `train.csv` at the same path `the_path`. Is there enough information in the temporary name/hash to *not* reload the *old* prepared dataset (e.g., timestamp of the file)? Or is it going to reload the *old* prepared file? ", "Thanks for the feedback, we'll work in improving this aspect of the documentation.\r\n\r\n> Where are these files stored? I guess not in the temporary directory that is removed...\r\n\r\nWe're using the Arrow file format to load datasets. Therefore each time you load a dataset, it is prepared as an arrow file on your disk. By default the file is located in the ~/.cache/huggingface/datasets/<dataset_name>/<config_id>/<version> directory.\r\n\r\n> What information is used to create the directory/filenames where the files are stored?\r\n\r\nThe config_id contains a hash that takes into account:\r\n- the dataset loader used and its source code (e.g. the \"csv\" loader)\r\n- the arguments passed to the loader (e.g. the csv delimiter)\r\n- metadata of the local data files if any (e.g. their timestamps)\r\n\r\n> I'm concerned about the following scenario: if I have a file, let's say train.csv at path the_path, run once, the dataset is prepared, some models are run, etc. Now let's say there is an issue and I recreate train.csv at the same path the_path. Is there enough information in the temporary name/hash to not reload the old prepared dataset (e.g., timestamp of the file)? Or is it going to reload the old prepared file?\r\n\r\nYes the timestamp of the local csv file is taken into account. If you edit your csv file, the config_id will change and loading the dataset will create a new arrow file.", "Thank you for all your clarifications, really helpful! \r\n\r\nIf you have the bandwidth, please do revisit the api wrt cache disabling. Anywhere in the computer stack (hardware included) where you disable the cache, one assumes there is no caching that happens. ", "That makes total sense indeed !\r\nI think we can do the change", "I have another question about caching, this time in the case where FORCE_REDOWNLOAD is used to load the dataset, the datasets cache is one directory as defined by HF_HOME and there are multiple concurrent jobs running in a cluster using the same local dataset (i.e., same local files in the cluster). Does anything in the naming convention and/or file access/locking that you're using prevent race conditions between the concurrent jobs on the caching of the local dataset they all use?\r\n\r\nI noticed some errors (can provide more details if helpful) in load_dataset/prepare_split that lead to my question above. \r\n\r\nLet me know if my question is clear, I can elaborate more if needed @lhoestq Thank you!", "I got another error that convinces me there is a race condition (one of the test files had zero samples at prediction time). I think it comes down to the fact that the `config_id` above (used in the naming for the cache) has no information on who's touching the data. If I have 2 concurrent jobs, both loading the same dataset and forcing redownload, they may step on each other foot/caching of the dataset. ", "We're using a locking mechanism to prevent two processes from writing at the same time. The locking is based on the `filelock` module.\r\nAlso directories that are being written use a suffix \".incomplete\" so that reading is not possible on a dataset being written.\r\n\r\nDo you think you could provide a simple code to reproduce the race condition you experienced ?", "I can provide details about the code I'm running (it's really-really close to some official samples from the huggingface transformers examples, I can point to the exact sample file, I kept a record of that). I can also describe in which conditions this race occurs (I'm convinced it has to do with forcing the redownloading of the dataset, I've been running hundreds of experiments before and didn't have a problem before I forced the redownload). I also can provide samples of the different stack errors I get and some details about the level of concurrency of jobs I was running. I can also try to imagine how the race manifests (I'm fairly sure that it's a combo of one job cleaning up and another job being in the middle of the run).\r\n\r\nHowever, I have to cleanup all this to make sure I'm no spilling any info I shouldn't be spilling. I'll try to do it by the end of the week, if you think all this is helpful. \r\n\r\nFor now, I have a workaround. Don't use forcing redownloading. And to be ultra careful (although I don't think this is a problem), I run a series of jobs that will prepare the datasets and I know there is no concurrency wrt the dataset. Once that's done (and I believe even having multiple jobs loading the datasets at the same time doesn't create problems, as long as REUSE_DATASET_IF_EXISTS is the policy for loading the dataset, so the filelock mechanism you're using is working in that scenario), the prepared datasets will be reused, no race possible in any way. \r\n\r\nThanks for all the details you provided, it helped me understand the underlying implementation and coming up with workarounds when I ran into issues. ", "Hi! I have the same challenge with caching, where the **.cache** folder is required even though it isn't possible for me.\r\n\r\nI'd like to run transformers in Snowflake, using Snowpark for Python, this would mean I could provide configurable transformers in real-time for business users without having data leave an environment (for security reasons). With no need for data transfer,n the compute is faster. It is a large use case - is it possible to entirely disable caching in certain scenarios?\r\n@lhoestq ?\r\n", "You can try to change the location of the cache folder using the `HF_CACHE_HOME` environment variable, and set a location where you have read/write access.", "Thanks @lhoestq \r\n\r\nI wanted to do that, however, snowflake does not allow it to write at all. I'm asking around to see if they can help me out with that issue 😅" ]

[]

[]

[ { "color": "a2eeef", "default": true, "description": "New feature or request", "id": 1935892871, "name": "enhancement", "node_id": "MDU6TGFiZWwxOTM1ODkyODcx", "url": "https://api.github.com/repos/huggingface/datasets/labels/enhancement" }, { "color": "c5def5", "default": false, "description": "Generic discussion on the library", "id": 2067400324, "name": "generic discussion", "node_id": "MDU6TGFiZWwyMDY3NDAwMzI0", "url": "https://api.github.com/repos/huggingface/datasets/labels/generic%20discussion" } ]

[ "I like this feature! I think the first question we should decide on is how to convert all datasets into the same format. In T5, the authors decided to format every dataset into a text-to-text format. If the dataset had \"multiple\" inputs like MNLI, the inputs were concatenated. So in MNLI the input:\r\n\r\n> - **Hypothesis**: The St. Louis Cardinals have always won.\r\n> \r\n> - **Premise**: yeah well losing is i mean i’m i’m originally from Saint Louis and Saint Louis Cardinals when they were there were uh a mostly a losing team but \r\n\r\nwas flattened to a single input:\r\n\r\n> mnli hypothesis: The St. Louis Cardinals have always won. premise:\r\n> yeah well losing is i mean i’m i’m originally from Saint Louis and Saint Louis Cardinals\r\n> when they were there were uh a mostly a losing team but.\r\n\r\nThis flattening is actually a very simple operation in `nlp` already. You would just need to do the following:\r\n\r\n```python \r\ndef flatten_inputs(example):\r\n return {\"input\": \"mnli hypothesis: \" + example['hypothesis'] + \" premise: \" + example['premise']}\r\n\r\nt5_ready_mnli_ds = mnli_ds.map(flatten_inputs, remove_columns=[<all columns except output>])\r\n```\r\n\r\nSo I guess converting the datasets into the same format can be left to the user for now. \r\nThen the question is how we can merge the datasets. I would probably be in favor of a simple \r\n\r\n```python \r\ndataset.add()\r\n```\r\n\r\nfunction that checks if the dataset is of the same format and if yes merges the two datasets. Finally, how should the sampling be implemented? **Examples-proportional mixing** corresponds to just merging the datasets and shuffling. For the other two sampling approaches we would need some higher-level features, maybe even a `dataset.sample()` function for merged datasets. \r\n\r\nWhat are your thoughts on this @thomwolf @lhoestq @ghomasHudson @enzoampil ?", "I agree that we should leave the flattening of the dataset to the user for now. Especially because although the T5 framing seems obvious, there are slight variations on how the T5 authors do it in comparison to other approaches such as gpt-3 and decaNLP.\r\n\r\nIn terms of sampling, Examples-proportional mixing does seem the simplest to implement so would probably be a good starting point.\r\n\r\nTemperature-scaled mixing would probably most useful, offering flexibility as it can simulate the other 2 methods by setting the temperature parameter. There is a [relevant part of the T5 repo](https://github.com/google-research/text-to-text-transfer-transformer/blob/03c94165a7d52e4f7230e5944a0541d8c5710788/t5/data/utils.py#L889-L1118) which should help with implementation.\r\n\r\nAccording to the T5 authors, equal-mixing performs worst. Among the other two methods, tuning the K value (the artificial dataset size limit) has a large impact.\r\n", "I agree with going with temperature-scaled mixing for its flexibility!\r\n\r\nFor the function that combines the datasets, I also find `dataset.add()` okay while also considering that users may want it to be easy to combine a list of say 10 data sources in one go.\r\n\r\n`dataset.sample()` should also be good. By the looks of it, we're planning to have as main parameters: `temperature`, and `K`.\r\n\r\nOn converting the datasets to the same format, I agree that we can leave these to the users for now. But, I do imagine it'd be an awesome feature for the future to have this automatically handled, based on a chosen *approach* to formatting :smile: \r\n\r\nE.g. T5, GPT-3, decaNLP, original raw formatting, or a contributed way of formatting in text-to-text. ", "This is an interesting discussion indeed and it would be nice to make multi-task easier.\r\n\r\nProbably the best would be to have a new type of dataset especially designed for that in order to easily combine and sample from the multiple datasets.\r\n\r\nThis way we could probably handle the combination of datasets with differing schemas as well (unlike T5).", "@thomwolf Are you suggesting making a wrapper class which can take existing datasets as arguments and do all the required sampling/combining, to present the same interface as a normal dataset?\r\n\r\nThat doesn't seem too complicated to implement.\r\n", "I guess we're looking at the end user writing something like:\r\n``` python\r\nds = nlp.load_dataset('multitask-t5',datasets=[\"squad\",\"cnn_dm\",...], k=1000, t=2.0)\r\n```\r\nUsing the t5 method of combining here (or this could be a function passed in as an arg) \r\n\r\nPassing kwargs to each 'sub-dataset' might become tricky.", "From thinking upon @thomwolf 's suggestion, I've started experimenting:\r\n```python\r\nclass MultitaskDataset(DatasetBuilder):\r\n def __init__(self, *args, **kwargs):\r\n super(MultitaskDataset, self).__init__(*args, **kwargs)\r\n self._datasets = kwargs.get(\"datasets\")\r\n\r\n def _info(self):\r\n return nlp.DatasetInfo(\r\n description=_DESCRIPTION,\r\n features=nlp.Features({\r\n \"source\": nlp.Value(\"string\"),\r\n \"target\": nlp.Sequence(nlp.Value(\"string\"))\r\n })\r\n )\r\n\r\n def _get_common_splits(self):\r\n '''Finds the common splits present in all self._datasets'''\r\n min_set = None\r\n for dataset in self._datasets:\r\n if min_set != None:\r\n min_set.intersection(set(dataset.keys()))\r\n else:\r\n min_set = set(dataset.keys())\r\n return min_set\r\n\r\n....\r\n\r\n# Maybe this?:\r\nsquad = nlp.load_dataset(\"squad\")\r\ncnn_dm = nlp.load_dataset(\"cnn_dailymail\",\"3.0.0\")\r\nmultitask_dataset = nlp.load_dataset(\r\n 'multitask_dataset',\r\n datasets=[squad,cnn_dailymail], \r\n k=1000, \r\n t=2.0\r\n)\r\n\r\n```\r\n\r\nDoes anyone know what methods of `MultitaskDataset` I would need to implement? Maybe `as_dataset` and `download_and_prepare`? Most of these should be just calling the methods of the sub-datasets. \r\n\r\nI'm assuming DatasetBuilder is better than the more specific `GeneratorBasedBuilder`, `BeamBasedBuilder`, etc....\r\n\r\nOne of the other problems is that the dataset size is unknown till you construct it (as you can pick the sub-datasets). Am hoping not to need to make changes to `nlp.load_dataset` just for this class.\r\n\r\nI'd appreciate it if anyone more familiar with nlp's internal workings could tell me if I'm on the right track!", "I think I would probably go for a `MultiDataset` wrapper around a list of `Dataset`.\r\n\r\nI'm not sure we need to give it `k` and `t` parameters at creation, it can maybe be something along the lines of:\r\n```python\r\nsquad = nlp.load_dataset(\"squad\")\r\ncnn_dm = nlp.load_dataset(\"cnn_dailymail\",\"3.0.0\")\r\n\r\nmultitask_dataset = nlp.MultiDataset(squad, cnn_dm)\r\n\r\nbatch = multitask_dataset.sample(10, temperature=2.0, k=1000)\r\n```\r\n\r\nThe first proof-of-concept for multi-task datasets could definitely require that the provided datasets have the same name/type for columns (if needed you easily rename/cast a column prior to instantiating the `MultiDataset`).\r\n\r\nIt's good to think about it for some time though and don't overfit too much on the T5 examples (in particular for the ways/kwargs for sampling among datasets).", "The problem with changing `k` and `t` per sampling is that you'd have to somehow remember which examples you'd already returned while re-weighting the remaining examples based on the new `k` and `t`values. It seems possible but complicated (I can't really see a reason why you'd want to change the weighting of datasets after you constructed the multidataset).\r\n\r\nWouldn't it be convenient if it implemented the dataset interface? Then if someone has code using a single nlp dataset, they can replace it with a multitask combination of more datasets without having to change other code. We would at least need to be able to pass it into a `DataLoader`.\r\n\r\n", "A very janky (but working) implementation of `multitask_dataset.sample()` could be something like this:\r\n```python\r\nimport nlp\r\nimport torch\r\n\r\nclass MultiDataset():\r\n def __init__(self, *args, temperature=2.0, k=1000, maximum=None, scale=1):\r\n self.datasets = args\r\n self._dataloaders = {}\r\n for split in self._get_common_splits():\r\n split_datasets = [ds[split] for ds in self.datasets]\r\n mixing_rates = self._calc_mixing_rates(split_datasets,temperature, k, maximum, scale)\r\n weights = []\r\n for i in range(len(self.datasets)):\r\n weights += [mixing_rates[i]]*len(self.datasets[i][split])\r\n self._dataloaders[split] = torch.utils.data.DataLoader(torch.utils.data.ConcatDataset(split_datasets),\r\n sampler=torch.utils.data.sampler.WeightedRandomSampler(\r\n num_samples=len(weights),\r\n weights = weights,\r\n replacement=True),\r\n shuffle=False)\r\n\r\n def _get_common_splits(self):\r\n '''Finds the common splits present in all self.datasets'''\r\n min_set = None\r\n for dataset in self.datasets:\r\n if min_set != None:\r\n min_set.intersection(set(dataset.keys()))\r\n else:\r\n min_set = set(dataset.keys())\r\n return min_set\r\n\r\n\r\n def _calc_mixing_rates(self,datasets, temperature=2.0, k=1000, maximum=None, scale=1):\r\n '''Work out the weighting of each dataset based on t and k'''\r\n mixing_rates = []\r\n for dataset in datasets:\r\n rate = len(dataset)\r\n rate *= scale\r\n if maximum:\r\n rate = min(rate, maximum)\r\n if temperature != 1.0:\r\n rate = rate ** (1.0/temperature)\r\n mixing_rates.append(rate)\r\n return mixing_rates\r\n\r\n def sample(self,n,split):\r\n batch = []\r\n for example in self._dataloaders[split]:\r\n batch.append(example)\r\n n -= 1\r\n if n == 0:\r\n return batch\r\n\r\n\r\ndef flatten(dataset,flatten_fn):\r\n for k in dataset.keys():\r\n if isinstance(dataset[k],nlp.Dataset):\r\n dataset[k] = dataset[k].map(flatten_fn,remove_columns=dataset[k].column_names)\r\n\r\n# Squad\r\ndef flatten_squad(example):\r\n return {\"source\": \"squad context: \" + example['context'] + \" question: \" + example['question'],\"target\":example[\"answers\"][\"text\"]}\r\nsquad = nlp.load_dataset(\"squad\")\r\nflatten(squad,flatten_squad)\r\n\r\n# CNN_DM\r\ndef flatten_cnn_dm(example):\r\n return {\"source\": \"cnn_dm: \" + example['article'],\"target\":[example[\"highlights\"]]}\r\ncnn_dm = nlp.load_dataset(\"cnn_dailymail\", \"3.0.0\")\r\nflatten(cnn_dm,flatten_cnn_dm)\r\n\r\nmultitask_dataset = MultiDataset(squad, cnn_dm)\r\nbatch = multitask_dataset.sample(100,\"train\")\r\n```\r\n\r\nThere's definitely a more sensible way than embedding `DataLoader`s inside. ", "There is an interesting related investigation by @zphang here https://colab.research.google.com/github/zphang/zphang.github.io/blob/master/files/notebooks/Multi_task_Training_with_Transformers_NLP.ipynb", "Good spot! Here are my thoughts:\r\n\r\n- Aside: Adding `MultitaskModel` to transformers might be a thing to raise - even though having task-specific heads has become unfashionable in recent times in favour of text-to-text type models.\r\n- Adding the task name as an extra field also seems useful for these kind of models which have task-specific heads\r\n- There is some validation of our approach that the user should be expected to `map` datasets into a common form.\r\n- The size-proportional sampling (also called \"Examples-proportional mixing\") used here doesn't perform too badly in the T5 paper (it's comparable to temperature-scaled mixing in many cases but less flexible. This is only reasonable with a `K` maximum size parameter to prevent very large datasets dominating). This might be good for a first prototype using:\r\n ```python\r\n def __iter__(self):\r\n \"\"\"\r\n For each batch, sample a task, and yield a batch from the respective\r\n task Dataloader.\r\n\r\n We use size-proportional sampling, but you could easily modify this\r\n to sample from some-other distribution.\r\n \"\"\"\r\n task_choice_list = []\r\n for i, task_name in enumerate(self.task_name_list):\r\n task_choice_list += [i] * self.num_batches_dict[task_name]\r\n task_choice_list = np.array(task_choice_list)\r\n np.random.shuffle(task_choice_list)\r\n\r\n dataloader_iter_dict = {\r\n task_name: iter(dataloader) \r\n for task_name, dataloader in self.dataloader_dict.items()\r\n }\r\n for task_choice in task_choice_list:\r\n task_name = self.task_name_list[task_choice]\r\n yield next(dataloader_iter_dict[task_name]) \r\n ```\r\n We'd just need to pull samples from the raw datasets and not from `DataLoader`s for each task. We can assume the user has done `dataset.shuffle()` if they want to.\r\n\r\n Other sampling methods can later be implemented by changing how the `task_choice_list` is generated. This should allow more flexibility and not tie us to specific methods for sampling among datasets.\r\n", "Another thought: Multitasking over benchmarks (represented as Meta-datasets in nlp) is probably a common use case. Would be nice to pass an entire benchmark to our `MultiDataset` wrapper rather than having to pass individual components.", "Here's a fully working implementation based on the `__iter__` function of @zphang.\r\n\r\n- I've generated the task choice list in the constructor as it allows us to index into the MultiDataset just like a normal dataset. I'm changing `task_choice_list` into a list of `(dataset_idx, example_idx)` so each entry references a unique dataset example. The shuffling has to be done before this as we don't want to shuffle within each task (we assume this is done by the user if this is what they intend).\r\n- I'm slightly concerned this list could become very large if many large datasets were used. Can't see a way round it at the moment though.\r\n- I've used `task.info.builder_name` as the dataset name. Not sure if this is correct.\r\n- I'd love to add some of the other `Dataset` methods (map, slicing by column, etc...). Would be great to implement the whole interface so a single dataset can be simply replaced by this.\r\n- This does everything on the individual example-level. If some application required batches all from a single task in turn we can't really do that.\r\n\r\n```python\r\nimport nlp\r\nimport numpy as np\r\n\r\nclass MultiDataset:\r\n def __init__(self,tasks):\r\n self.tasks = tasks\r\n\r\n # Create random order of tasks\r\n # Using size-proportional sampling\r\n task_choice_list = []\r\n for i, task in enumerate(self.tasks):\r\n task_choice_list += [i] * len(task)\r\n task_choice_list = np.array(task_choice_list)\r\n np.random.shuffle(task_choice_list)\r\n\r\n # Add index into each dataset\r\n # - We don't want to shuffle within each task\r\n counters = {}\r\n self.task_choice_list = []\r\n for i in range(len(task_choice_list)):\r\n idx = counters.get(task_choice_list[i],0)\r\n self.task_choice_list.append((task_choice_list[i],idx))\r\n counters[task_choice_list[i]] = idx + 1\r\n\r\n\r\n def __len__(self):\r\n return np.sum([len(t) for t in self.tasks])\r\n\r\n def __repr__(self):\r\n task_str = \", \".join([str(t) for t in self.tasks])\r\n return f\"MultiDataset(tasks: {task_str})\"\r\n\r\n def __getitem__(self,key):\r\n if isinstance(key, int):\r\n task_idx, example_idx = self.task_choice_list[key]\r\n task = self.tasks[task_idx]\r\n example = task[example_idx]\r\n example[\"task_name\"] = task.info.builder_name\r\n return example\r\n elif isinstance(key, slice):\r\n raise NotImplementedError()\r\n\r\n def __iter__(self):\r\n for i in range(len(self)):\r\n yield self[i]\r\n\r\n\r\ndef load_multitask(*datasets):\r\n '''Create multitask datasets per split'''\r\n\r\n def _get_common_splits(datasets):\r\n '''Finds the common splits present in all self.datasets'''\r\n min_set = None\r\n for dataset in datasets:\r\n if min_set != None:\r\n min_set.intersection(set(dataset.keys()))\r\n else:\r\n min_set = set(dataset.keys())\r\n return min_set\r\n\r\n common_splits = _get_common_splits(datasets)\r\n out = {}\r\n for split in common_splits:\r\n out[split] = MultiDataset([d[split] for d in datasets])\r\n return out\r\n\r\n\r\n##########################################\r\n# Dataset Flattening\r\n\r\ndef flatten(dataset,flatten_fn):\r\n for k in dataset.keys():\r\n if isinstance(dataset[k],nlp.Dataset):\r\n dataset[k] = dataset[k].map(flatten_fn,remove_columns=dataset[k].column_names)\r\n\r\n# Squad\r\ndef flatten_squad(example):\r\n return {\"source\": \"squad context: \" + example['context'] + \" question: \" + example['question'],\r\n \"target\":example[\"answers\"][\"text\"]}\r\nsquad = nlp.load_dataset(\"squad\")\r\nflatten(squad,flatten_squad)\r\n\r\n# CNN_DM\r\ndef flatten_cnn_dm(example):\r\n return {\"source\": \"cnn_dm: \" + example['article'],\"target\":[example[\"highlights\"]]}\r\ncnn_dm = nlp.load_dataset(\"cnn_dailymail\", \"3.0.0\")\r\nflatten(cnn_dm,flatten_cnn_dm)\r\n\r\n#############################################\r\n\r\nmtds = load_multitask(squad,cnn_dm)\r\n\r\nfor example in mtds[\"train\"]:\r\n print(example[\"task_name\"],example[\"target\"])\r\n```\r\nLet me know if you have any thoughts. I've started using this in some of my projects and it seems to work. If people are happy with the general approach for a first version, I can make a pull request.", "Hey! Happy to jump into the discussion here. I'm still getting familiar with bits of this code, but the reasons I sampled over data loaders rather than datasets is 1) ensuring that each sampled batch corresponds to only 1 task (in case of different inputs formats/downstream models) and 2) potentially having different batch sizes per task (e.g. some tasks have very long/short inputs). How are you currently dealing with these in your PR?", "The short answer is - I'm not! Everything is currently on a per-example basis. It would be fairly simple to add a `batch_size` argument which would ensure that every `batch_size` examples come from the same task. That should suit most use-cases (unless you wanted to ensure batches all came from the same task and apply something like `SortishSampler` on each task first)\r\n\r\nYour notebook was really inspiring by the way - thanks!", "@zphang is having different batch sizes per task actually helpful? Would be interesting to know as it's not something I've come across as a technique used by any MTL papers.", "mt-dnn's [batcher.py](https://github.com/namisan/mt-dnn/blob/master/mt_dnn/batcher.py) might be worth looking at.", "> @zphang is having different batch sizes per task actually helpful? Would be interesting to know as it's not something I've come across as a technique used by any MTL papers.\r\n\r\nI think having different batch sizes per task is particularly helpful in some scenarios where each task has different amount of data. For example, the problem I'm currently facing is one task has tens of thousands of samples while one task has a couple hundreds. I think in this case different batch size could help. But if using the same batch size is a lot simpler to implement, I guess it makes sense to go with that.", "I think that instead of proportional to size sampling you should specify weights or probabilities for drawing a batch from each dataset. We should also ensure that the smaller datasets are repeated so that the encoder layer doesn't overtrain on the largest dataset.", "Are there any references for people doing different batch sizes per task in the literature? I've only seen constant batch sizes with differing numbers of batches for each task which seems sufficient to prevent the impact of large datasets (Read 3.5.3 of the [T5 paper](https://arxiv.org/pdf/1910.10683.pdf) for example).\r\n\r\n", "Hi,\r\nregarding building T5 dataset , I think we can use datasets https://github.com/huggingface/datasets and then need something similar to tf.data.experimental.sample_from_datasets, do you know if similar functionality exist in pytorch? Which can sample multiple datasets with the given rates. thanks. ", "Is this feature part of a `datasets` release yet? ", "> Here's a fully working implementation based on the `__iter__` function of @zphang.\r\n> \r\n> * I've generated the task choice list in the constructor as it allows us to index into the MultiDataset just like a normal dataset. I'm changing `task_choice_list` into a list of `(dataset_idx, example_idx)` so each entry references a unique dataset example. The shuffling has to be done before this as we don't want to shuffle within each task (we assume this is done by the user if this is what they intend).\r\n> * I'm slightly concerned this list could become very large if many large datasets were used. Can't see a way round it at the moment though.\r\n> * I've used `task.info.builder_name` as the dataset name. Not sure if this is correct.\r\n> * I'd love to add some of the other `Dataset` methods (map, slicing by column, etc...). Would be great to implement the whole interface so a single dataset can be simply replaced by this.\r\n> * This does everything on the individual example-level. If some application required batches all from a single task in turn we can't really do that.\r\n> \r\n> ```python\r\n> import nlp\r\n> import numpy as np\r\n> \r\n> class MultiDataset:\r\n> def __init__(self,tasks):\r\n> self.tasks = tasks\r\n> \r\n> # Create random order of tasks\r\n> # Using size-proportional sampling\r\n> task_choice_list = []\r\n> for i, task in enumerate(self.tasks):\r\n> task_choice_list += [i] * len(task)\r\n> task_choice_list = np.array(task_choice_list)\r\n> np.random.shuffle(task_choice_list)\r\n> \r\n> # Add index into each dataset\r\n> # - We don't want to shuffle within each task\r\n> counters = {}\r\n> self.task_choice_list = []\r\n> for i in range(len(task_choice_list)):\r\n> idx = counters.get(task_choice_list[i],0)\r\n> self.task_choice_list.append((task_choice_list[i],idx))\r\n> counters[task_choice_list[i]] = idx + 1\r\n> \r\n> \r\n> def __len__(self):\r\n> return np.sum([len(t) for t in self.tasks])\r\n> \r\n> def __repr__(self):\r\n> task_str = \", \".join([str(t) for t in self.tasks])\r\n> return f\"MultiDataset(tasks: {task_str})\"\r\n> \r\n> def __getitem__(self,key):\r\n> if isinstance(key, int):\r\n> task_idx, example_idx = self.task_choice_list[key]\r\n> task = self.tasks[task_idx]\r\n> example = task[example_idx]\r\n> example[\"task_name\"] = task.info.builder_name\r\n> return example\r\n> elif isinstance(key, slice):\r\n> raise NotImplementedError()\r\n> \r\n> def __iter__(self):\r\n> for i in range(len(self)):\r\n> yield self[i]\r\n> \r\n> \r\n> def load_multitask(*datasets):\r\n> '''Create multitask datasets per split'''\r\n> \r\n> def _get_common_splits(datasets):\r\n> '''Finds the common splits present in all self.datasets'''\r\n> min_set = None\r\n> for dataset in datasets:\r\n> if min_set != None:\r\n> min_set.intersection(set(dataset.keys()))\r\n> else:\r\n> min_set = set(dataset.keys())\r\n> return min_set\r\n> \r\n> common_splits = _get_common_splits(datasets)\r\n> out = {}\r\n> for split in common_splits:\r\n> out[split] = MultiDataset([d[split] for d in datasets])\r\n> return out\r\n> \r\n> \r\n> ##########################################\r\n> # Dataset Flattening\r\n> \r\n> def flatten(dataset,flatten_fn):\r\n> for k in dataset.keys():\r\n> if isinstance(dataset[k],nlp.Dataset):\r\n> dataset[k] = dataset[k].map(flatten_fn,remove_columns=dataset[k].column_names)\r\n> \r\n> # Squad\r\n> def flatten_squad(example):\r\n> return {\"source\": \"squad context: \" + example['context'] + \" question: \" + example['question'],\r\n> \"target\":example[\"answers\"][\"text\"]}\r\n> squad = nlp.load_dataset(\"squad\")\r\n> flatten(squad,flatten_squad)\r\n> \r\n> # CNN_DM\r\n> def flatten_cnn_dm(example):\r\n> return {\"source\": \"cnn_dm: \" + example['article'],\"target\":[example[\"highlights\"]]}\r\n> cnn_dm = nlp.load_dataset(\"cnn_dailymail\", \"3.0.0\")\r\n> flatten(cnn_dm,flatten_cnn_dm)\r\n> \r\n> #############################################\r\n> \r\n> mtds = load_multitask(squad,cnn_dm)\r\n> \r\n> for example in mtds[\"train\"]:\r\n> print(example[\"task_name\"],example[\"target\"])\r\n> ```\r\n> \r\n> Let me know if you have any thoughts. I've started using this in some of my projects and it seems to work. If people are happy with the general approach for a first version, I can make a pull request.\r\n\r\nNot sure if this is what I'm looking for, but I implemented a version of Examples-Proportional mixing supporting only the basic feature [here](https://stackoverflow.com/a/74070116/10732321), seems to work in my project. ", "You can use `interleave_datasets` to mix several datasets together. By default it alternates between all the datasets, but you can also provide sampling probabilities if you want to oversample from one of the datasets\r\n\r\n```python\r\nfrom datasets import load_dataset, interleave_datasets\r\n\r\nsquad = load_dataset(\"squad\", split=\"train\")\r\ncnn_dm = load_dataset(\"cnn_dailymail\", \"3.0.0\", split=\"train\")\r\nds = interleave_datasets([squad, cnn_dm])\r\n\r\nprint(ds[0])\r\n# {'id': '5733be284776f41900661182',\r\n# 'title': 'University_of_Notre_Dame',\r\n# 'context': 'Architecturally, the school has a Catholic character...',\r\n# 'question': 'To whom did the Virgin Mary allegedly appear in 1858 in Lourdes France?',\r\n# 'answers': {'text': ['Saint Bernadette Soubirous'], 'answer_start': [515]},\r\n# 'article': None,\r\n# 'highlights': None}\r\nprint(ds[1])\r\n# {'id': '42c027e4ff9730fbb3de84c1af0d2c506e41c3e4',\r\n# 'title': None,\r\n# 'context': None,\r\n# 'question': None,\r\n# 'answers': None,\r\n# 'article': 'LONDON, England (Reuters) -- Harry Potter star Daniel Radcliffe...',\r\n# 'highlights': \"Harry Potter star Daniel Radcliffe...\"}\r\n```\r\n\r\nsee docs at https://huggingface.co/docs/datasets/v2.6.1/en/package_reference/main_classes#datasets.interleave_datasets", "I also have this implementation of multi-task sampler here which I used it to tune T5: https://github.com/rabeehk/hyperformer/blob/main/hyperformer/data/multitask_sampler.py " ]

[]

[ "I `git rebase`ed my branch to `upstream/master` as suggested in point 7 of <https://huggingface.co/docs/datasets/share_dataset.html> and subsequently used `git pull` to be able to push to my remote branch. However, I think this messed up the history.\r\n\r\nPlease let me know if I should create a clean new PR with my changes.\r\n\r\nUpdate: I also fixed the dataset name in the Dataset Card.", "Dear @SBrandeis , @lhoestq . I am not sure how to fix the PR with respect to the additional files that are currently included in the commits. Could you provide me with an example? Otherwise I would be happy to close/re-open another PR. Please let me know if anything is missing for the review.", "Hi @tdiggelm, thanks for the contribution! This dataset is really awesome.\r\nI believe creating a new branch from master and opening a new PR with your changes is the simplest option since no review has been done yet. Feel free to ping us when it's done.", "> Hi @tdiggelm, thanks for the contribution! This dataset is really awesome.\r\n> I believe creating a new branch from master and opening a new PR with your changes is the simplest option since no review has been done yet. Feel free to ping us when it's done.\r\n\r\nThank you very much for your quick reply! Will do ASAP and ping you when done.", "closing in favor of #1623" ]

[ { "color": "a2eeef", "default": true, "description": "New feature or request", "id": 1935892871, "name": "enhancement", "node_id": "MDU6TGFiZWwxOTM1ODkyODcx", "url": "https://api.github.com/repos/huggingface/datasets/labels/enhancement" } ]

[ "Hi @stas00, thanks for pointing out this issue with logging.\r\n\r\nI agree that `datasets` can sometimes be too verbose... I can create a PR and we could discuss there the choice of the log levels for different parts of the code." ]

[ { "color": "d73a4a", "default": true, "description": "Something isn't working", "id": 1935892857, "name": "bug", "node_id": "MDU6TGFiZWwxOTM1ODkyODU3", "url": "https://api.github.com/repos/huggingface/datasets/labels/bug" } ]

[ "knock knock", "Hi, thanks for reporting! I've linked a PR that should fix the issue.", "I've checked the PR and it looks great, thanks a lot!" ]

[]

[ "Hi ! In the arrow file we store all the integers as uint8.\r\nSo your arrow file should weigh around `height x width x n_channels x n_images` bytes.\r\n\r\nWhat feature type do your TFDS dataset have ?\r\n\r\nIf it uses a `tfds.features.Image` type, then what is stored is the encoded data (as png or jpg for example). Since these encodings are made for compression, the resulting tfrecord is smaller that the arrow file.\r\n\r\nWe are working on adding a similar feature in `datasets`: the ability to store the encoded data instead of the raw integers for images, but also for audio data. This way, arrow files will have similar sizes as tfrecords for images.", "Thanks for the prompt response. You're right about the encoding, I have the `tfds.features.Image` feature type you mentioned.\r\nHowever, as described in the `dataset_info.json`, my dataset is made of 1479 (224x224x3) images. 1479 x 224 x 224 x 3 = 222630912 bytes which is far from the actual size 520803408 bytes. \r\n\r\nAnyway I look forward to the Image feature type in `datasets`. ", "@lhoestq I changed the data structure so I have a 2D Array feature type instead of a 3D Array by grouping the two last dimensions ( a 224x672 2D Array instead of a 224x224x3 3D Array). The file size is now 223973964 bytes, nearly half the previous size! Which is around of what I would expect.\r\nI found similar behavior in existing `datasets` collection, when comparing black and white vs color image, for example MNIST vs CIFAR. ", "Interesting !\r\nThis may be because of the offsets that are stored with the array data.\r\n\r\nCurrently the offsets are stored even if the `shape` of the arrays is fixed. This was needed because of some issues with pyarrow a few months ago. I think these issues have been addressed now, so we can probably try to remove them to make the file lighter.\r\n\r\nIdeally in your case the floats data should be 220 MB for both Array2D and Array3D", "Yeah for sure, can you be a bit more specific about where the offset is stored in the code base ? And any reference to pyarrow issues if you have some. I would be very interested in contributing to `datasets` by trying to fix this issue. ", "Pyarrow has two types of lists: variable length lists and fixed size lists.\r\nCurrently we store the ArrayXD data as variable length lists. They take more disk space because they must store both actual data and offsets.\r\nIn the `datasets` code this is done here:\r\n\r\nhttps://github.com/huggingface/nlp/blob/dbac87c8a083f806467f5afc4ec9b401a7e4c15c/src/datasets/features.py#L346-L352\r\n\r\nTo use a fixed length list, one should use the `list_size` argument of `pyarrow.list_()`.\r\nI believe this would work directly modulo some changes in the numpy conversion here:\r\n\r\nhttps://github.com/huggingface/nlp/blob/dbac87c8a083f806467f5afc4ec9b401a7e4c15c/src/datasets/features.py#L381-L395" ]

[]

[ "I have the same issue. Tried to download a few of them and not a single one is downloaded successfully.\r\n\r\nThis is the output:\r\n```\r\n>>> dataset = load_dataset('blended_skill_talk', split='train')\r\nUsing custom data configuration default <-- This step never ends\r\n```", "This was fixed in #644 \r\nI'll do a new release soon :)\r\n\r\nIn the meantime you can run it by installing from source", "Closing since version 1.1.0 got released with Windows support :) \r\nLet me know if it works for you now" ]

[ { "color": "d73a4a", "default": true, "description": "Something isn't working", "id": 1935892857, "name": "bug", "node_id": "MDU6TGFiZWwxOTM1ODkyODU3", "url": "https://api.github.com/repos/huggingface/datasets/labels/bug" } ]

[ "Hi @lewtun, thanks for reporting.\r\n\r\nIt seems that our library fails at inferring the dtype of the columns:\r\n- `milestone`\r\n- `performed_via_github_app` \r\n\r\n(and assigns them `null` dtype)." ]

[]

[]

[]

[ "_The documentation is not available anymore as the PR was closed or merged._", "<details>\n<summary>Show benchmarks</summary>\n\nPyArrow==6.0.0\n\n<details>\n<summary>Show updated benchmarks!</summary>\n\n### Benchmark: benchmark_array_xd.json\n\n| metric | read_batch_formatted_as_numpy after write_array2d | read_batch_formatted_as_numpy after write_flattened_sequence | read_batch_formatted_as_numpy after write_nested_sequence | read_batch_unformated after write_array2d | read_batch_unformated after write_flattened_sequence | read_batch_unformated after write_nested_sequence | read_col_formatted_as_numpy after write_array2d | read_col_formatted_as_numpy after write_flattened_sequence | read_col_formatted_as_numpy after write_nested_sequence | read_col_unformated after write_array2d | read_col_unformated after write_flattened_sequence | read_col_unformated after write_nested_sequence | read_formatted_as_numpy after write_array2d | read_formatted_as_numpy after write_flattened_sequence | read_formatted_as_numpy after write_nested_sequence | read_unformated after write_array2d | read_unformated after write_flattened_sequence | read_unformated after write_nested_sequence | write_array2d | write_flattened_sequence | write_nested_sequence |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.008753 / 0.011353 (-0.002600) | 0.004877 / 0.011008 (-0.006131) | 0.098320 / 0.038508 (0.059812) | 0.034123 / 0.023109 (0.011014) | 0.289539 / 0.275898 (0.013641) | 0.323584 / 0.323480 (0.000104) | 0.007455 / 0.007986 (-0.000531) | 0.004763 / 0.004328 (0.000434) | 0.074350 / 0.004250 (0.070100) | 0.039018 / 0.037052 (0.001966) | 0.294319 / 0.258489 (0.035830) | 0.348686 / 0.293841 (0.054845) | 0.037814 / 0.128546 (-0.090732) | 0.011808 / 0.075646 (-0.063838) | 0.333808 / 0.419271 (-0.085464) | 0.047758 / 0.043533 (0.004225) | 0.298533 / 0.255139 (0.043394) | 0.320790 / 0.283200 (0.037590) | 0.095909 / 0.141683 (-0.045774) | 1.434422 / 1.452155 (-0.017732) | 1.509703 / 1.492716 (0.016987) |\n\n### Benchmark: benchmark_getitem\\_100B.json\n\n| metric | get_batch_of\\_1024\\_random_rows | get_batch_of\\_1024\\_rows | get_first_row | get_last_row |\n|--------|---|---|---|---|\n| new / old (diff) | 0.201728 / 0.018006 (0.183722) | 0.432243 / 0.000490 (0.431753) | 0.002760 / 0.000200 (0.002560) | 0.000080 / 0.000054 (0.000026) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.026090 / 0.037411 (-0.011321) | 0.105914 / 0.014526 (0.091388) | 0.115869 / 0.176557 (-0.060688) | 0.178291 / 0.737135 (-0.558844) | 0.121435 / 0.296338 (-0.174904) |\n\n### Benchmark: benchmark_iterating.json\n\n| metric | read 5000 | read 50000 | read_batch 50000 10 | read_batch 50000 100 | read_batch 50000 1000 | read_formatted numpy 5000 | read_formatted pandas 5000 | read_formatted tensorflow 5000 | read_formatted torch 5000 | read_formatted_batch numpy 5000 10 | read_formatted_batch numpy 5000 1000 | shuffled read 5000 | shuffled read 50000 | shuffled read_batch 50000 10 | shuffled read_batch 50000 100 | shuffled read_batch 50000 1000 | shuffled read_formatted numpy 5000 | shuffled read_formatted_batch numpy 5000 10 | shuffled read_formatted_batch numpy 5000 1000 |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.402304 / 0.215209 (0.187095) | 3.995183 / 2.077655 (1.917529) | 1.794548 / 1.504120 (0.290428) | 1.603034 / 1.541195 (0.061839) | 1.643836 / 1.468490 (0.175346) | 0.694934 / 4.584777 (-3.889843) | 3.695128 / 3.745712 (-0.050584) | 2.018582 / 5.269862 (-3.251279) | 1.294315 / 4.565676 (-3.271362) | 0.085346 / 0.424275 (-0.338929) | 0.012201 / 0.007607 (0.004594) | 0.510057 / 0.226044 (0.284012) | 5.123404 / 2.268929 (2.854476) | 2.319089 / 55.444624 (-53.125535) | 1.930935 / 6.876477 (-4.945542) | 1.939700 / 2.142072 (-0.202372) | 0.848282 / 4.805227 (-3.956945) | 0.165561 / 6.500664 (-6.335103) | 0.062028 / 0.075469 (-0.013441) |\n\n### Benchmark: benchmark_map_filter.json\n\n| metric | filter | map fast-tokenizer batched | map identity | map identity batched | map no-op batched | map no-op batched numpy | map no-op batched pandas | map no-op batched pytorch | map no-op batched tensorflow |\n|--------|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 1.220576 / 1.841788 (-0.621212) | 14.413853 / 8.074308 (6.339544) | 14.027156 / 10.191392 (3.835764) | 0.170109 / 0.680424 (-0.510315) | 0.029412 / 0.534201 (-0.504789) | 0.443898 / 0.579283 (-0.135386) | 0.433059 / 0.434364 (-0.001305) | 0.533465 / 0.540337 (-0.006872) | 0.626562 / 1.386936 (-0.760374) |\n\n</details>\nPyArrow==latest\n\n<details>\n<summary>Show updated benchmarks!</summary>\n\n### Benchmark: benchmark_array_xd.json\n\n| metric | read_batch_formatted_as_numpy after write_array2d | read_batch_formatted_as_numpy after write_flattened_sequence | read_batch_formatted_as_numpy after write_nested_sequence | read_batch_unformated after write_array2d | read_batch_unformated after write_flattened_sequence | read_batch_unformated after write_nested_sequence | read_col_formatted_as_numpy after write_array2d | read_col_formatted_as_numpy after write_flattened_sequence | read_col_formatted_as_numpy after write_nested_sequence | read_col_unformated after write_array2d | read_col_unformated after write_flattened_sequence | read_col_unformated after write_nested_sequence | read_formatted_as_numpy after write_array2d | read_formatted_as_numpy after write_flattened_sequence | read_formatted_as_numpy after write_nested_sequence | read_unformated after write_array2d | read_unformated after write_flattened_sequence | read_unformated after write_nested_sequence | write_array2d | write_flattened_sequence | write_nested_sequence |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.007148 / 0.011353 (-0.004205) | 0.005019 / 0.011008 (-0.005989) | 0.073132 / 0.038508 (0.034624) | 0.032763 / 0.023109 (0.009654) | 0.329309 / 0.275898 (0.053411) | 0.361658 / 0.323480 (0.038178) | 0.005683 / 0.007986 (-0.002302) | 0.003793 / 0.004328 (-0.000536) | 0.071858 / 0.004250 (0.067608) | 0.045160 / 0.037052 (0.008107) | 0.335852 / 0.258489 (0.077363) | 0.384274 / 0.293841 (0.090433) | 0.036647 / 0.128546 (-0.091899) | 0.012217 / 0.075646 (-0.063430) | 0.086265 / 0.419271 (-0.333007) | 0.049223 / 0.043533 (0.005690) | 0.331460 / 0.255139 (0.076321) | 0.353175 / 0.283200 (0.069975) | 0.102214 / 0.141683 (-0.039469) | 1.491451 / 1.452155 (0.039296) | 1.553702 / 1.492716 (0.060985) |\n\n### Benchmark: benchmark_getitem\\_100B.json\n\n| metric | get_batch_of\\_1024\\_random_rows | get_batch_of\\_1024\\_rows | get_first_row | get_last_row |\n|--------|---|---|---|---|\n| new / old (diff) | 0.222972 / 0.018006 (0.204966) | 0.432862 / 0.000490 (0.432372) | 0.000421 / 0.000200 (0.000221) | 0.000059 / 0.000054 (0.000005) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.028401 / 0.037411 (-0.009010) | 0.109331 / 0.014526 (0.094805) | 0.119246 / 0.176557 (-0.057311) | 0.187997 / 0.737135 (-0.549138) | 0.124212 / 0.296338 (-0.172127) |\n\n### Benchmark: benchmark_iterating.json\n\n| metric | read 5000 | read 50000 | read_batch 50000 10 | read_batch 50000 100 | read_batch 50000 1000 | read_formatted numpy 5000 | read_formatted pandas 5000 | read_formatted tensorflow 5000 | read_formatted torch 5000 | read_formatted_batch numpy 5000 10 | read_formatted_batch numpy 5000 1000 | shuffled read 5000 | shuffled read 50000 | shuffled read_batch 50000 10 | shuffled read_batch 50000 100 | shuffled read_batch 50000 1000 | shuffled read_formatted numpy 5000 | shuffled read_formatted_batch numpy 5000 10 | shuffled read_formatted_batch numpy 5000 1000 |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.427240 / 0.215209 (0.212031) | 4.271619 / 2.077655 (2.193964) | 2.104948 / 1.504120 (0.600828) | 1.910624 / 1.541195 (0.369430) | 1.943812 / 1.468490 (0.475322) | 0.711466 / 4.584777 (-3.873311) | 3.778987 / 3.745712 (0.033275) | 2.976258 / 5.269862 (-2.293604) | 1.807591 / 4.565676 (-2.758086) | 0.088286 / 0.424275 (-0.335989) | 0.012461 / 0.007607 (0.004854) | 0.527554 / 0.226044 (0.301509) | 5.279461 / 2.268929 (3.010532) | 2.517911 / 55.444624 (-52.926713) | 2.176557 / 6.876477 (-4.699920) | 2.205322 / 2.142072 (0.063249) | 0.855012 / 4.805227 (-3.950215) | 0.170007 / 6.500664 (-6.330658) | 0.065273 / 0.075469 (-0.010196) |\n\n### Benchmark: benchmark_map_filter.json\n\n| metric | filter | map fast-tokenizer batched | map identity | map identity batched | map no-op batched | map no-op batched numpy | map no-op batched pandas | map no-op batched pytorch | map no-op batched tensorflow |\n|--------|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 1.282785 / 1.841788 (-0.559003) | 14.819500 / 8.074308 (6.745192) | 13.282211 / 10.191392 (3.090819) | 0.161804 / 0.680424 (-0.518620) | 0.017615 / 0.534201 (-0.516586) | 0.420159 / 0.579283 (-0.159124) | 0.441304 / 0.434364 (0.006940) | 0.531704 / 0.540337 (-0.008634) | 0.627477 / 1.386936 (-0.759459) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#b91070b9c09673e2e148eec458036ab6a62ac042 \"CML watermark\")\n", "Hmm I think we need to add more tests. Not sure what would happen with :\r\n- decodable features that may end up decoded twice \r\n- formatted datasets \r\n\r\nI'd be in favor of reverting this until we checked those" ]

[]

[]

[]

[]

[]

[ "@lhoestq - tiny change in the dummy data command, should be good to merge." ]

[ { "color": "d73a4a", "default": true, "description": "Something isn't working", "id": 1935892857, "name": "bug", "node_id": "MDU6TGFiZWwxOTM1ODkyODU3", "url": "https://api.github.com/repos/huggingface/datasets/labels/bug" } ]

[ "Hi ! Thanks for reporting. Indeed this is a current limitation of the usage we have of `dill` in `datasets`. I'd suggest you use your workaround for now until we find a way to fix this. Maybe functions that are not coming from a module not installed with pip should be dumped completely, rather than only taking their locations into account", "I agree. Sounds like a solution for it would be pretty dirty, even [cloudpickle](https://stackoverflow.com/a/16891169) doesn't help in this case.\r\nIn the meanwhile I think that adding a warning and the workaround somewhere in the documentation can be helpful.", "For anyone interested, I see that with `dill==0.3.6` the workaround I suggested doesn't work anymore.\r\nI opened an issue about it: https://github.com/uqfoundation/dill/issues/572.\r\n\r\n " ]

[ { "color": "a2eeef", "default": true, "description": "New feature or request", "id": 1935892871, "name": "enhancement", "node_id": "MDU6TGFiZWwxOTM1ODkyODcx", "url": "https://api.github.com/repos/huggingface/datasets/labels/enhancement" } ]

[ "duplicate of #2655" ]

[]

[ "_The documentation is not available anymore as the PR was closed or merged._" ]

[]

[ "> Should I add a section in https://github.com/huggingface/datasets/blob/master/docs/source/load_hub.rst? (there is no section for get_dataset_infos)\r\n\r\nYes totally :) This tutorial should indeed mention this, given how fundamental it is" ]

[]

[]

[ { "color": "d73a4a", "default": true, "description": "Something isn't working", "id": 1935892857, "name": "bug", "node_id": "MDU6TGFiZWwxOTM1ODkyODU3", "url": "https://api.github.com/repos/huggingface/datasets/labels/bug" } ]

[ "Hi @liyucheng09, thanks for reporting.\r\n\r\nApparently this issue has to do with your environment setup. One question: is your data in an NFS share? Some people have reported this error when using `fcntl` to write to an NFS share... If this is the case, then it might be that your NFS just may not be set up to provide file locks. You should ask your system administrator, or try these commands in the terminal:\r\n```shell\r\nsudo systemctl enable rpc-statd\r\nsudo systemctl start rpc-statd\r\n```" ]

[]

[ "@lhoestq ready for a second review" ]

[]

[]

[]

[]

[ { "color": "d73a4a", "default": true, "description": "Something isn't working", "id": 1935892857, "name": "bug", "node_id": "MDU6TGFiZWwxOTM1ODkyODU3", "url": "https://api.github.com/repos/huggingface/datasets/labels/bug" } ]

[ "Hi ! I guess the caching mechanism should have considered the new `filter` to be different from the old one, and don't use cached results from the old `filter`.\r\nTo avoid other users from having this issue we could make the caching differentiate the two, what do you think ?", "If it's easy enough to implement, then yes please 😄 But this issue can be low-priority, since I've only encountered it in a couple of `transformers` CI tests.", "Well it can cause issue with anyone that updates `datasets` and re-run some code that uses filter, so I'm creating a PR", "I just merged a fix, let me know if you're still having this kind of issues :)\r\n\r\nWe'll do a release soon to make this fix available", "Definitely works on several manual cases with our dummy datasets, thank you @lhoestq !", "Fixed by #2947." ]

[ { "color": "a2eeef", "default": true, "description": "New feature or request", "id": 1935892871, "name": "enhancement", "node_id": "MDU6TGFiZWwxOTM1ODkyODcx", "url": "https://api.github.com/repos/huggingface/datasets/labels/enhancement" } ]

[ "#2274 has been merged. You can now merge master into this branch and use `assert_arrow_metadata_are_synced_with_dataset_features(dset)` to make sure that the metadata are good :)" ]

[]

[]

[ { "color": "2edb81", "default": false, "description": "A bug in a dataset script provided in the library", "id": 2067388877, "name": "dataset bug", "node_id": "MDU6TGFiZWwyMDY3Mzg4ODc3", "url": "https://api.github.com/repos/huggingface/datasets/labels/dataset%20bug" } ]

[ "Hi ! It runs on my side without issues. I tried\r\n```python\r\nfrom datasets import load_dataset\r\nload_dataset(\"boolq\")\r\n```\r\n\r\nWhat version of datasets and tensorflow are your runnning ?\r\nAlso if you manage to get a minimal reproducible script (on google colab for example) that would be useful.", "hey\ni do the exact same commands. for me it fails i guess might be issues with\ncaching maybe?\nthanks\nbest\nrabeeh\n\nOn Tue, Nov 24, 2020, 10:24 AM Quentin Lhoest <[email protected]>\nwrote:\n\n> Hi ! It runs on my side without issues. I tried\n>\n> from datasets import load_datasetload_dataset(\"boolq\")\n>\n> What version of datasets and tensorflow are your runnning ?\n> Also if you manage to get a minimal reproducible script (on google colab\n> for example) that would be useful.\n>\n> —\n> You are receiving this because you authored the thread.\n> Reply to this email directly, view it on GitHub\n> <https://github.com/huggingface/datasets/issues/879#issuecomment-732769114>,\n> or unsubscribe\n> <https://github.com/notifications/unsubscribe-auth/ABP4ZCGGDR2FUMRKZTIY5CTSRN3VXANCNFSM4T7R3U6A>\n> .\n>\n", "Could you check if it works on the master branch ?\r\nYou can use `load_dataset(\"boolq\", script_version=\"master\")` to do so.\r\nWe did some changes recently in boolq to remove the TF dependency and we changed the way the data files are downloaded in https://github.com/huggingface/datasets/pull/881" ]

[]

[ "_The documentation is not available anymore as the PR was closed or merged._" ]

[]

[ "_The documentation is not available anymore as the PR was closed or merged._", "> This makes me think we shouldn't advise the use of load_dataset in dataset scripts, since it doesn't guarantee that the cache will work as expected (the cache directory is not set correctly, and the required disk space for downloaded files is not recorded)\r\n\r\n@lhoestq, do you think it could be a good idea to add a comment in this script WARNING that using load_dataset in a script is not good practice and that people should avoid using that script as a template to create other scripts? ", "good idea ! :)" ]

[ { "color": "d73a4a", "default": true, "description": "Something isn't working", "id": 1935892857, "name": "bug", "node_id": "MDU6TGFiZWwxOTM1ODkyODU3", "url": "https://api.github.com/repos/huggingface/datasets/labels/bug" } ]

[ "I think the test passes because it simply ignore what's after `gzip://`.\r\n\r\nThe returned urlpath is expected to look like `gzip://filename::url`, and the filename is currently considered to be what's after the final `/`, hence the result.\r\n\r\nWe can decide to change this and simply have `gzip://::url`, this way we don't need to guess the filename, what do you think ?" ]

[]

[ "_The documentation is not available anymore as the PR was closed or merged._", "Comment by @lhoestq:\r\n> Do you think we need this in `datasets` if it's fixed on the moon landing side ? In the aiohttp doc they consider those symbols as \"non-safe\" ", "The lib `requests` does not perform that requote on redirect URLs.", "Indeed, the `requests` library does perform a requoting, but this does not unquote `%27`:\r\n```python\r\nIn [1]: from requests.utils import requote_uri\r\n\r\nIn [2]: url = \"https://netloc/path?param=param%27%27value\"\r\n\r\nIn [3]: url\r\nOut[3]: 'https://netloc/path?param=param%27%27value'\r\n\r\nIn [4]: requote_uri(url)\r\nOut[4]: 'https://netloc/path?param=param%27%27value'\r\n```\r\n\r\nHowever, the `aiohttp` library uses `yarl.ULR` and this does unquote `%27`:\r\n```python\r\nIn [5]: from yarl import URL\r\n\r\nIn [6]: url\r\nOut[6]: 'https://netloc/path?param=param%27%27value'\r\n\r\nIn [7]: str(URL(url))\r\nOut[7]: \"https://netloc/path?param=param''value\"\r\n```\r\n\r\nIf we pass `requote_redirect_url=False` to `aiohttp`, then it passes `encoded=True` to `yarl.ULR`: https://github.com/aio-libs/aiohttp/blob/4635161ee8e7ad321cca46e01ce5bfeb1ad8bf26/aiohttp/client.py#L578-L580\r\n```python\r\nparsed_url = URL(\r\n r_url, encoded=not self._requote_redirect_url\r\n)\r\n```\r\nwhich does not unquote `%27`:\r\n```python\r\nIn [8]: url\r\nOut[8]: 'https://netloc/path?param=param%27%27value'\r\n\r\nIn [9]: str(URL(url, encoded=True))\r\nOut[9]: 'https://netloc/path?param=param%27%27value'\r\n```", "See the issues we opened in the respective libraries:\r\n- aiohttp\r\n - aio-libs/aiohttp#7183\r\n- requests\r\n - psf/requests#6341", "<details>\n<summary>Show benchmarks</summary>\n\nPyArrow==6.0.0\n\n<details>\n<summary>Show updated benchmarks!</summary>\n\n### Benchmark: benchmark_array_xd.json\n\n| metric | read_batch_formatted_as_numpy after write_array2d | read_batch_formatted_as_numpy after write_flattened_sequence | read_batch_formatted_as_numpy after write_nested_sequence | read_batch_unformated after write_array2d | read_batch_unformated after write_flattened_sequence | read_batch_unformated after write_nested_sequence | read_col_formatted_as_numpy after write_array2d | read_col_formatted_as_numpy after write_flattened_sequence | read_col_formatted_as_numpy after write_nested_sequence | read_col_unformated after write_array2d | read_col_unformated after write_flattened_sequence | read_col_unformated after write_nested_sequence | read_formatted_as_numpy after write_array2d | read_formatted_as_numpy after write_flattened_sequence | read_formatted_as_numpy after write_nested_sequence | read_unformated after write_array2d | read_unformated after write_flattened_sequence | read_unformated after write_nested_sequence | write_array2d | write_flattened_sequence | write_nested_sequence |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.012399 / 0.011353 (0.001047) | 0.006388 / 0.011008 (-0.004620) | 0.134173 / 0.038508 (0.095665) | 0.037059 / 0.023109 (0.013949) | 0.420697 / 0.275898 (0.144799) | 0.473981 / 0.323480 (0.150502) | 0.009857 / 0.007986 (0.001871) | 0.004791 / 0.004328 (0.000463) | 0.106886 / 0.004250 (0.102636) | 0.044871 / 0.037052 (0.007818) | 0.429843 / 0.258489 (0.171354) | 0.461569 / 0.293841 (0.167728) | 0.057285 / 0.128546 (-0.071261) | 0.018809 / 0.075646 (-0.056837) | 0.432613 / 0.419271 (0.013342) | 0.058086 / 0.043533 (0.014553) | 0.413064 / 0.255139 (0.157925) | 0.444407 / 0.283200 (0.161207) | 0.119102 / 0.141683 (-0.022581) | 1.875954 / 1.452155 (0.423799) | 1.916392 / 1.492716 (0.423676) |\n\n### Benchmark: benchmark_getitem\\_100B.json\n\n| metric | get_batch_of\\_1024\\_random_rows | get_batch_of\\_1024\\_rows | get_first_row | get_last_row |\n|--------|---|---|---|---|\n| new / old (diff) | 0.267489 / 0.018006 (0.249483) | 0.567554 / 0.000490 (0.567064) | 0.005901 / 0.000200 (0.005701) | 0.000134 / 0.000054 (0.000079) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.031248 / 0.037411 (-0.006164) | 0.123014 / 0.014526 (0.108489) | 0.140001 / 0.176557 (-0.036556) | 0.191476 / 0.737135 (-0.545659) | 0.141687 / 0.296338 (-0.154652) |\n\n### Benchmark: benchmark_iterating.json\n\n| metric | read 5000 | read 50000 | read_batch 50000 10 | read_batch 50000 100 | read_batch 50000 1000 | read_formatted numpy 5000 | read_formatted pandas 5000 | read_formatted tensorflow 5000 | read_formatted torch 5000 | read_formatted_batch numpy 5000 10 | read_formatted_batch numpy 5000 1000 | shuffled read 5000 | shuffled read 50000 | shuffled read_batch 50000 10 | shuffled read_batch 50000 100 | shuffled read_batch 50000 1000 | shuffled read_formatted numpy 5000 | shuffled read_formatted_batch numpy 5000 10 | shuffled read_formatted_batch numpy 5000 1000 |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.637481 / 0.215209 (0.422272) | 6.255969 / 2.077655 (4.178314) | 2.559811 / 1.504120 (1.055691) | 2.118154 / 1.541195 (0.576960) | 2.079487 / 1.468490 (0.610997) | 1.201079 / 4.584777 (-3.383698) | 5.592625 / 3.745712 (1.846913) | 5.143344 / 5.269862 (-0.126517) | 2.764716 / 4.565676 (-1.800960) | 0.142539 / 0.424275 (-0.281736) | 0.015541 / 0.007607 (0.007934) | 0.771407 / 0.226044 (0.545363) | 7.631657 / 2.268929 (5.362728) | 3.279684 / 55.444624 (-52.164940) | 2.587566 / 6.876477 (-4.288911) | 2.624622 / 2.142072 (0.482549) | 1.427878 / 4.805227 (-3.377350) | 0.257759 / 6.500664 (-6.242906) | 0.078616 / 0.075469 (0.003147) |\n\n### Benchmark: benchmark_map_filter.json\n\n| metric | filter | map fast-tokenizer batched | map identity | map identity batched | map no-op batched | map no-op batched numpy | map no-op batched pandas | map no-op batched pytorch | map no-op batched tensorflow |\n|--------|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 1.609305 / 1.841788 (-0.232483) | 18.258792 / 8.074308 (10.184484) | 20.345242 / 10.191392 (10.153850) | 0.267366 / 0.680424 (-0.413058) | 0.047035 / 0.534201 (-0.487166) | 0.568881 / 0.579283 (-0.010402) | 0.662763 / 0.434364 (0.228399) | 0.668927 / 0.540337 (0.128590) | 0.755766 / 1.386936 (-0.631170) |\n\n</details>\nPyArrow==latest\n\n<details>\n<summary>Show updated benchmarks!</summary>\n\n### Benchmark: benchmark_array_xd.json\n\n| metric | read_batch_formatted_as_numpy after write_array2d | read_batch_formatted_as_numpy after write_flattened_sequence | read_batch_formatted_as_numpy after write_nested_sequence | read_batch_unformated after write_array2d | read_batch_unformated after write_flattened_sequence | read_batch_unformated after write_nested_sequence | read_col_formatted_as_numpy after write_array2d | read_col_formatted_as_numpy after write_flattened_sequence | read_col_formatted_as_numpy after write_nested_sequence | read_col_unformated after write_array2d | read_col_unformated after write_flattened_sequence | read_col_unformated after write_nested_sequence | read_formatted_as_numpy after write_array2d | read_formatted_as_numpy after write_flattened_sequence | read_formatted_as_numpy after write_nested_sequence | read_unformated after write_array2d | read_unformated after write_flattened_sequence | read_unformated after write_nested_sequence | write_array2d | write_flattened_sequence | write_nested_sequence |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.010017 / 0.011353 (-0.001336) | 0.006816 / 0.011008 (-0.004192) | 0.105038 / 0.038508 (0.066529) | 0.038689 / 0.023109 (0.015580) | 0.482113 / 0.275898 (0.206215) | 0.540072 / 0.323480 (0.216592) | 0.007738 / 0.007986 (-0.000248) | 0.005134 / 0.004328 (0.000806) | 0.102203 / 0.004250 (0.097953) | 0.054080 / 0.037052 (0.017028) | 0.501057 / 0.258489 (0.242568) | 0.567186 / 0.293841 (0.273345) | 0.060330 / 0.128546 (-0.068217) | 0.020059 / 0.075646 (-0.055587) | 0.123102 / 0.419271 (-0.296170) | 0.063426 / 0.043533 (0.019893) | 0.494171 / 0.255139 (0.239032) | 0.538238 / 0.283200 (0.255039) | 0.119613 / 0.141683 (-0.022069) | 1.853728 / 1.452155 (0.401574) | 1.984621 / 1.492716 (0.491904) |\n\n### Benchmark: benchmark_getitem\\_100B.json\n\n| metric | get_batch_of\\_1024\\_random_rows | get_batch_of\\_1024\\_rows | get_first_row | get_last_row |\n|--------|---|---|---|---|\n| new / old (diff) | 0.282511 / 0.018006 (0.264505) | 0.563190 / 0.000490 (0.562700) | 0.000465 / 0.000200 (0.000265) | 0.000086 / 0.000054 (0.000032) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.029267 / 0.037411 (-0.008144) | 0.135618 / 0.014526 (0.121093) | 0.146286 / 0.176557 (-0.030271) | 0.188570 / 0.737135 (-0.548565) | 0.155839 / 0.296338 (-0.140499) |\n\n### Benchmark: benchmark_iterating.json\n\n| metric | read 5000 | read 50000 | read_batch 50000 10 | read_batch 50000 100 | read_batch 50000 1000 | read_formatted numpy 5000 | read_formatted pandas 5000 | read_formatted tensorflow 5000 | read_formatted torch 5000 | read_formatted_batch numpy 5000 10 | read_formatted_batch numpy 5000 1000 | shuffled read 5000 | shuffled read 50000 | shuffled read_batch 50000 10 | shuffled read_batch 50000 100 | shuffled read_batch 50000 1000 | shuffled read_formatted numpy 5000 | shuffled read_formatted_batch numpy 5000 10 | shuffled read_formatted_batch numpy 5000 1000 |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.671660 / 0.215209 (0.456451) | 6.718775 / 2.077655 (4.641120) | 3.004601 / 1.504120 (1.500481) | 2.640504 / 1.541195 (1.099309) | 2.666788 / 1.468490 (1.198298) | 1.242655 / 4.584777 (-3.342122) | 5.780119 / 3.745712 (2.034407) | 3.247935 / 5.269862 (-2.021927) | 2.114007 / 4.565676 (-2.451669) | 0.147546 / 0.424275 (-0.276729) | 0.014408 / 0.007607 (0.006801) | 0.824407 / 0.226044 (0.598362) | 8.278185 / 2.268929 (6.009257) | 3.733463 / 55.444624 (-51.711161) | 2.976732 / 6.876477 (-3.899745) | 3.132758 / 2.142072 (0.990686) | 1.446095 / 4.805227 (-3.359132) | 0.258628 / 6.500664 (-6.242036) | 0.085513 / 0.075469 (0.010043) |\n\n### Benchmark: benchmark_map_filter.json\n\n| metric | filter | map fast-tokenizer batched | map identity | map identity batched | map no-op batched | map no-op batched numpy | map no-op batched pandas | map no-op batched pytorch | map no-op batched tensorflow |\n|--------|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 1.702681 / 1.841788 (-0.139106) | 18.725123 / 8.074308 (10.650815) | 19.622808 / 10.191392 (9.431416) | 0.215845 / 0.680424 (-0.464579) | 0.029246 / 0.534201 (-0.504955) | 0.554819 / 0.579283 (-0.024464) | 0.630926 / 0.434364 (0.196562) | 0.637663 / 0.540337 (0.097325) | 0.837948 / 1.386936 (-0.548988) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#c4a4f96ef0a4ec4b25f0872f160fa1eb9d2e711c \"CML watermark\")\n", "<details>\n<summary>Show benchmarks</summary>\n\nPyArrow==6.0.0\n\n<details>\n<summary>Show updated benchmarks!</summary>\n\n### Benchmark: benchmark_array_xd.json\n\n| metric | read_batch_formatted_as_numpy after write_array2d | read_batch_formatted_as_numpy after write_flattened_sequence | read_batch_formatted_as_numpy after write_nested_sequence | read_batch_unformated after write_array2d | read_batch_unformated after write_flattened_sequence | read_batch_unformated after write_nested_sequence | read_col_formatted_as_numpy after write_array2d | read_col_formatted_as_numpy after write_flattened_sequence | read_col_formatted_as_numpy after write_nested_sequence | read_col_unformated after write_array2d | read_col_unformated after write_flattened_sequence | read_col_unformated after write_nested_sequence | read_formatted_as_numpy after write_array2d | read_formatted_as_numpy after write_flattened_sequence | read_formatted_as_numpy after write_nested_sequence | read_unformated after write_array2d | read_unformated after write_flattened_sequence | read_unformated after write_nested_sequence | write_array2d | write_flattened_sequence | write_nested_sequence |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.008540 / 0.011353 (-0.002813) | 0.004538 / 0.011008 (-0.006470) | 0.101507 / 0.038508 (0.062999) | 0.029751 / 0.023109 (0.006641) | 0.292608 / 0.275898 (0.016710) | 0.354734 / 0.323480 (0.031254) | 0.007430 / 0.007986 (-0.000556) | 0.003365 / 0.004328 (-0.000964) | 0.078703 / 0.004250 (0.074452) | 0.034858 / 0.037052 (-0.002194) | 0.303518 / 0.258489 (0.045029) | 0.336523 / 0.293841 (0.042682) | 0.033741 / 0.128546 (-0.094805) | 0.011460 / 0.075646 (-0.064186) | 0.319551 / 0.419271 (-0.099721) | 0.041102 / 0.043533 (-0.002431) | 0.295914 / 0.255139 (0.040775) | 0.322142 / 0.283200 (0.038943) | 0.084694 / 0.141683 (-0.056989) | 1.481308 / 1.452155 (0.029153) | 1.530271 / 1.492716 (0.037554) |\n\n### Benchmark: benchmark_getitem\\_100B.json\n\n| metric | get_batch_of\\_1024\\_random_rows | get_batch_of\\_1024\\_rows | get_first_row | get_last_row |\n|--------|---|---|---|---|\n| new / old (diff) | 0.180516 / 0.018006 (0.162510) | 0.405741 / 0.000490 (0.405251) | 0.002806 / 0.000200 (0.002606) | 0.000072 / 0.000054 (0.000018) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.023359 / 0.037411 (-0.014052) | 0.096950 / 0.014526 (0.082424) | 0.103991 / 0.176557 (-0.072566) | 0.143700 / 0.737135 (-0.593435) | 0.106764 / 0.296338 (-0.189575) |\n\n### Benchmark: benchmark_iterating.json\n\n| metric | read 5000 | read 50000 | read_batch 50000 10 | read_batch 50000 100 | read_batch 50000 1000 | read_formatted numpy 5000 | read_formatted pandas 5000 | read_formatted tensorflow 5000 | read_formatted torch 5000 | read_formatted_batch numpy 5000 10 | read_formatted_batch numpy 5000 1000 | shuffled read 5000 | shuffled read 50000 | shuffled read_batch 50000 10 | shuffled read_batch 50000 100 | shuffled read_batch 50000 1000 | shuffled read_formatted numpy 5000 | shuffled read_formatted_batch numpy 5000 10 | shuffled read_formatted_batch numpy 5000 1000 |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.416966 / 0.215209 (0.201757) | 4.145601 / 2.077655 (2.067946) | 1.838258 / 1.504120 (0.334139) | 1.629396 / 1.541195 (0.088201) | 1.649707 / 1.468490 (0.181217) | 0.689624 / 4.584777 (-3.895153) | 3.414584 / 3.745712 (-0.331129) | 1.874295 / 5.269862 (-3.395566) | 1.251930 / 4.565676 (-3.313746) | 0.081782 / 0.424275 (-0.342493) | 0.012868 / 0.007607 (0.005261) | 0.523904 / 0.226044 (0.297859) | 5.251032 / 2.268929 (2.982104) | 2.301549 / 55.444624 (-53.143075) | 1.942110 / 6.876477 (-4.934367) | 2.023014 / 2.142072 (-0.119058) | 0.816492 / 4.805227 (-3.988736) | 0.150107 / 6.500664 (-6.350558) | 0.065118 / 0.075469 (-0.010351) |\n\n### Benchmark: benchmark_map_filter.json\n\n| metric | filter | map fast-tokenizer batched | map identity | map identity batched | map no-op batched | map no-op batched numpy | map no-op batched pandas | map no-op batched pytorch | map no-op batched tensorflow |\n|--------|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 1.226433 / 1.841788 (-0.615355) | 13.852569 / 8.074308 (5.778261) | 13.862779 / 10.191392 (3.671387) | 0.146361 / 0.680424 (-0.534062) | 0.028652 / 0.534201 (-0.505549) | 0.398251 / 0.579283 (-0.181032) | 0.403590 / 0.434364 (-0.030774) | 0.492184 / 0.540337 (-0.048154) | 0.581040 / 1.386936 (-0.805896) |\n\n</details>\nPyArrow==latest\n\n<details>\n<summary>Show updated benchmarks!</summary>\n\n### Benchmark: benchmark_array_xd.json\n\n| metric | read_batch_formatted_as_numpy after write_array2d | read_batch_formatted_as_numpy after write_flattened_sequence | read_batch_formatted_as_numpy after write_nested_sequence | read_batch_unformated after write_array2d | read_batch_unformated after write_flattened_sequence | read_batch_unformated after write_nested_sequence | read_col_formatted_as_numpy after write_array2d | read_col_formatted_as_numpy after write_flattened_sequence | read_col_formatted_as_numpy after write_nested_sequence | read_col_unformated after write_array2d | read_col_unformated after write_flattened_sequence | read_col_unformated after write_nested_sequence | read_formatted_as_numpy after write_array2d | read_formatted_as_numpy after write_flattened_sequence | read_formatted_as_numpy after write_nested_sequence | read_unformated after write_array2d | read_unformated after write_flattened_sequence | read_unformated after write_nested_sequence | write_array2d | write_flattened_sequence | write_nested_sequence |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.006859 / 0.011353 (-0.004494) | 0.004632 / 0.011008 (-0.006376) | 0.076653 / 0.038508 (0.038145) | 0.027865 / 0.023109 (0.004755) | 0.354472 / 0.275898 (0.078573) | 0.385462 / 0.323480 (0.061982) | 0.005125 / 0.007986 (-0.002861) | 0.003420 / 0.004328 (-0.000909) | 0.076018 / 0.004250 (0.071768) | 0.040197 / 0.037052 (0.003144) | 0.353675 / 0.258489 (0.095186) | 0.394911 / 0.293841 (0.101070) | 0.032909 / 0.128546 (-0.095637) | 0.011713 / 0.075646 (-0.063933) | 0.085921 / 0.419271 (-0.333350) | 0.044462 / 0.043533 (0.000929) | 0.349997 / 0.255139 (0.094858) | 0.375207 / 0.283200 (0.092008) | 0.091288 / 0.141683 (-0.050394) | 1.536515 / 1.452155 (0.084361) | 1.581878 / 1.492716 (0.089162) |\n\n### Benchmark: benchmark_getitem\\_100B.json\n\n| metric | get_batch_of\\_1024\\_random_rows | get_batch_of\\_1024\\_rows | get_first_row | get_last_row |\n|--------|---|---|---|---|\n| new / old (diff) | 0.273284 / 0.018006 (0.255277) | 0.424457 / 0.000490 (0.423967) | 0.044659 / 0.000200 (0.044459) | 0.000247 / 0.000054 (0.000192) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.025473 / 0.037411 (-0.011938) | 0.100014 / 0.014526 (0.085488) | 0.108551 / 0.176557 (-0.068006) | 0.147913 / 0.737135 (-0.589223) | 0.112729 / 0.296338 (-0.183610) |\n\n### Benchmark: benchmark_iterating.json\n\n| metric | read 5000 | read 50000 | read_batch 50000 10 | read_batch 50000 100 | read_batch 50000 1000 | read_formatted numpy 5000 | read_formatted pandas 5000 | read_formatted tensorflow 5000 | read_formatted torch 5000 | read_formatted_batch numpy 5000 10 | read_formatted_batch numpy 5000 1000 | shuffled read 5000 | shuffled read 50000 | shuffled read_batch 50000 10 | shuffled read_batch 50000 100 | shuffled read_batch 50000 1000 | shuffled read_formatted numpy 5000 | shuffled read_formatted_batch numpy 5000 10 | shuffled read_formatted_batch numpy 5000 1000 |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.448162 / 0.215209 (0.232953) | 4.472701 / 2.077655 (2.395046) | 2.078384 / 1.504120 (0.574264) | 1.861292 / 1.541195 (0.320097) | 1.920482 / 1.468490 (0.451991) | 0.706968 / 4.584777 (-3.877809) | 3.433109 / 3.745712 (-0.312603) | 1.898684 / 5.269862 (-3.371178) | 1.174375 / 4.565676 (-3.391302) | 0.083666 / 0.424275 (-0.340609) | 0.012388 / 0.007607 (0.004781) | 0.546011 / 0.226044 (0.319966) | 5.487514 / 2.268929 (3.218585) | 2.534124 / 55.444624 (-52.910500) | 2.168441 / 6.876477 (-4.708036) | 2.203458 / 2.142072 (0.061386) | 0.813333 / 4.805227 (-3.991894) | 0.153169 / 6.500664 (-6.347495) | 0.067151 / 0.075469 (-0.008318) |\n\n### Benchmark: benchmark_map_filter.json\n\n| metric | filter | map fast-tokenizer batched | map identity | map identity batched | map no-op batched | map no-op batched numpy | map no-op batched pandas | map no-op batched pytorch | map no-op batched tensorflow |\n|--------|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 1.277815 / 1.841788 (-0.563972) | 13.920545 / 8.074308 (5.846237) | 13.473801 / 10.191392 (3.282409) | 0.129035 / 0.680424 (-0.551389) | 0.016737 / 0.534201 (-0.517464) | 0.388413 / 0.579283 (-0.190870) | 0.388785 / 0.434364 (-0.045579) | 0.481735 / 0.540337 (-0.058602) | 0.576390 / 1.386936 (-0.810546) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#c4a4f96ef0a4ec4b25f0872f160fa1eb9d2e711c \"CML watermark\")\n" ]

[]

[ "IMO, the problem with this dataset is that it is not really a text/nlp dataset. These are just collections of tweet ids. So, ultimately, one needs to crawl twitter to get the actual text.", "That's true.\r\n\r\n", "at least it's clear in the description that one needs to collect the tweets : \r\n```\r\nThis resource is a result of a generic method for collecting parallel tweets.\r\n```", "Looks like this is failing for other datasets. Should I rebase it and push again?\r\nAlso rebasing and pushing is reflecting changes in many other files (ultimately forcing me to open a new branch and a new PR) any way to avoid this?", "No let me merge this one directly, it's fine", "merging since the CI is fixed on master" ]

[]

[ "_The documentation is not available anymore as the PR was closed or merged._", "I also ran the code in https://github.com/huggingface/datasets/issues/5111 and it works fine now :)", "This is ready for review :)" ]

[]

[]

[]

[ "do you have an idea of why it can occur @huggingface/datasets? The dataset consists of a single parquet file.", "Thanks for reporting @severo.\r\n\r\nI'm not able to reproduce that error. I get instead:\r\n```\r\nFileNotFoundError: [Errno 2] No such file or directory: 'orix/data/ChiSig/唐合乐-9-3.jpg'\r\n```\r\n\r\nWhich pyarrow version are you using? Mine is 6.0.1. ", "OK, I get now your error when not streaming.", "OK!\r\n\r\nIf it's useful, the pyarrow version is 7.0.0:\r\n\r\nhttps://github.com/huggingface/datasets-server/blob/487c39d87998f8d5a35972f1027d6c8e588e622d/services/worker/poetry.lock#L1537-L1543", "Apparently, there is something weird with that Parquet file: its schema is:\r\n```\r\nimages: extension<arrow.py_extension_type<pyarrow.lib.UnknownExtensionType>>\r\n```\r\n\r\nI have forced a right schema:\r\n```python\r\nfrom datasets import Features, Image, load_dataset\r\n\r\nfeatures = Features({\"images\": Image()})\r\nds = datasets.load_dataset(\"parquet\", split=\"train\", data_files=\"train-00000-of-00001.parquet\", features=features)\r\n```\r\nand then recreated a new Parquet file:\r\n```python\r\nds.to_parquet(\"train.parquet\")\r\n```\r\n\r\nNow this Parquet file has the right schema:\r\n```\r\nimages: struct<bytes: binary, path: string>\r\n child 0, bytes: binary\r\n child 1, path: string\r\n```\r\nand can be loaded normally:\r\n```python\r\nIn [26]: ds = load_dataset(\"parquet\", split=\"train\", data_files=\"dataset.parquet\")\r\nn [27]: ds\r\nOut[27]: \r\nDataset({\r\n features: ['images'],\r\n num_rows: 20\r\n})\r\n```" ]

[]

[ "_The documentation is not available anymore as the PR was closed or merged._", "<details>\n<summary>Show benchmarks</summary>\n\nPyArrow==6.0.0\n\n<details>\n<summary>Show updated benchmarks!</summary>\n\n### Benchmark: benchmark_array_xd.json\n\n| metric | read_batch_formatted_as_numpy after write_array2d | read_batch_formatted_as_numpy after write_flattened_sequence | read_batch_formatted_as_numpy after write_nested_sequence | read_batch_unformated after write_array2d | read_batch_unformated after write_flattened_sequence | read_batch_unformated after write_nested_sequence | read_col_formatted_as_numpy after write_array2d | read_col_formatted_as_numpy after write_flattened_sequence | read_col_formatted_as_numpy after write_nested_sequence | read_col_unformated after write_array2d | read_col_unformated after write_flattened_sequence | read_col_unformated after write_nested_sequence | read_formatted_as_numpy after write_array2d | read_formatted_as_numpy after write_flattened_sequence | read_formatted_as_numpy after write_nested_sequence | read_unformated after write_array2d | read_unformated after write_flattened_sequence | read_unformated after write_nested_sequence | write_array2d | write_flattened_sequence | write_nested_sequence |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.012220 / 0.011353 (0.000867) | 0.005943 / 0.011008 (-0.005065) | 0.128223 / 0.038508 (0.089715) | 0.037352 / 0.023109 (0.014242) | 0.397143 / 0.275898 (0.121245) | 0.483935 / 0.323480 (0.160455) | 0.010279 / 0.007986 (0.002293) | 0.004842 / 0.004328 (0.000513) | 0.101403 / 0.004250 (0.097153) | 0.042935 / 0.037052 (0.005883) | 0.421642 / 0.258489 (0.163153) | 0.456328 / 0.293841 (0.162487) | 0.065639 / 0.128546 (-0.062907) | 0.019820 / 0.075646 (-0.055826) | 0.426090 / 0.419271 (0.006818) | 0.069583 / 0.043533 (0.026051) | 0.402662 / 0.255139 (0.147523) | 0.428826 / 0.283200 (0.145626) | 0.116760 / 0.141683 (-0.024923) | 1.806216 / 1.452155 (0.354061) | 1.852629 / 1.492716 (0.359913) |\n\n### Benchmark: benchmark_getitem\\_100B.json\n\n| metric | get_batch_of\\_1024\\_random_rows | get_batch_of\\_1024\\_rows | get_first_row | get_last_row |\n|--------|---|---|---|---|\n| new / old (diff) | 0.226555 / 0.018006 (0.208548) | 0.584693 / 0.000490 (0.584203) | 0.008612 / 0.000200 (0.008412) | 0.000205 / 0.000054 (0.000150) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.028393 / 0.037411 (-0.009018) | 0.123355 / 0.014526 (0.108829) | 0.134423 / 0.176557 (-0.042133) | 0.188536 / 0.737135 (-0.548600) | 0.141595 / 0.296338 (-0.154743) |\n\n### Benchmark: benchmark_iterating.json\n\n| metric | read 5000 | read 50000 | read_batch 50000 10 | read_batch 50000 100 | read_batch 50000 1000 | read_formatted numpy 5000 | read_formatted pandas 5000 | read_formatted tensorflow 5000 | read_formatted torch 5000 | read_formatted_batch numpy 5000 10 | read_formatted_batch numpy 5000 1000 | shuffled read 5000 | shuffled read 50000 | shuffled read_batch 50000 10 | shuffled read_batch 50000 100 | shuffled read_batch 50000 1000 | shuffled read_formatted numpy 5000 | shuffled read_formatted_batch numpy 5000 10 | shuffled read_formatted_batch numpy 5000 1000 |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.589359 / 0.215209 (0.374150) | 5.974655 / 2.077655 (3.897001) | 2.465580 / 1.504120 (0.961460) | 2.007618 / 1.541195 (0.466424) | 2.078788 / 1.468490 (0.610298) | 1.216646 / 4.584777 (-3.368131) | 5.217516 / 3.745712 (1.471804) | 3.107188 / 5.269862 (-2.162674) | 2.251641 / 4.565676 (-2.314036) | 0.138640 / 0.424275 (-0.285635) | 0.015046 / 0.007607 (0.007439) | 0.780092 / 0.226044 (0.554048) | 7.749564 / 2.268929 (5.480635) | 3.080708 / 55.444624 (-52.363917) | 2.393897 / 6.876477 (-4.482579) | 2.387738 / 2.142072 (0.245665) | 1.458844 / 4.805227 (-3.346384) | 0.252476 / 6.500664 (-6.248188) | 0.076594 / 0.075469 (0.001125) |\n\n### Benchmark: benchmark_map_filter.json\n\n| metric | filter | map fast-tokenizer batched | map identity | map identity batched | map no-op batched | map no-op batched numpy | map no-op batched pandas | map no-op batched pytorch | map no-op batched tensorflow |\n|--------|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 1.540868 / 1.841788 (-0.300919) | 17.295684 / 8.074308 (9.221376) | 19.669300 / 10.191392 (9.477908) | 0.250315 / 0.680424 (-0.430109) | 0.045068 / 0.534201 (-0.489133) | 0.538840 / 0.579283 (-0.040443) | 0.584443 / 0.434364 (0.150079) | 0.614476 / 0.540337 (0.074138) | 0.729928 / 1.386936 (-0.657008) |\n\n</details>\nPyArrow==latest\n\n<details>\n<summary>Show updated benchmarks!</summary>\n\n### Benchmark: benchmark_array_xd.json\n\n| metric | read_batch_formatted_as_numpy after write_array2d | read_batch_formatted_as_numpy after write_flattened_sequence | read_batch_formatted_as_numpy after write_nested_sequence | read_batch_unformated after write_array2d | read_batch_unformated after write_flattened_sequence | read_batch_unformated after write_nested_sequence | read_col_formatted_as_numpy after write_array2d | read_col_formatted_as_numpy after write_flattened_sequence | read_col_formatted_as_numpy after write_nested_sequence | read_col_unformated after write_array2d | read_col_unformated after write_flattened_sequence | read_col_unformated after write_nested_sequence | read_formatted_as_numpy after write_array2d | read_formatted_as_numpy after write_flattened_sequence | read_formatted_as_numpy after write_nested_sequence | read_unformated after write_array2d | read_unformated after write_flattened_sequence | read_unformated after write_nested_sequence | write_array2d | write_flattened_sequence | write_nested_sequence |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.009218 / 0.011353 (-0.002135) | 0.006261 / 0.011008 (-0.004747) | 0.125541 / 0.038508 (0.087033) | 0.034405 / 0.023109 (0.011296) | 0.468381 / 0.275898 (0.192483) | 0.503336 / 0.323480 (0.179856) | 0.006839 / 0.007986 (-0.001146) | 0.004724 / 0.004328 (0.000396) | 0.097875 / 0.004250 (0.093625) | 0.051278 / 0.037052 (0.014225) | 0.473323 / 0.258489 (0.214834) | 0.537392 / 0.293841 (0.243551) | 0.055588 / 0.128546 (-0.072958) | 0.021041 / 0.075646 (-0.054605) | 0.416952 / 0.419271 (-0.002320) | 0.070128 / 0.043533 (0.026595) | 0.465224 / 0.255139 (0.210085) | 0.504678 / 0.283200 (0.221478) | 0.112504 / 0.141683 (-0.029179) | 1.865865 / 1.452155 (0.413710) | 1.988296 / 1.492716 (0.495580) |\n\n### Benchmark: benchmark_getitem\\_100B.json\n\n| metric | get_batch_of\\_1024\\_random_rows | get_batch_of\\_1024\\_rows | get_first_row | get_last_row |\n|--------|---|---|---|---|\n| new / old (diff) | 0.314170 / 0.018006 (0.296164) | 0.526726 / 0.000490 (0.526236) | 0.018691 / 0.000200 (0.018491) | 0.000128 / 0.000054 (0.000073) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.033772 / 0.037411 (-0.003639) | 0.124796 / 0.014526 (0.110270) | 0.134700 / 0.176557 (-0.041856) | 0.190595 / 0.737135 (-0.546541) | 0.143205 / 0.296338 (-0.153133) |\n\n### Benchmark: benchmark_iterating.json\n\n| metric | read 5000 | read 50000 | read_batch 50000 10 | read_batch 50000 100 | read_batch 50000 1000 | read_formatted numpy 5000 | read_formatted pandas 5000 | read_formatted tensorflow 5000 | read_formatted torch 5000 | read_formatted_batch numpy 5000 10 | read_formatted_batch numpy 5000 1000 | shuffled read 5000 | shuffled read 50000 | shuffled read_batch 50000 10 | shuffled read_batch 50000 100 | shuffled read_batch 50000 1000 | shuffled read_formatted numpy 5000 | shuffled read_formatted_batch numpy 5000 10 | shuffled read_formatted_batch numpy 5000 1000 |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.656708 / 0.215209 (0.441499) | 6.470503 / 2.077655 (4.392848) | 2.866430 / 1.504120 (1.362310) | 2.506846 / 1.541195 (0.965651) | 2.548669 / 1.468490 (1.080179) | 1.226695 / 4.584777 (-3.358082) | 5.117866 / 3.745712 (1.372153) | 3.032822 / 5.269862 (-2.237040) | 1.999152 / 4.565676 (-2.566524) | 0.142974 / 0.424275 (-0.281301) | 0.015011 / 0.007607 (0.007404) | 0.799729 / 0.226044 (0.573684) | 8.286313 / 2.268929 (6.017385) | 3.636482 / 55.444624 (-51.808142) | 2.888038 / 6.876477 (-3.988439) | 2.924982 / 2.142072 (0.782910) | 1.471996 / 4.805227 (-3.333231) | 0.257119 / 6.500664 (-6.243545) | 0.077294 / 0.075469 (0.001825) |\n\n### Benchmark: benchmark_map_filter.json\n\n| metric | filter | map fast-tokenizer batched | map identity | map identity batched | map no-op batched | map no-op batched numpy | map no-op batched pandas | map no-op batched pytorch | map no-op batched tensorflow |\n|--------|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 1.608290 / 1.841788 (-0.233497) | 17.599119 / 8.074308 (9.524811) | 18.917086 / 10.191392 (8.725694) | 0.236237 / 0.680424 (-0.444187) | 0.026061 / 0.534201 (-0.508140) | 0.527359 / 0.579283 (-0.051925) | 0.589176 / 0.434364 (0.154812) | 0.602310 / 0.540337 (0.061973) | 0.726756 / 1.386936 (-0.660180) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png \"CML watermark\")\n", "Issue reported to `pydantic`: \r\n- https://github.com/pydantic/pydantic/issues/4885\r\n\r\nFixing PR at `pydantic`:\r\n- https://github.com/pydantic/pydantic/pull/4886" ]

[]

[ "merging since the CI is fixed on master" ]

[]

[ "_The documentation is not available anymore as the PR was closed or merged._", "Spent the day familiarising myself with the huggingface line of products, and happened to run into some small issues here and there. Magically, I've found exactly one small issue in `transformers`, one in `accelerate` and now one in `datasets`, hah!\r\n\r\nAs for this PR, the issue seems solved according to the [new PR documentation](https://moon-ci-docs.huggingface.co/docs/datasets/pr_4986/en/process#map):\r\n\r\n" ]

[]

[ "_The documentation is not available anymore as the PR was closed or merged._", "Better yet have someone enable pre-commit CI https://pre-commit.ci/ and it will apply the pre-commit fixes to the PR automatically as an additional commit.", "@Skylion007 hi! I agree with @nateraw here, I'd better not force to use pre-commit so I'm not setting it up in the CI for now. And regarding end-of-file - currently it's being done by `black`. \r\n", "<details>\n<summary>Show benchmarks</summary>\n\nPyArrow==6.0.0\n\n<details>\n<summary>Show updated benchmarks!</summary>\n\n### Benchmark: benchmark_array_xd.json\n\n| metric | read_batch_formatted_as_numpy after write_array2d | read_batch_formatted_as_numpy after write_flattened_sequence | read_batch_formatted_as_numpy after write_nested_sequence | read_batch_unformated after write_array2d | read_batch_unformated after write_flattened_sequence | read_batch_unformated after write_nested_sequence | read_col_formatted_as_numpy after write_array2d | read_col_formatted_as_numpy after write_flattened_sequence | read_col_formatted_as_numpy after write_nested_sequence | read_col_unformated after write_array2d | read_col_unformated after write_flattened_sequence | read_col_unformated after write_nested_sequence | read_formatted_as_numpy after write_array2d | read_formatted_as_numpy after write_flattened_sequence | read_formatted_as_numpy after write_nested_sequence | read_unformated after write_array2d | read_unformated after write_flattened_sequence | read_unformated after write_nested_sequence | write_array2d | write_flattened_sequence | write_nested_sequence |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.008704 / 0.011353 (-0.002649) | 0.004448 / 0.011008 (-0.006560) | 0.099530 / 0.038508 (0.061022) | 0.029739 / 0.023109 (0.006629) | 0.329267 / 0.275898 (0.053369) | 0.368805 / 0.323480 (0.045325) | 0.006852 / 0.007986 (-0.001133) | 0.004575 / 0.004328 (0.000246) | 0.076838 / 0.004250 (0.072588) | 0.033885 / 0.037052 (-0.003167) | 0.336340 / 0.258489 (0.077851) | 0.384880 / 0.293841 (0.091039) | 0.034051 / 0.128546 (-0.094495) | 0.011638 / 0.075646 (-0.064009) | 0.321650 / 0.419271 (-0.097622) | 0.041202 / 0.043533 (-0.002330) | 0.330841 / 0.255139 (0.075702) | 0.361329 / 0.283200 (0.078130) | 0.084864 / 0.141683 (-0.056819) | 1.454005 / 1.452155 (0.001850) | 1.542167 / 1.492716 (0.049451) |\n\n### Benchmark: benchmark_getitem\\_100B.json\n\n| metric | get_batch_of\\_1024\\_random_rows | get_batch_of\\_1024\\_rows | get_first_row | get_last_row |\n|--------|---|---|---|---|\n| new / old (diff) | 0.196207 / 0.018006 (0.178200) | 0.400675 / 0.000490 (0.400185) | 0.000403 / 0.000200 (0.000203) | 0.000059 / 0.000054 (0.000005) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.022694 / 0.037411 (-0.014717) | 0.095139 / 0.014526 (0.080613) | 0.104129 / 0.176557 (-0.072427) | 0.168688 / 0.737135 (-0.568447) | 0.109243 / 0.296338 (-0.187096) |\n\n### Benchmark: benchmark_iterating.json\n\n| metric | read 5000 | read 50000 | read_batch 50000 10 | read_batch 50000 100 | read_batch 50000 1000 | read_formatted numpy 5000 | read_formatted pandas 5000 | read_formatted tensorflow 5000 | read_formatted torch 5000 | read_formatted_batch numpy 5000 10 | read_formatted_batch numpy 5000 1000 | shuffled read 5000 | shuffled read 50000 | shuffled read_batch 50000 10 | shuffled read_batch 50000 100 | shuffled read_batch 50000 1000 | shuffled read_formatted numpy 5000 | shuffled read_formatted_batch numpy 5000 10 | shuffled read_formatted_batch numpy 5000 1000 |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.427520 / 0.215209 (0.212311) | 4.237726 / 2.077655 (2.160071) | 2.191887 / 1.504120 (0.687767) | 1.987750 / 1.541195 (0.446555) | 1.996540 / 1.468490 (0.528050) | 0.696416 / 4.584777 (-3.888361) | 3.454536 / 3.745712 (-0.291176) | 2.023600 / 5.269862 (-3.246261) | 1.336394 / 4.565676 (-3.229282) | 0.082933 / 0.424275 (-0.341342) | 0.012572 / 0.007607 (0.004965) | 0.534330 / 0.226044 (0.308285) | 5.347588 / 2.268929 (3.078659) | 2.640397 / 55.444624 (-52.804228) | 2.338266 / 6.876477 (-4.538211) | 2.431969 / 2.142072 (0.289897) | 0.821335 / 4.805227 (-3.983893) | 0.151905 / 6.500664 (-6.348759) | 0.067983 / 0.075469 (-0.007486) |\n\n### Benchmark: benchmark_map_filter.json\n\n| metric | filter | map fast-tokenizer batched | map identity | map identity batched | map no-op batched | map no-op batched numpy | map no-op batched pandas | map no-op batched pytorch | map no-op batched tensorflow |\n|--------|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 1.228841 / 1.841788 (-0.612947) | 13.660437 / 8.074308 (5.586128) | 13.729442 / 10.191392 (3.538050) | 0.165835 / 0.680424 (-0.514589) | 0.028753 / 0.534201 (-0.505448) | 0.400143 / 0.579283 (-0.179140) | 0.403714 / 0.434364 (-0.030650) | 0.492168 / 0.540337 (-0.048170) | 0.581151 / 1.386936 (-0.805785) |\n\n</details>\nPyArrow==latest\n\n<details>\n<summary>Show updated benchmarks!</summary>\n\n### Benchmark: benchmark_array_xd.json\n\n| metric | read_batch_formatted_as_numpy after write_array2d | read_batch_formatted_as_numpy after write_flattened_sequence | read_batch_formatted_as_numpy after write_nested_sequence | read_batch_unformated after write_array2d | read_batch_unformated after write_flattened_sequence | read_batch_unformated after write_nested_sequence | read_col_formatted_as_numpy after write_array2d | read_col_formatted_as_numpy after write_flattened_sequence | read_col_formatted_as_numpy after write_nested_sequence | read_col_unformated after write_array2d | read_col_unformated after write_flattened_sequence | read_col_unformated after write_nested_sequence | read_formatted_as_numpy after write_array2d | read_formatted_as_numpy after write_flattened_sequence | read_formatted_as_numpy after write_nested_sequence | read_unformated after write_array2d | read_unformated after write_flattened_sequence | read_unformated after write_nested_sequence | write_array2d | write_flattened_sequence | write_nested_sequence |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.006289 / 0.011353 (-0.005064) | 0.004419 / 0.011008 (-0.006589) | 0.077220 / 0.038508 (0.038712) | 0.027170 / 0.023109 (0.004060) | 0.344988 / 0.275898 (0.069090) | 0.374150 / 0.323480 (0.050670) | 0.004842 / 0.007986 (-0.003144) | 0.003289 / 0.004328 (-0.001039) | 0.076200 / 0.004250 (0.071950) | 0.036287 / 0.037052 (-0.000766) | 0.345764 / 0.258489 (0.087275) | 0.387439 / 0.293841 (0.093599) | 0.031547 / 0.128546 (-0.096999) | 0.011586 / 0.075646 (-0.064060) | 0.086599 / 0.419271 (-0.332672) | 0.042338 / 0.043533 (-0.001195) | 0.355384 / 0.255139 (0.100246) | 0.369474 / 0.283200 (0.086275) | 0.090945 / 0.141683 (-0.050738) | 1.488632 / 1.452155 (0.036477) | 1.554606 / 1.492716 (0.061890) |\n\n### Benchmark: benchmark_getitem\\_100B.json\n\n| metric | get_batch_of\\_1024\\_random_rows | get_batch_of\\_1024\\_rows | get_first_row | get_last_row |\n|--------|---|---|---|---|\n| new / old (diff) | 0.212962 / 0.018006 (0.194956) | 0.399647 / 0.000490 (0.399157) | 0.003055 / 0.000200 (0.002856) | 0.000083 / 0.000054 (0.000029) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.024349 / 0.037411 (-0.013062) | 0.100342 / 0.014526 (0.085817) | 0.105657 / 0.176557 (-0.070899) | 0.175139 / 0.737135 (-0.561997) | 0.110014 / 0.296338 (-0.186324) |\n\n### Benchmark: benchmark_iterating.json\n\n| metric | read 5000 | read 50000 | read_batch 50000 10 | read_batch 50000 100 | read_batch 50000 1000 | read_formatted numpy 5000 | read_formatted pandas 5000 | read_formatted tensorflow 5000 | read_formatted torch 5000 | read_formatted_batch numpy 5000 10 | read_formatted_batch numpy 5000 1000 | shuffled read 5000 | shuffled read 50000 | shuffled read_batch 50000 10 | shuffled read_batch 50000 100 | shuffled read_batch 50000 1000 | shuffled read_formatted numpy 5000 | shuffled read_formatted_batch numpy 5000 10 | shuffled read_formatted_batch numpy 5000 1000 |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.434785 / 0.215209 (0.219575) | 4.346950 / 2.077655 (2.269295) | 2.045411 / 1.504120 (0.541291) | 1.844258 / 1.541195 (0.303064) | 1.889503 / 1.468490 (0.421013) | 0.704530 / 4.584777 (-3.880247) | 3.362435 / 3.745712 (-0.383277) | 2.797205 / 5.269862 (-2.472656) | 1.504431 / 4.565676 (-3.061245) | 0.083331 / 0.424275 (-0.340945) | 0.012274 / 0.007607 (0.004666) | 0.531123 / 0.226044 (0.305078) | 5.322588 / 2.268929 (3.053660) | 2.483875 / 55.444624 (-52.960750) | 2.147218 / 6.876477 (-4.729258) | 2.164024 / 2.142072 (0.021952) | 0.807191 / 4.805227 (-3.998036) | 0.151189 / 6.500664 (-6.349475) | 0.068027 / 0.075469 (-0.007442) |\n\n### Benchmark: benchmark_map_filter.json\n\n| metric | filter | map fast-tokenizer batched | map identity | map identity batched | map no-op batched | map no-op batched numpy | map no-op batched pandas | map no-op batched pytorch | map no-op batched tensorflow |\n|--------|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 1.316001 / 1.841788 (-0.525787) | 13.892785 / 8.074308 (5.818477) | 13.485982 / 10.191392 (3.294590) | 0.138904 / 0.680424 (-0.541520) | 0.016748 / 0.534201 (-0.517453) | 0.379840 / 0.579283 (-0.199443) | 0.384854 / 0.434364 (-0.049510) | 0.464275 / 0.540337 (-0.076063) | 0.553622 / 1.386936 (-0.833314) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#a940972a9a38543b2066129dc6e7987e08dca082 \"CML watermark\")\n", "<details>\n<summary>Show benchmarks</summary>\n\nPyArrow==6.0.0\n\n<details>\n<summary>Show updated benchmarks!</summary>\n\n### Benchmark: benchmark_array_xd.json\n\n| metric | read_batch_formatted_as_numpy after write_array2d | read_batch_formatted_as_numpy after write_flattened_sequence | read_batch_formatted_as_numpy after write_nested_sequence | read_batch_unformated after write_array2d | read_batch_unformated after write_flattened_sequence | read_batch_unformated after write_nested_sequence | read_col_formatted_as_numpy after write_array2d | read_col_formatted_as_numpy after write_flattened_sequence | read_col_formatted_as_numpy after write_nested_sequence | read_col_unformated after write_array2d | read_col_unformated after write_flattened_sequence | read_col_unformated after write_nested_sequence | read_formatted_as_numpy after write_array2d | read_formatted_as_numpy after write_flattened_sequence | read_formatted_as_numpy after write_nested_sequence | read_unformated after write_array2d | read_unformated after write_flattened_sequence | read_unformated after write_nested_sequence | write_array2d | write_flattened_sequence | write_nested_sequence |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.009179 / 0.011353 (-0.002174) | 0.005080 / 0.011008 (-0.005929) | 0.099061 / 0.038508 (0.060553) | 0.035252 / 0.023109 (0.012143) | 0.293496 / 0.275898 (0.017598) | 0.360365 / 0.323480 (0.036886) | 0.007757 / 0.007986 (-0.000229) | 0.003985 / 0.004328 (-0.000343) | 0.076021 / 0.004250 (0.071771) | 0.042286 / 0.037052 (0.005233) | 0.316542 / 0.258489 (0.058053) | 0.341711 / 0.293841 (0.047870) | 0.037970 / 0.128546 (-0.090576) | 0.011977 / 0.075646 (-0.063670) | 0.333341 / 0.419271 (-0.085931) | 0.049211 / 0.043533 (0.005678) | 0.297401 / 0.255139 (0.042262) | 0.313424 / 0.283200 (0.030224) | 0.105719 / 0.141683 (-0.035964) | 1.487879 / 1.452155 (0.035724) | 1.529785 / 1.492716 (0.037068) |\n\n### Benchmark: benchmark_getitem\\_100B.json\n\n| metric | get_batch_of\\_1024\\_random_rows | get_batch_of\\_1024\\_rows | get_first_row | get_last_row |\n|--------|---|---|---|---|\n| new / old (diff) | 0.201062 / 0.018006 (0.183056) | 0.438024 / 0.000490 (0.437534) | 0.002129 / 0.000200 (0.001929) | 0.000083 / 0.000054 (0.000028) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.026422 / 0.037411 (-0.010989) | 0.104863 / 0.014526 (0.090337) | 0.114934 / 0.176557 (-0.061623) | 0.179173 / 0.737135 (-0.557962) | 0.119734 / 0.296338 (-0.176604) |\n\n### Benchmark: benchmark_iterating.json\n\n| metric | read 5000 | read 50000 | read_batch 50000 10 | read_batch 50000 100 | read_batch 50000 1000 | read_formatted numpy 5000 | read_formatted pandas 5000 | read_formatted tensorflow 5000 | read_formatted torch 5000 | read_formatted_batch numpy 5000 10 | read_formatted_batch numpy 5000 1000 | shuffled read 5000 | shuffled read 50000 | shuffled read_batch 50000 10 | shuffled read_batch 50000 100 | shuffled read_batch 50000 1000 | shuffled read_formatted numpy 5000 | shuffled read_formatted_batch numpy 5000 10 | shuffled read_formatted_batch numpy 5000 1000 |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.397195 / 0.215209 (0.181986) | 3.959945 / 2.077655 (1.882290) | 1.794059 / 1.504120 (0.289939) | 1.606814 / 1.541195 (0.065619) | 1.674681 / 1.468490 (0.206191) | 0.680130 / 4.584777 (-3.904646) | 3.742730 / 3.745712 (-0.002982) | 2.021793 / 5.269862 (-3.248069) | 1.322726 / 4.565676 (-3.242950) | 0.084519 / 0.424275 (-0.339756) | 0.012012 / 0.007607 (0.004405) | 0.510076 / 0.226044 (0.284032) | 5.084163 / 2.268929 (2.815234) | 2.241032 / 55.444624 (-53.203592) | 1.911936 / 6.876477 (-4.964540) | 1.947992 / 2.142072 (-0.194080) | 0.838779 / 4.805227 (-3.966448) | 0.165103 / 6.500664 (-6.335561) | 0.060722 / 0.075469 (-0.014747) |\n\n### Benchmark: benchmark_map_filter.json\n\n| metric | filter | map fast-tokenizer batched | map identity | map identity batched | map no-op batched | map no-op batched numpy | map no-op batched pandas | map no-op batched pytorch | map no-op batched tensorflow |\n|--------|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 1.180274 / 1.841788 (-0.661514) | 14.285364 / 8.074308 (6.211056) | 12.941205 / 10.191392 (2.749813) | 0.153815 / 0.680424 (-0.526609) | 0.028554 / 0.534201 (-0.505647) | 0.441551 / 0.579283 (-0.137732) | 0.434906 / 0.434364 (0.000542) | 0.516120 / 0.540337 (-0.024217) | 0.603062 / 1.386936 (-0.783874) |\n\n</details>\nPyArrow==latest\n\n<details>\n<summary>Show updated benchmarks!</summary>\n\n### Benchmark: benchmark_array_xd.json\n\n| metric | read_batch_formatted_as_numpy after write_array2d | read_batch_formatted_as_numpy after write_flattened_sequence | read_batch_formatted_as_numpy after write_nested_sequence | read_batch_unformated after write_array2d | read_batch_unformated after write_flattened_sequence | read_batch_unformated after write_nested_sequence | read_col_formatted_as_numpy after write_array2d | read_col_formatted_as_numpy after write_flattened_sequence | read_col_formatted_as_numpy after write_nested_sequence | read_col_unformated after write_array2d | read_col_unformated after write_flattened_sequence | read_col_unformated after write_nested_sequence | read_formatted_as_numpy after write_array2d | read_formatted_as_numpy after write_flattened_sequence | read_formatted_as_numpy after write_nested_sequence | read_unformated after write_array2d | read_unformated after write_flattened_sequence | read_unformated after write_nested_sequence | write_array2d | write_flattened_sequence | write_nested_sequence |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.007287 / 0.011353 (-0.004066) | 0.004998 / 0.011008 (-0.006010) | 0.074997 / 0.038508 (0.036489) | 0.033209 / 0.023109 (0.010100) | 0.336836 / 0.275898 (0.060938) | 0.365562 / 0.323480 (0.042082) | 0.005739 / 0.007986 (-0.002246) | 0.003942 / 0.004328 (-0.000387) | 0.074681 / 0.004250 (0.070430) | 0.049530 / 0.037052 (0.012478) | 0.335642 / 0.258489 (0.077153) | 0.388874 / 0.293841 (0.095033) | 0.037198 / 0.128546 (-0.091349) | 0.011983 / 0.075646 (-0.063664) | 0.087601 / 0.419271 (-0.331671) | 0.053761 / 0.043533 (0.010228) | 0.334142 / 0.255139 (0.079003) | 0.351348 / 0.283200 (0.068148) | 0.107462 / 0.141683 (-0.034221) | 1.497015 / 1.452155 (0.044860) | 1.608287 / 1.492716 (0.115571) |\n\n### Benchmark: benchmark_getitem\\_100B.json\n\n| metric | get_batch_of\\_1024\\_random_rows | get_batch_of\\_1024\\_rows | get_first_row | get_last_row |\n|--------|---|---|---|---|\n| new / old (diff) | 0.255395 / 0.018006 (0.237389) | 0.439141 / 0.000490 (0.438651) | 0.021391 / 0.000200 (0.021191) | 0.000230 / 0.000054 (0.000176) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.028331 / 0.037411 (-0.009080) | 0.108744 / 0.014526 (0.094218) | 0.118201 / 0.176557 (-0.058355) | 0.189556 / 0.737135 (-0.547579) | 0.123112 / 0.296338 (-0.173226) |\n\n### Benchmark: benchmark_iterating.json\n\n| metric | read 5000 | read 50000 | read_batch 50000 10 | read_batch 50000 100 | read_batch 50000 1000 | read_formatted numpy 5000 | read_formatted pandas 5000 | read_formatted tensorflow 5000 | read_formatted torch 5000 | read_formatted_batch numpy 5000 10 | read_formatted_batch numpy 5000 1000 | shuffled read 5000 | shuffled read 50000 | shuffled read_batch 50000 10 | shuffled read_batch 50000 100 | shuffled read_batch 50000 1000 | shuffled read_formatted numpy 5000 | shuffled read_formatted_batch numpy 5000 10 | shuffled read_formatted_batch numpy 5000 1000 |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.431394 / 0.215209 (0.216185) | 4.296121 / 2.077655 (2.218466) | 2.126371 / 1.504120 (0.622251) | 1.978178 / 1.541195 (0.436983) | 2.082674 / 1.468490 (0.614184) | 0.701789 / 4.584777 (-3.882988) | 3.791495 / 3.745712 (0.045783) | 2.115267 / 5.269862 (-3.154594) | 1.342159 / 4.565676 (-3.223517) | 0.088132 / 0.424275 (-0.336143) | 0.011903 / 0.007607 (0.004295) | 0.528398 / 0.226044 (0.302354) | 5.270077 / 2.268929 (3.001148) | 2.498860 / 55.444624 (-52.945765) | 2.155515 / 6.876477 (-4.720962) | 2.192866 / 2.142072 (0.050793) | 0.859596 / 4.805227 (-3.945631) | 0.170544 / 6.500664 (-6.330120) | 0.063883 / 0.075469 (-0.011587) |\n\n### Benchmark: benchmark_map_filter.json\n\n| metric | filter | map fast-tokenizer batched | map identity | map identity batched | map no-op batched | map no-op batched numpy | map no-op batched pandas | map no-op batched pytorch | map no-op batched tensorflow |\n|--------|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 1.240679 / 1.841788 (-0.601109) | 14.497379 / 8.074308 (6.423071) | 12.881417 / 10.191392 (2.690025) | 0.147295 / 0.680424 (-0.533129) | 0.017465 / 0.534201 (-0.516736) | 0.424695 / 0.579283 (-0.154588) | 0.414929 / 0.434364 (-0.019435) | 0.536079 / 0.540337 (-0.004259) | 0.638245 / 1.386936 (-0.748691) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#a940972a9a38543b2066129dc6e7987e08dca082 \"CML watermark\")\n", "<details>\n<summary>Show benchmarks</summary>\n\nPyArrow==6.0.0\n\n<details>\n<summary>Show updated benchmarks!</summary>\n\n### Benchmark: benchmark_array_xd.json\n\n| metric | read_batch_formatted_as_numpy after write_array2d | read_batch_formatted_as_numpy after write_flattened_sequence | read_batch_formatted_as_numpy after write_nested_sequence | read_batch_unformated after write_array2d | read_batch_unformated after write_flattened_sequence | read_batch_unformated after write_nested_sequence | read_col_formatted_as_numpy after write_array2d | read_col_formatted_as_numpy after write_flattened_sequence | read_col_formatted_as_numpy after write_nested_sequence | read_col_unformated after write_array2d | read_col_unformated after write_flattened_sequence | read_col_unformated after write_nested_sequence | read_formatted_as_numpy after write_array2d | read_formatted_as_numpy after write_flattened_sequence | read_formatted_as_numpy after write_nested_sequence | read_unformated after write_array2d | read_unformated after write_flattened_sequence | read_unformated after write_nested_sequence | write_array2d | write_flattened_sequence | write_nested_sequence |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.008806 / 0.011353 (-0.002547) | 0.004712 / 0.011008 (-0.006297) | 0.102383 / 0.038508 (0.063875) | 0.030260 / 0.023109 (0.007151) | 0.330175 / 0.275898 (0.054277) | 0.376816 / 0.323480 (0.053337) | 0.008065 / 0.007986 (0.000079) | 0.003534 / 0.004328 (-0.000794) | 0.078824 / 0.004250 (0.074573) | 0.036704 / 0.037052 (-0.000349) | 0.331848 / 0.258489 (0.073359) | 0.351031 / 0.293841 (0.057190) | 0.033406 / 0.128546 (-0.095140) | 0.011543 / 0.075646 (-0.064103) | 0.322114 / 0.419271 (-0.097157) | 0.041249 / 0.043533 (-0.002284) | 0.309413 / 0.255139 (0.054274) | 0.329156 / 0.283200 (0.045956) | 0.088636 / 0.141683 (-0.053047) | 1.508226 / 1.452155 (0.056071) | 1.557203 / 1.492716 (0.064487) |\n\n### Benchmark: benchmark_getitem\\_100B.json\n\n| metric | get_batch_of\\_1024\\_random_rows | get_batch_of\\_1024\\_rows | get_first_row | get_last_row |\n|--------|---|---|---|---|\n| new / old (diff) | 0.196696 / 0.018006 (0.178690) | 0.426360 / 0.000490 (0.425870) | 0.001263 / 0.000200 (0.001064) | 0.000079 / 0.000054 (0.000024) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.023747 / 0.037411 (-0.013664) | 0.100756 / 0.014526 (0.086230) | 0.105817 / 0.176557 (-0.070739) | 0.172573 / 0.737135 (-0.564562) | 0.110705 / 0.296338 (-0.185634) |\n\n### Benchmark: benchmark_iterating.json\n\n| metric | read 5000 | read 50000 | read_batch 50000 10 | read_batch 50000 100 | read_batch 50000 1000 | read_formatted numpy 5000 | read_formatted pandas 5000 | read_formatted tensorflow 5000 | read_formatted torch 5000 | read_formatted_batch numpy 5000 10 | read_formatted_batch numpy 5000 1000 | shuffled read 5000 | shuffled read 50000 | shuffled read_batch 50000 10 | shuffled read_batch 50000 100 | shuffled read_batch 50000 1000 | shuffled read_formatted numpy 5000 | shuffled read_formatted_batch numpy 5000 10 | shuffled read_formatted_batch numpy 5000 1000 |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.436913 / 0.215209 (0.221704) | 4.365753 / 2.077655 (2.288099) | 2.201346 / 1.504120 (0.697226) | 1.978800 / 1.541195 (0.437605) | 1.951585 / 1.468490 (0.483094) | 0.699208 / 4.584777 (-3.885569) | 3.381492 / 3.745712 (-0.364220) | 2.966174 / 5.269862 (-2.303687) | 1.487521 / 4.565676 (-3.078156) | 0.082673 / 0.424275 (-0.341602) | 0.012436 / 0.007607 (0.004829) | 0.553276 / 0.226044 (0.327232) | 5.554081 / 2.268929 (3.285153) | 2.653286 / 55.444624 (-52.791339) | 2.404788 / 6.876477 (-4.471689) | 2.484610 / 2.142072 (0.342537) | 0.817073 / 4.805227 (-3.988154) | 0.151619 / 6.500664 (-6.349045) | 0.068259 / 0.075469 (-0.007210) |\n\n### Benchmark: benchmark_map_filter.json\n\n| metric | filter | map fast-tokenizer batched | map identity | map identity batched | map no-op batched | map no-op batched numpy | map no-op batched pandas | map no-op batched pytorch | map no-op batched tensorflow |\n|--------|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 1.273481 / 1.841788 (-0.568306) | 13.908825 / 8.074308 (5.834517) | 13.106695 / 10.191392 (2.915303) | 0.139609 / 0.680424 (-0.540815) | 0.028425 / 0.534201 (-0.505776) | 0.395626 / 0.579283 (-0.183657) | 0.405526 / 0.434364 (-0.028838) | 0.465628 / 0.540337 (-0.074709) | 0.542824 / 1.386936 (-0.844112) |\n\n</details>\nPyArrow==latest\n\n<details>\n<summary>Show updated benchmarks!</summary>\n\n### Benchmark: benchmark_array_xd.json\n\n| metric | read_batch_formatted_as_numpy after write_array2d | read_batch_formatted_as_numpy after write_flattened_sequence | read_batch_formatted_as_numpy after write_nested_sequence | read_batch_unformated after write_array2d | read_batch_unformated after write_flattened_sequence | read_batch_unformated after write_nested_sequence | read_col_formatted_as_numpy after write_array2d | read_col_formatted_as_numpy after write_flattened_sequence | read_col_formatted_as_numpy after write_nested_sequence | read_col_unformated after write_array2d | read_col_unformated after write_flattened_sequence | read_col_unformated after write_nested_sequence | read_formatted_as_numpy after write_array2d | read_formatted_as_numpy after write_flattened_sequence | read_formatted_as_numpy after write_nested_sequence | read_unformated after write_array2d | read_unformated after write_flattened_sequence | read_unformated after write_nested_sequence | write_array2d | write_flattened_sequence | write_nested_sequence |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.006821 / 0.011353 (-0.004532) | 0.004570 / 0.011008 (-0.006438) | 0.076568 / 0.038508 (0.038060) | 0.028109 / 0.023109 (0.004999) | 0.342768 / 0.275898 (0.066870) | 0.390680 / 0.323480 (0.067200) | 0.005056 / 0.007986 (-0.002930) | 0.003359 / 0.004328 (-0.000970) | 0.075835 / 0.004250 (0.071584) | 0.038888 / 0.037052 (0.001836) | 0.343489 / 0.258489 (0.085000) | 0.400766 / 0.293841 (0.106925) | 0.031816 / 0.128546 (-0.096730) | 0.011637 / 0.075646 (-0.064009) | 0.085474 / 0.419271 (-0.333797) | 0.041740 / 0.043533 (-0.001793) | 0.342501 / 0.255139 (0.087362) | 0.377467 / 0.283200 (0.094267) | 0.091532 / 0.141683 (-0.050151) | 1.457368 / 1.452155 (0.005213) | 1.537187 / 1.492716 (0.044471) |\n\n### Benchmark: benchmark_getitem\\_100B.json\n\n| metric | get_batch_of\\_1024\\_random_rows | get_batch_of\\_1024\\_rows | get_first_row | get_last_row |\n|--------|---|---|---|---|\n| new / old (diff) | 0.187507 / 0.018006 (0.169501) | 0.415706 / 0.000490 (0.415217) | 0.001816 / 0.000200 (0.001616) | 0.000072 / 0.000054 (0.000018) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.026251 / 0.037411 (-0.011161) | 0.106609 / 0.014526 (0.092083) | 0.109822 / 0.176557 (-0.066735) | 0.180462 / 0.737135 (-0.556674) | 0.114647 / 0.296338 (-0.181691) |\n\n### Benchmark: benchmark_iterating.json\n\n| metric | read 5000 | read 50000 | read_batch 50000 10 | read_batch 50000 100 | read_batch 50000 1000 | read_formatted numpy 5000 | read_formatted pandas 5000 | read_formatted tensorflow 5000 | read_formatted torch 5000 | read_formatted_batch numpy 5000 10 | read_formatted_batch numpy 5000 1000 | shuffled read 5000 | shuffled read 50000 | shuffled read_batch 50000 10 | shuffled read_batch 50000 100 | shuffled read_batch 50000 1000 | shuffled read_formatted numpy 5000 | shuffled read_formatted_batch numpy 5000 10 | shuffled read_formatted_batch numpy 5000 1000 |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.438804 / 0.215209 (0.223595) | 4.387960 / 2.077655 (2.310306) | 2.056804 / 1.504120 (0.552684) | 1.848584 / 1.541195 (0.307389) | 1.939470 / 1.468490 (0.470980) | 0.702539 / 4.584777 (-3.882238) | 3.419535 / 3.745712 (-0.326177) | 1.933889 / 5.269862 (-3.335973) | 1.189631 / 4.565676 (-3.376045) | 0.084105 / 0.424275 (-0.340170) | 0.012520 / 0.007607 (0.004913) | 0.538125 / 0.226044 (0.312081) | 5.370000 / 2.268929 (3.101072) | 2.497487 / 55.444624 (-52.947137) | 2.156054 / 6.876477 (-4.720423) | 2.225909 / 2.142072 (0.083837) | 0.811456 / 4.805227 (-3.993771) | 0.151461 / 6.500664 (-6.349203) | 0.066940 / 0.075469 (-0.008530) |\n\n### Benchmark: benchmark_map_filter.json\n\n| metric | filter | map fast-tokenizer batched | map identity | map identity batched | map no-op batched | map no-op batched numpy | map no-op batched pandas | map no-op batched pytorch | map no-op batched tensorflow |\n|--------|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 1.301246 / 1.841788 (-0.540542) | 14.459755 / 8.074308 (6.385447) | 13.147151 / 10.191392 (2.955759) | 0.129236 / 0.680424 (-0.551188) | 0.016427 / 0.534201 (-0.517774) | 0.380047 / 0.579283 (-0.199236) | 0.392217 / 0.434364 (-0.042147) | 0.470338 / 0.540337 (-0.069999) | 0.559800 / 1.386936 (-0.827136) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#a940972a9a38543b2066129dc6e7987e08dca082 \"CML watermark\")\n" ]

[ { "color": "94203D", "default": false, "description": "", "id": 2107841032, "name": "nlp-viewer", "node_id": "MDU6TGFiZWwyMTA3ODQxMDMy", "url": "https://api.github.com/repos/huggingface/datasets/labels/nlp-viewer" } ]

[ "Thanks for reporting !\r\nWe'll free some memory" ]

[]

[]

[]

[ "_The documentation is not available anymore as the PR was closed or merged._", "<details>\n<summary>Show benchmarks</summary>\n\nPyArrow==8.0.0\n\n<details>\n<summary>Show updated benchmarks!</summary>\n\n### Benchmark: benchmark_array_xd.json\n\n| metric | read_batch_formatted_as_numpy after write_array2d | read_batch_formatted_as_numpy after write_flattened_sequence | read_batch_formatted_as_numpy after write_nested_sequence | read_batch_unformated after write_array2d | read_batch_unformated after write_flattened_sequence | read_batch_unformated after write_nested_sequence | read_col_formatted_as_numpy after write_array2d | read_col_formatted_as_numpy after write_flattened_sequence | read_col_formatted_as_numpy after write_nested_sequence | read_col_unformated after write_array2d | read_col_unformated after write_flattened_sequence | read_col_unformated after write_nested_sequence | read_formatted_as_numpy after write_array2d | read_formatted_as_numpy after write_flattened_sequence | read_formatted_as_numpy after write_nested_sequence | read_unformated after write_array2d | read_unformated after write_flattened_sequence | read_unformated after write_nested_sequence | write_array2d | write_flattened_sequence | write_nested_sequence |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.007980 / 0.011353 (-0.003373) | 0.005351 / 0.011008 (-0.005657) | 0.096325 / 0.038508 (0.057817) | 0.034204 / 0.023109 (0.011095) | 0.328080 / 0.275898 (0.052182) | 0.361519 / 0.323480 (0.038039) | 0.005954 / 0.007986 (-0.002032) | 0.004106 / 0.004328 (-0.000222) | 0.072827 / 0.004250 (0.068576) | 0.050522 / 0.037052 (0.013470) | 0.326975 / 0.258489 (0.068486) | 0.373180 / 0.293841 (0.079339) | 0.037024 / 0.128546 (-0.091522) | 0.012347 / 0.075646 (-0.063299) | 0.332341 / 0.419271 (-0.086931) | 0.050695 / 0.043533 (0.007162) | 0.328298 / 0.255139 (0.073159) | 0.352808 / 0.283200 (0.069608) | 0.101637 / 0.141683 (-0.040046) | 1.435172 / 1.452155 (-0.016982) | 1.529797 / 1.492716 (0.037080) |\n\n### Benchmark: benchmark_getitem\\_100B.json\n\n| metric | get_batch_of\\_1024\\_random_rows | get_batch_of\\_1024\\_rows | get_first_row | get_last_row |\n|--------|---|---|---|---|\n| new / old (diff) | 0.305727 / 0.018006 (0.287721) | 0.583951 / 0.000490 (0.583462) | 0.011699 / 0.000200 (0.011499) | 0.000345 / 0.000054 (0.000290) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.027917 / 0.037411 (-0.009495) | 0.107698 / 0.014526 (0.093173) | 0.120572 / 0.176557 (-0.055985) | 0.176066 / 0.737135 (-0.561069) | 0.125348 / 0.296338 (-0.170991) |\n\n### Benchmark: benchmark_iterating.json\n\n| metric | read 5000 | read 50000 | read_batch 50000 10 | read_batch 50000 100 | read_batch 50000 1000 | read_formatted numpy 5000 | read_formatted pandas 5000 | read_formatted tensorflow 5000 | read_formatted torch 5000 | read_formatted_batch numpy 5000 10 | read_formatted_batch numpy 5000 1000 | shuffled read 5000 | shuffled read 50000 | shuffled read_batch 50000 10 | shuffled read_batch 50000 100 | shuffled read_batch 50000 1000 | shuffled read_formatted numpy 5000 | shuffled read_formatted_batch numpy 5000 10 | shuffled read_formatted_batch numpy 5000 1000 |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.411980 / 0.215209 (0.196771) | 4.113135 / 2.077655 (2.035480) | 1.868725 / 1.504120 (0.364605) | 1.677422 / 1.541195 (0.136227) | 1.796759 / 1.468490 (0.328269) | 0.701957 / 4.584777 (-3.882820) | 3.830742 / 3.745712 (0.085030) | 2.170444 / 5.269862 (-3.099418) | 1.345097 / 4.565676 (-3.220580) | 0.086661 / 0.424275 (-0.337614) | 0.013073 / 0.007607 (0.005466) | 0.519150 / 0.226044 (0.293106) | 5.193447 / 2.268929 (2.924518) | 2.391155 / 55.444624 (-53.053470) | 2.076610 / 6.876477 (-4.799867) | 2.245557 / 2.142072 (0.103484) | 0.846496 / 4.805227 (-3.958731) | 0.169246 / 6.500664 (-6.331418) | 0.066360 / 0.075469 (-0.009109) |\n\n### Benchmark: benchmark_map_filter.json\n\n| metric | filter | map fast-tokenizer batched | map identity | map identity batched | map no-op batched | map no-op batched numpy | map no-op batched pandas | map no-op batched pytorch | map no-op batched tensorflow |\n|--------|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 1.196344 / 1.841788 (-0.645444) | 15.640363 / 8.074308 (7.566055) | 14.936144 / 10.191392 (4.744752) | 0.163613 / 0.680424 (-0.516811) | 0.017900 / 0.534201 (-0.516301) | 0.425377 / 0.579283 (-0.153906) | 0.431119 / 0.434364 (-0.003245) | 0.513669 / 0.540337 (-0.026669) | 0.592970 / 1.386936 (-0.793966) |\n\n</details>\nPyArrow==latest\n\n<details>\n<summary>Show updated benchmarks!</summary>\n\n### Benchmark: benchmark_array_xd.json\n\n| metric | read_batch_formatted_as_numpy after write_array2d | read_batch_formatted_as_numpy after write_flattened_sequence | read_batch_formatted_as_numpy after write_nested_sequence | read_batch_unformated after write_array2d | read_batch_unformated after write_flattened_sequence | read_batch_unformated after write_nested_sequence | read_col_formatted_as_numpy after write_array2d | read_col_formatted_as_numpy after write_flattened_sequence | read_col_formatted_as_numpy after write_nested_sequence | read_col_unformated after write_array2d | read_col_unformated after write_flattened_sequence | read_col_unformated after write_nested_sequence | read_formatted_as_numpy after write_array2d | read_formatted_as_numpy after write_flattened_sequence | read_formatted_as_numpy after write_nested_sequence | read_unformated after write_array2d | read_unformated after write_flattened_sequence | read_unformated after write_nested_sequence | write_array2d | write_flattened_sequence | write_nested_sequence |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.007958 / 0.011353 (-0.003395) | 0.005707 / 0.011008 (-0.005301) | 0.075377 / 0.038508 (0.036869) | 0.037126 / 0.023109 (0.014016) | 0.344589 / 0.275898 (0.068691) | 0.381060 / 0.323480 (0.057580) | 0.006592 / 0.007986 (-0.001393) | 0.004479 / 0.004328 (0.000151) | 0.074456 / 0.004250 (0.070206) | 0.054087 / 0.037052 (0.017035) | 0.344942 / 0.258489 (0.086453) | 0.393174 / 0.293841 (0.099333) | 0.037926 / 0.128546 (-0.090620) | 0.012638 / 0.075646 (-0.063009) | 0.087743 / 0.419271 (-0.331529) | 0.050081 / 0.043533 (0.006548) | 0.340406 / 0.255139 (0.085267) | 0.361487 / 0.283200 (0.078287) | 0.108546 / 0.141683 (-0.033137) | 1.424626 / 1.452155 (-0.027529) | 1.553958 / 1.492716 (0.061242) |\n\n### Benchmark: benchmark_getitem\\_100B.json\n\n| metric | get_batch_of\\_1024\\_random_rows | get_batch_of\\_1024\\_rows | get_first_row | get_last_row |\n|--------|---|---|---|---|\n| new / old (diff) | 0.329922 / 0.018006 (0.311916) | 0.523239 / 0.000490 (0.522749) | 0.012164 / 0.000200 (0.011964) | 0.000137 / 0.000054 (0.000082) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.031935 / 0.037411 (-0.005477) | 0.115680 / 0.014526 (0.101154) | 0.130062 / 0.176557 (-0.046494) | 0.180679 / 0.737135 (-0.556457) | 0.135548 / 0.296338 (-0.160790) |\n\n### Benchmark: benchmark_iterating.json\n\n| metric | read 5000 | read 50000 | read_batch 50000 10 | read_batch 50000 100 | read_batch 50000 1000 | read_formatted numpy 5000 | read_formatted pandas 5000 | read_formatted tensorflow 5000 | read_formatted torch 5000 | read_formatted_batch numpy 5000 10 | read_formatted_batch numpy 5000 1000 | shuffled read 5000 | shuffled read 50000 | shuffled read_batch 50000 10 | shuffled read_batch 50000 100 | shuffled read_batch 50000 1000 | shuffled read_formatted numpy 5000 | shuffled read_formatted_batch numpy 5000 10 | shuffled read_formatted_batch numpy 5000 1000 |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.429648 / 0.215209 (0.214439) | 4.303342 / 2.077655 (2.225687) | 1.999395 / 1.504120 (0.495275) | 1.810354 / 1.541195 (0.269160) | 1.963132 / 1.468490 (0.494642) | 0.701654 / 4.584777 (-3.883122) | 3.844687 / 3.745712 (0.098975) | 2.153425 / 5.269862 (-3.116436) | 1.351541 / 4.565676 (-3.214135) | 0.086292 / 0.424275 (-0.337983) | 0.012491 / 0.007607 (0.004883) | 0.523144 / 0.226044 (0.297099) | 5.243283 / 2.268929 (2.974355) | 2.465849 / 55.444624 (-52.978775) | 2.154505 / 6.876477 (-4.721972) | 2.245500 / 2.142072 (0.103428) | 0.838902 / 4.805227 (-3.966326) | 0.169441 / 6.500664 (-6.331223) | 0.065631 / 0.075469 (-0.009838) |\n\n### Benchmark: benchmark_map_filter.json\n\n| metric | filter | map fast-tokenizer batched | map identity | map identity batched | map no-op batched | map no-op batched numpy | map no-op batched pandas | map no-op batched pytorch | map no-op batched tensorflow |\n|--------|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 1.262175 / 1.841788 (-0.579612) | 15.424650 / 8.074308 (7.350342) | 15.000718 / 10.191392 (4.809326) | 0.186328 / 0.680424 (-0.494096) | 0.018076 / 0.534201 (-0.516125) | 0.433458 / 0.579283 (-0.145825) | 0.424213 / 0.434364 (-0.010151) | 0.546568 / 0.540337 (0.006231) | 0.643529 / 1.386936 (-0.743407) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#ea7298bf121d7ae8079f0a59deb67c2fa1d4df6a \"CML watermark\")\n" ]

[ { "color": "d73a4a", "default": true, "description": "Something isn't working", "id": 1935892857, "name": "bug", "node_id": "MDU6TGFiZWwxOTM1ODkyODU3", "url": "https://api.github.com/repos/huggingface/datasets/labels/bug" } ]

[]

[ { "color": "a2eeef", "default": true, "description": "New feature or request", "id": 1935892871, "name": "enhancement", "node_id": "MDU6TGFiZWwxOTM1ODkyODcx", "url": "https://api.github.com/repos/huggingface/datasets/labels/enhancement" } ]

[ "Hi ! We use the Arrow columnar format under the hood, which doesn't support such dictionaries: each field must have a fixed type and exist in each sample.\r\n\r\nInstead you can restructure your data like\r\n```\r\n{\r\n \"index\": 0,\r\n \"keys\": [\"2 * x + y >= 3\"],\r\n \"values\": [[\"2 * x + y >= 3\", \"4 * x + 2 * y >= 6\"]],\r\n }\r\n},\r\n...\r\n{\r\n \"index\": 9999,\r\n \"keys\": [\"x >= 6\"],\r\n \"values\": [[\"x >= 6\", \"x >= 0\", \"x >= -1\"]],\r\n},\r\n...\r\n```" ]

[]

[ "Thanks for the review. \r\nCheers!", "Hi @hfawaz, this week we are doing the 🤗 `datasets` sprint (see some details [here](https://discuss.huggingface.co/t/open-to-the-community-one-week-team-effort-to-reach-v2-0-of-hf-datasets-library/2176)).\r\n\r\nNothing more to do on your side but it means that if you register on the thread I linked above, you can have some goodies for the present dataset that you have already added (and a special goodie if you want to spend more time and add 2 other datasets as well).\r\n\r\nIf you want to join, just tell me (or post on the thread on the HuggingFace forum: https://discuss.huggingface.co/t/open-to-the-community-one-week-team-effort-to-reach-v2-0-of-hf-datasets-library/2176)", "Hello @thomwolf \r\nThanks for the feedback and for this invitation, indeed I would be glad to join you guys (you can add me). \r\nI will see if I have the time to implement a couple of datasets. \r\nCheers! ", "@hfawaz invited you to the slack with your uha email.\r\n\r\nCheck your spam folder if you can't find the invitation :)", "Oh thanks, but can you invite me on my gmail: [email protected] \r\nUHA is my old organization, I haven't had the time to update my online profiles yet.\r\nThank you " ]

[]

[]

[ { "color": "a2eeef", "default": true, "description": "New feature or request", "id": 1935892871, "name": "enhancement", "node_id": "MDU6TGFiZWwxOTM1ODkyODcx", "url": "https://api.github.com/repos/huggingface/datasets/labels/enhancement" } ]