Datasets:

---

Francesco-A

/

github-issues_huggingface-datasets

like 0

[ { "color": "0e8a16", "default": false, "description": "Contribution to a dataset script", "id": 4564477500, "name": "dataset contribution", "node_id": "LA_kwDODunzps8AAAABEBBmPA", "url": "https://api.github.com/repos/huggingface/datasets/labels/dataset%20contribution" } ]

[ "Hi @albertvillanova we've made a major overhaul of the loading script including all configurations we're making available. Could you please review it again?", "@albertvillanova :ping_pong: ", "Thanks @lhbonifacio for adding this dataset.\r\nHi there, i got an error about mmarco:\r\nConnectionError: Couldn't reach 'unicamp-dl/mmarco' on the Hub (ConnectionError)\r\ncode:\r\n`from datasets import list_datasets, load_dataset\r\ndataset = load_dataset('unicamp-dl/mmarco', language='portuguese')`\r\n\r\nAny help will be appreciated!", "Hi @catqaq, we updated the loading script. Now you can load the datasets with:\r\n\r\n```python\r\ndataset = load_dataset('unicamp-dl/mmarco', 'portuguese')\r\n```\r\n\r\nYou can check the list of supported languages and usage examples in [this link](https://huggingface.co/datasets/unicamp-dl/mmarco). Feel free to contact us if you have any issues.", "\r\n\r\n\r\n> \r\n\r\n\r\n\r\n> Hi @catqaq, we updated the loading script. Now you can load the datasets with:\r\n> \r\n> ```python\r\n> dataset = load_dataset('unicamp-dl/mmarco', 'portuguese')\r\n> ```\r\n> \r\n> You can check the list of supported languages and usage examples in [this link](https://huggingface.co/datasets/unicamp-dl/mmarco). Feel free to contact us if you have any issues.\r\n\r\nThanks for your quick updates. So, how can i get the fixed version, install from the source? It seems that the merging is blocked.", "@catqaq you can load mMARCO using the namespace `unicamp-dl/mmarco` while this PR remains under review.", "Thanks for your contribution, @lhbonifacio and @hugoabonizio. And sorry for the late response.\r\n\r\nWe are removing the dataset scripts from this GitHub repo and moving them to the Hugging Face Hub: https://huggingface.co/datasets\r\n\r\nAs you already created this dataset under your organization namespace (https://huggingface.co/datasets/unicamp-dl/mmarco), I think we can safely close this PR.\r\n\r\nWe would suggest you complete your dataset card with the YAML tags, to make it searchable and discoverable.\r\n\r\nPlease, feel free to tell us if you need some 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 ! Indeed it says \"downloading and preparing\" but in your case it didn't need to download anything since you used local files (it would have thrown an error otherwise). I think we can improve the logging to make it clearer in this case", "@lhoestq Thank you for explaining. I am sorry but I was not clear about my intention. I didn't want to kill internet traffic; I wanted to kill all write activity. In other words, you can imagine that my storage has only read access but crashes on write.\r\n\r\nWhen run_clm.py is invoked with the same parameters, the hash in the cache directory \"datacache/trainpy.v2/json/default-471372bed4b51b53/0.0.0/...\" doesn't change, and my job can load cached data properly. This is great.\r\n\r\nUnfortunately, when params change (which happens sometimes), the hash changes and the old cache is invalid. datasets builder would create a new cache directory with the new hash and create JSON builder there, even though every JSON builder is the same. I didn't find a way to avoid such behavior.\r\n\r\nThis problem can be resolved when using datasets.map() for tokenizing and grouping text. This function allows me to specify output filenames with --cache_file_names, so that the cached files are always valid.\r\n\r\nThis is the code that I used to freeze cache filenames for tokenization. I wish I could do the same to datasets.load_dataset()\r\n```\r\n tokenized_datasets = raw_datasets.map(\r\n tokenize_function,\r\n batched=True,\r\n num_proc=data_args.preprocessing_num_workers,\r\n remove_columns=column_names,\r\n load_from_cache_file=not data_args.overwrite_cache,\r\n desc=\"Running tokenizer on dataset\",\r\n cache_file_names={k: os.path.join(model_args.cache_dir, f'{k}-tokenized') for k in raw_datasets},\r\n )\r\n```", "Hi ! `load_dataset` may re-generate your dataset if some parameters changed indeed. If you want to freeze a dataset loaded with `load_dataset`, I think the best solution is just to save it somewhere on your disk with `.save_to_disk(my_dataset_dir)` and reload it with `load_from_disk(my_dataset_dir)`. This way you will be able to reload the dataset without having to run `load_dataset`" ]

[ { "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 ! It looks like an issue with pytorch.\r\n\r\nCould you try to run `import torch` and see if it raises an error ?", "> Hi ! It looks like an issue with pytorch.\r\n> \r\n> Could you try to run `import torch` and see if it raises an error ?\r\n\r\nIt works. Thank you!" ]

[]

[]

[]

[ "Hi! \r\n\r\nThis would only work if `bigscience/xP3` was a no-code dataset, but it isn't (it has a Python builder script).\r\n\r\nBut this should work: \r\n```python\r\nload_dataset(\"json\", data_files=\"https://huggingface.co/datasets/bigscience/xP3/resolve/main/ur/xp3_facebook_flores_spa_Latn-urd_Arab_devtest_ab-spa_Latn-urd_Arab.jsonl\")\r\n```\r\n\r\n", "I see, it's not compatible w/ regex right?\r\ne.g.\r\n`load_dataset(\"json\", data_files=\"https://huggingface.co/datasets/bigscience/xP3/resolve/main/ur/*\")`", "> I see, it's not compatible w/ regex right? e.g. `load_dataset(\"json\", data_files=\"https://huggingface.co/datasets/bigscience/xP3/resolve/main/ur/*\")`\r\n\r\nIt should work for patterns that \"reference\" the local filesystem, but to make this work with the Hub, we must implement https://github.com/huggingface/datasets/issues/5281 first.\r\n\r\nIn the meantime, you can fetch these glob files with `HfFileSystem` and pass them as a list to `load_dataset`:\r\n```python\r\nfrom datasets import load_dataset\r\nfrom huggingface_hub import HfFileSystem, hf_hub_url # `HfFileSystem` requires the latest version of `huggingface_hub`\r\n\r\nfs = HfFileSystem()\r\nglob_files = fs.glob(\"datasets/bigscience/xP3/ur/*\")\r\n# convert fsspec URLs to HTTP URLs\r\nresolved_paths = [fs.resolve_path(file) for file in glob_files]\r\ndata_files = [hf_hub_url(resolved_path.repo_id, resolved_path.path_in_repo, repo_type=resolved_path.repo_type) for resolved_path in resolved_paths]\r\n\r\nds = load_dataset(\"json\", data_files=data_files)\r\n```" ]

[]

[]

[]

[ "Hello. All aforementioned changes have been made. I've also re-run black on miam.py. :-)", "I will run isort again. Hopefully it resolves the current check_code_quality test failure.", "Once the review period is over, feel free to open a PR to add all the missing information ;)", "Hi! I will follow up right now with one more pull request as I have new anonymous citation information to include." ]

[]

[ "Seems like it works when setting ```block_size=2100000000``` or something arbitrarily large though.", "Can you give us some stats on the data files you use as inputs?", "Basically ~600MB txt files(UTF-8) * 59. \r\ncontents like ```안녕하세요, 이것은 예제로 한번 말해보는 텍스트입니다. 그냥 이렇다고요.<|endoftext|>\\n```\r\n\r\nAlso, it gets stuck for a loooong time at ```Testing the mapped function outputs```, for more than 12 hours(currently ongoing)", "It gets stuck while doing `.map()` ? Are you using multiprocessing ?\r\nIf you could provide a code snippet it could be very useful", "From transformers/examples/language-modeling/run-language-modeling.py :\r\n```\r\ndef get_dataset(\r\n args: DataTrainingArguments,\r\n tokenizer: PreTrainedTokenizer,\r\n evaluate: bool = False,\r\n cache_dir: Optional[str] = None,\r\n):\r\n file_path = args.eval_data_file if evaluate else args.train_data_file\r\n if True:\r\n dataset = load_dataset(\"text\", data_files=glob.glob(file_path), split='train', use_threads=True, \r\n ignore_verifications=True, save_infos=True, block_size=104857600)\r\n dataset = dataset.map(lambda ex: tokenizer(ex[\"text\"], add_special_tokens=True,\r\n truncation=True, max_length=args.block_size), batched=True)\r\n dataset.set_format(type='torch', columns=['input_ids'])\r\n return dataset\r\n if args.line_by_line:\r\n return LineByLineTextDataset(tokenizer=tokenizer, file_path=file_path, block_size=args.block_size)\r\n else:\r\n return TextDataset(\r\n tokenizer=tokenizer,\r\n file_path=file_path,\r\n block_size=args.block_size,\r\n overwrite_cache=args.overwrite_cache,\r\n cache_dir=cache_dir,\r\n )\r\n```\r\n\r\nNo, I'm not using multiprocessing.", "I am not able to reproduce on my side :/\r\n\r\nCould you send the version of `datasets` and `pyarrow` you're using ?\r\nCould you try to update the lib and try again ?\r\nOr do you think you could try to reproduce it on google colab ?", "Huh, weird. It's fixed on my side too.\r\nBut now ```Caching processed dataset``` is taking forever - how can I disable it? Any flags?", "Right after `Caching processed dataset`, your function is applied to the dataset and there's a progress bar that shows how much time is left. How much time does it take for you ?\r\n\r\nAlso caching isn't supposed to slow down your processing. But if you still want to disable it you can do `.map(..., load_from_cache_file=False)`", "Ah, it’s much faster now(Takes around 15~20min). \r\nBTW, any way to set default tensor output as plain tensors with distributed training? The ragged tensors are incompatible with tpustrategy :(", "> Ah, it’s much faster now(Takes around 15~20min).\r\n\r\nGlad to see that it's faster now. What did you change exactly ?\r\n\r\n> BTW, any way to set default tensor output as plain tensors with distributed training? The ragged tensors are incompatible with tpustrategy :(\r\n\r\nOh I didn't know about that. Feel free to open an issue to mention that.\r\nI guess what you can do for now is set the dataset format to numpy instead of tensorflow, and use a wrapper of the dataset that converts the numpy arrays to tf tensors.\r\n\r\n", ">>> Glad to see that it's faster now. What did you change exactly ?\r\nI don't know, it just worked...? Sorry I couldn't be more helpful.\r\n\r\nSetting with numpy array is a great idea! Thanks." ]

[]

[ "Hi,\r\n\r\nif you inspect the raw data, you can find there are 946 occurrences of `-DOCSTART- -X- -X- O` in the train split and `14041 + 946 = 14987`, which is exactly the number of sentences the authors report. `-DOCSTART-` is a special line that acts as a boundary between two different documents and is filtered out in our implementation.\r\n\r\n@lhoestq What do you think about including these lines? ([Link](https://github.com/flairNLP/flair/issues/1097) to a similar issue in the flairNLP repo)", "We should mention in the Conll2003 dataset card that these lines have been removed indeed.\r\n\r\nIf some users are interested in using these lines (maybe to recombine documents ?) then we can add a parameter to the conll2003 dataset to include them.\r\n\r\nBut IMO the default config should stay the current one (without the `-DOCSTART-` stuff), so that you can directly train NER models without additional preprocessing. Let me know what you think", "@lhoestq Yes, I agree adding a small note should be sufficient.\r\n\r\nCurrently, NLTK's `ConllCorpusReader` ignores the `-DOCSTART-` lines so I think it's ok if we do the same. If there is an interest in the future to use these lines, then we can include them.", "I added a mention of this in conll2003's dataset card:\r\nhttps://github.com/huggingface/datasets/blob/fc9796920da88486c3b97690969aabf03d6b4088/datasets/conll2003/README.md#conll2003\r\n\r\nEdit: just saw your PR @mariosasko (noticed it too late ^^)\r\nLet me take a look at it :)" ]

[]

[ "_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) | 0.009139 / 0.011353 (-0.002214) | 0.005634 / 0.011008 (-0.005374) | 0.129587 / 0.038508 (0.091079) | 0.038298 / 0.023109 (0.015189) | 0.428149 / 0.275898 (0.152251) | 0.443744 / 0.323480 (0.120264) | 0.007501 / 0.007986 (-0.000485) | 0.005999 / 0.004328 (0.001671) | 0.100796 / 0.004250 (0.096546) | 0.053236 / 0.037052 (0.016184) | 0.423868 / 0.258489 (0.165379) | 0.460110 / 0.293841 (0.166269) | 0.041255 / 0.128546 (-0.087291) | 0.013790 / 0.075646 (-0.061856) | 0.438398 / 0.419271 (0.019127) | 0.063086 / 0.043533 (0.019553) | 0.414826 / 0.255139 (0.159687) | 0.460652 / 0.283200 (0.177453) | 0.121223 / 0.141683 (-0.020460) | 1.754430 / 1.452155 (0.302275) | 1.900037 / 1.492716 (0.407320) |\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) | 0.027222 / 0.018006 (0.009216) | 0.617666 / 0.000490 (0.617176) | 0.022443 / 0.000200 (0.022243) | 0.000820 / 0.000054 (0.000766) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.030397 / 0.037411 (-0.007014) | 0.125732 / 0.014526 (0.111206) | 0.149805 / 0.176557 (-0.026752) | 0.234048 / 0.737135 (-0.503087) | 0.143108 / 0.296338 (-0.153231) |\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) | 0.631189 / 0.215209 (0.415980) | 6.182871 / 2.077655 (4.105216) | 2.635730 / 1.504120 (1.131610) | 2.231429 / 1.541195 (0.690235) | 2.438360 / 1.468490 (0.969870) | 0.861170 / 4.584777 (-3.723607) | 5.785984 / 3.745712 (2.040272) | 2.758358 / 5.269862 (-2.511504) | 1.678095 / 4.565676 (-2.887582) | 0.105961 / 0.424275 (-0.318314) | 0.013659 / 0.007607 (0.006052) | 0.762943 / 0.226044 (0.536898) | 7.774399 / 2.268929 (5.505471) | 3.319027 / 55.444624 (-52.125598) | 2.700248 / 6.876477 (-4.176229) | 3.008581 / 2.142072 (0.866509) | 1.122522 / 4.805227 (-3.682705) | 0.214832 / 6.500664 (-6.285832) | 0.085281 / 0.075469 (0.009811) |\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) | 1.647610 / 1.841788 (-0.194177) | 18.178316 / 8.074308 (10.104008) | 21.199177 / 10.191392 (11.007785) | 0.247063 / 0.680424 (-0.433361) | 0.030443 / 0.534201 (-0.503758) | 0.512527 / 0.579283 (-0.066757) | 0.640758 / 0.434364 (0.206394) | 0.639986 / 0.540337 (0.099649) | 0.760113 / 1.386936 (-0.626823) |\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) | 0.008293 / 0.011353 (-0.003060) | 0.005360 / 0.011008 (-0.005648) | 0.102932 / 0.038508 (0.064424) | 0.037457 / 0.023109 (0.014347) | 0.444114 / 0.275898 (0.168216) | 0.512855 / 0.323480 (0.189375) | 0.007030 / 0.007986 (-0.000956) | 0.004954 / 0.004328 (0.000625) | 0.095757 / 0.004250 (0.091507) | 0.051239 / 0.037052 (0.014187) | 0.471118 / 0.258489 (0.212629) | 0.517764 / 0.293841 (0.223923) | 0.041953 / 0.128546 (-0.086593) | 0.013748 / 0.075646 (-0.061898) | 0.118089 / 0.419271 (-0.301182) | 0.060159 / 0.043533 (0.016626) | 0.466011 / 0.255139 (0.210872) | 0.489180 / 0.283200 (0.205980) | 0.123250 / 0.141683 (-0.018433) | 1.714738 / 1.452155 (0.262584) | 1.838571 / 1.492716 (0.345855) |\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) | 0.267792 / 0.018006 (0.249785) | 0.624313 / 0.000490 (0.623824) | 0.007315 / 0.000200 (0.007115) | 0.000136 / 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) | 0.033751 / 0.037411 (-0.003661) | 0.122819 / 0.014526 (0.108293) | 0.148270 / 0.176557 (-0.028286) | 0.198581 / 0.737135 (-0.538554) | 0.144845 / 0.296338 (-0.151494) |\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) | 0.620631 / 0.215209 (0.405422) | 6.224665 / 2.077655 (4.147010) | 2.856592 / 1.504120 (1.352473) | 2.525089 / 1.541195 (0.983894) | 2.600198 / 1.468490 (1.131708) | 0.872038 / 4.584777 (-3.712739) | 5.571650 / 3.745712 (1.825937) | 5.907643 / 5.269862 (0.637782) | 2.348770 / 4.565676 (-2.216906) | 0.111665 / 0.424275 (-0.312610) | 0.013886 / 0.007607 (0.006278) | 0.762154 / 0.226044 (0.536109) | 7.792686 / 2.268929 (5.523758) | 3.601122 / 55.444624 (-51.843503) | 2.939412 / 6.876477 (-3.937064) | 2.973430 / 2.142072 (0.831358) | 1.065016 / 4.805227 (-3.740211) | 0.221701 / 6.500664 (-6.278963) | 0.088157 / 0.075469 (0.012688) |\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) | 1.771061 / 1.841788 (-0.070727) | 18.826926 / 8.074308 (10.752618) | 21.283830 / 10.191392 (11.092438) | 0.239233 / 0.680424 (-0.441191) | 0.026159 / 0.534201 (-0.508042) | 0.487074 / 0.579283 (-0.092209) | 0.623241 / 0.434364 (0.188877) | 0.600506 / 0.540337 (0.060169) | 0.691271 / 1.386936 (-0.695665) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#1bbe2c3496498a6415765b517ac4bc600a02ad06 \"CML watermark\")\n" ]

[ { "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" }, { "color": "bfdadc", "default": false, "description": "Vision datasets", "id": 3608941089, "name": "vision", "node_id": "LA_kwDODunzps7XHBIh", "url": "https://api.github.com/repos/huggingface/datasets/labels/vision" } ]

[ "Hello! I have recently been involved with a medical image segmentation project myself and was going through the `The Medical Segmentation Decathlon Dataset` as well. \r\nI haven't yet had experience adding datasets to this repository yet but would love to get started. Should I take this issue?\r\nIf yes, I've got two questions -\r\n1. There are 10 different datasets available, so are all datasets to be added in a single PR, or one at a time? \r\n2. Since it's a competition, masks for the test-set are not available. How is that to be tackled? Sorry if it's a silly question, I have recently started exploring `datasets`.", "Hi! Sure, feel free to take this issue. You can self-assign the issue by commenting `#self-assign`.\r\n\r\nTo answer your questions:\r\n1. It makes the most sense to add each one as a separate config, so one dataset script with 10 configs in a single PR.\r\n2. Just set masks in the test set to `None`.\r\n\r\nNote that the images/masks in this dataset are in NIfTI format, which our `Image` feature currently doesn't support, so I think it's best to yield the paths to the images/masks in the script and add a preprocessing section to the card where we explain how to load/process the images/masks with `nibabel` (I can help with that). \r\n\r\n", "> Note that the images/masks in this dataset are in NIfTI format, which our `Image` feature currently doesn't support, so I think it's best to yield the paths to the images/masks in the script and add a preprocessing section to the card where we explain how to load/process the images/masks with `nibabel` (I can help with that).\r\n\r\nGotcha, thanks. Will start working on the issue and let you know in case of any doubt.", "#self-assign", "This is great! There is a first model on the HUb that uses this dataset! https://huggingface.co/MONAI/example_spleen_segmentation" ]

[ { "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" } ]

[ "Done! cf #1117 and #1022" ]

[ { "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" } ]

[ "This is an interesting idea !\r\nDo you have ideas about how to approach the decoding and the normalization ?", "@lhoestq\r\nI think we first need to save the transformation chain to a list in `self._fingerprint`. Then we can\r\n- decode all the current saved datasets to see if there is already one that is equivalent to the transformation we need now.\r\n- or, calculate all the possible hash value of the current chain for comparison so that we could continue to use hashing.\r\nIf we find one, we can adjust the list in `self._fingerprint` to it.\r\n\r\nAs for the transformation reordering rules, we can just start with some manual rules, like two sort on the same column should merge to one, filter and select can change orders.\r\n\r\nAnd for encoding and decoding, we can just manually specify `sort` is 0, `shuffling` is 2 and create a base-n number or use some general algorithm like `base64.urlsafe_b64encode`.\r\n\r\nBecause we are not doing lazy evaluation now, we may not be able to normalize the transformation to its minimal form. If we want to support that, we can provde a `Sequential` api and let user input a list or transformation, so that user would not use the intermediate datasets. This would look like tf.data.Dataset." ]

[ { "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 @vpj, thanks for reporting.\r\n\r\nI'm sorry, but I can't reproduce your bug: I load \"20220301.simple\"in 9 seconds:\r\n```shell\r\ntime python -c \"from datasets import load_dataset; load_dataset('wikipedia', '20220301.simple')\"\r\n\r\nDownloading and preparing dataset wikipedia/20220301.simple (download: 228.58 MiB, generated: 224.18 MiB, post-processed: Unknown size, total: 452.76 MiB) to .../.cache/huggingface/datasets/wikipedia/20220301.simple/2.0.0/aa542ed919df55cc5d3347f42dd4521d05ca68751f50dbc32bae2a7f1e167559...\r\nDownloading: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1.66k/1.66k [00:00<00:00, 1.02MB/s]\r\nDownloading: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 235M/235M [00:02<00:00, 82.8MB/s]\r\nDataset wikipedia downloaded and prepared to .../.cache/huggingface/datasets/wikipedia/20220301.simple/2.0.0/aa542ed919df55cc5d3347f42dd4521d05ca68751f50dbc32bae2a7f1e167559. Subsequent calls will reuse this data.\r\n100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 290.75it/s]\r\n\r\nreal\t0m9.693s\r\nuser\t0m6.002s\r\nsys\t0m3.260s\r\n```\r\n\r\nCould you please check your environment info, as requested when opening this issue?\r\n```\r\n## Environment info\r\n<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->\r\n- `datasets` version:\r\n- Platform:\r\n- Python version:\r\n- PyArrow version:\r\n```\r\nMaybe you are using an old version of `datasets`...", "Downloading and processing `wikipedia simple` dataset completed in under 11sec on M1 Mac. Could you please check `dataset` version as mentioned by @albertvillanova? Also check system specs, if system is under load processing could take some time I guess." ]

[]

[ "Looks like we need to trim the ones with too many configs, will look into it tomorrow!" ]

[ { "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" } ]

[ "Thanks for reporting this second issue as well. We definitely want to make streaming datasets fully working in a distributed setup and with the best performance. Right now it only supports single process.\r\n\r\nIn this issue it seems that the streaming capabilities that we offer to dataset builders are not transferred to the forked process (so it fails to open remote files and start streaming data from them). In particular `open` is supposed to be mocked by our `xopen` function that is an extended open that supports remote files. Let me try to fix this" ]

[]

[]

[ { "color": "0e8a16", "default": false, "description": "Contribution to a dataset script", "id": 4564477500, "name": "dataset contribution", "node_id": "LA_kwDODunzps8AAAABEBBmPA", "url": "https://api.github.com/repos/huggingface/datasets/labels/dataset%20contribution" } ]

[ "_The documentation is not available anymore as the PR was closed or merged._", "As discussed with @lewtun, we are closing this PR, because it requires first the task names to be aligned between AutoTrain and datasets." ]

[]

[]

[ { "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 @dconathan, thanks for reporting.\r\n\r\nWe rely on Arrow as a backend, and as far as I know currently support for `float16` in Arrow is not fully implemented in Python (C++), hence the `ArrowNotImplementedError` you get.\r\n\r\nSee, e.g.: https://arrow.apache.org/docs/status.html?highlight=float16#data-types", "Thanks for the link…. didn’t realize arrow didn’t support it yet. Should it be removed from https://huggingface.co/docs/datasets/v2.4.0/en/package_reference/main_classes#datasets.Value until Arrow supports it?", "Yes, you are right: maybe we should either remove it from our docs or add a comment explaining the issue.\r\n\r\nThe thing is that in Arrow it is partially supported: you can create `float16` values, but you can't cast them from/to other types. And current implementation of `Value` always tries to perform a cast from `float64` to `float16`.", "Maybe we can just add a note in the `Value` documentation ?", "Would you accept a PR to fix this? @lhoestq Do you have an idea of how hard it would be to fix?", "I think the issue comes mostly from pyarrow not supporting `float16` completely.\r\n\r\nFor example you stil can't cast from/to `float16`\r\n```python\r\nimport numpy as np\r\nimport pyarrow as pa\r\n\r\npa.array(range(5)).cast(pa.float16())\r\n# ArrowNotImplementedError: Unsupported cast from int64 to halffloat using function cast_half_float\r\npa.array(range(5), pa.float32()).cast(pa.float16())\r\n# ArrowNotImplementedError: Unsupported cast from float to halffloat using function cast_half_float\r\npa.array(range(5), pa.float16())\r\n# ArrowTypeError: Expected np.float16 instance\r\npa.array(np.arange(5, dtype=np.float16())).cast(pa.float32())\r\n# ArrowNotImplementedError: Unsupported cast from halffloat to float using function cast_float\r\n```", "Hmm it seems like we can either:\r\n1. try to fix pyarrow upstream\r\n2. half-support float16 with some workaround to make sure we don't ever do casting internally\r\n" ]

[ { "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" }, { "color": "fef2c0", "default": false, "description": "", "id": 3287858981, "name": "streaming", "node_id": "MDU6TGFiZWwzMjg3ODU4OTgx", "url": "https://api.github.com/repos/huggingface/datasets/labels/streaming" } ]

[ "#self-assign", "@alvarobartt I've assigned you the issue since I'm not actively working on it.", "Cool thanks @mariosasko I'll try to fix it in the upcoming days, thanks!", "@lhoestq so in order to address what’s not completed in this issue, do you think it makes sense to add a param `features` to `IterableDataset.map` so that the output features right after the `map` are defined there? ", "Yes that would be ideal IMO, thanks again for the help :)", "@lhoestq cool then if you agree I can work on that! I’ll also update the docs accordingly once done, thanks!", "I've already started with a PR as a draft @lhoestq, should we also try to look for a way to explicitly request pre-fetching right after a map operation is applied, so that the features are inferred if the user says explicitly so? Thanks!", "> should we also try to look for a way to explicitly request pre-fetching right after a map operation is applied, so that the features are inferred if the user says explicitly so?\r\n\r\nRight now one can use `ds = ds._resolve_features()` do to so. It can be used after `map` or `load_dataset` if the features are not known. Maybe we can make this method public ?" ]

[]

[]

[]

[ "_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) | 0.005855 / 0.011353 (-0.005498) | 0.003556 / 0.011008 (-0.007452) | 0.079430 / 0.038508 (0.040922) | 0.056754 / 0.023109 (0.033645) | 0.311718 / 0.275898 (0.035820) | 0.346731 / 0.323480 (0.023251) | 0.004414 / 0.007986 (-0.003571) | 0.002835 / 0.004328 (-0.001493) | 0.062138 / 0.004250 (0.057888) | 0.044259 / 0.037052 (0.007206) | 0.314681 / 0.258489 (0.056192) | 0.359802 / 0.293841 (0.065961) | 0.026684 / 0.128546 (-0.101862) | 0.008023 / 0.075646 (-0.067623) | 0.260148 / 0.419271 (-0.159123) | 0.043734 / 0.043533 (0.000202) | 0.312081 / 0.255139 (0.056942) | 0.340004 / 0.283200 (0.056805) | 0.019559 / 0.141683 (-0.122124) | 1.488758 / 1.452155 (0.036604) | 1.510828 / 1.492716 (0.018111) |\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) | 0.181376 / 0.018006 (0.163370) | 0.441726 / 0.000490 (0.441236) | 0.001722 / 0.000200 (0.001522) | 0.000066 / 0.000054 (0.000012) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.023760 / 0.037411 (-0.013651) | 0.071847 / 0.014526 (0.057321) | 0.082642 / 0.176557 (-0.093915) | 0.145555 / 0.737135 (-0.591580) | 0.084554 / 0.296338 (-0.211784) |\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) | 0.401688 / 0.215209 (0.186479) | 4.000994 / 2.077655 (1.923339) | 2.047109 / 1.504120 (0.542989) | 1.891874 / 1.541195 (0.350679) | 1.970599 / 1.468490 (0.502109) | 0.500646 / 4.584777 (-4.084131) | 3.006623 / 3.745712 (-0.739089) | 4.248359 / 5.269862 (-1.021503) | 2.613946 / 4.565676 (-1.951730) | 0.057921 / 0.424275 (-0.366354) | 0.006407 / 0.007607 (-0.001200) | 0.470676 / 0.226044 (0.244631) | 4.722280 / 2.268929 (2.453352) | 2.448530 / 55.444624 (-52.996095) | 2.175841 / 6.876477 (-4.700635) | 2.352287 / 2.142072 (0.210214) | 0.589049 / 4.805227 (-4.216179) | 0.125145 / 6.500664 (-6.375519) | 0.060829 / 0.075469 (-0.014640) |\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) | 1.189225 / 1.841788 (-0.652563) | 16.753085 / 8.074308 (8.678777) | 13.086512 / 10.191392 (2.895120) | 0.132371 / 0.680424 (-0.548052) | 0.016933 / 0.534201 (-0.517268) | 0.328258 / 0.579283 (-0.251025) | 0.344074 / 0.434364 (-0.090290) | 0.374042 / 0.540337 (-0.166296) | 0.515307 / 1.386936 (-0.871629) |\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) | 0.005963 / 0.011353 (-0.005390) | 0.003484 / 0.011008 (-0.007525) | 0.062618 / 0.038508 (0.024110) | 0.057217 / 0.023109 (0.034108) | 0.426760 / 0.275898 (0.150862) | 0.464422 / 0.323480 (0.140942) | 0.005276 / 0.007986 (-0.002709) | 0.002872 / 0.004328 (-0.001456) | 0.062636 / 0.004250 (0.058385) | 0.045953 / 0.037052 (0.008900) | 0.433221 / 0.258489 (0.174732) | 0.475087 / 0.293841 (0.181246) | 0.027217 / 0.128546 (-0.101329) | 0.007965 / 0.075646 (-0.067681) | 0.067749 / 0.419271 (-0.351522) | 0.041235 / 0.043533 (-0.002298) | 0.425424 / 0.255139 (0.170285) | 0.453390 / 0.283200 (0.170190) | 0.020217 / 0.141683 (-0.121466) | 1.436354 / 1.452155 (-0.015801) | 1.492372 / 1.492716 (-0.000345) |\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) | 0.226896 / 0.018006 (0.208889) | 0.411935 / 0.000490 (0.411445) | 0.000356 / 0.000200 (0.000156) | 0.000054 / 0.000054 (-0.000000) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.024705 / 0.037411 (-0.012706) | 0.076232 / 0.014526 (0.061706) | 0.086949 / 0.176557 (-0.089608) | 0.141867 / 0.737135 (-0.595269) | 0.088199 / 0.296338 (-0.208140) |\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) | 0.419748 / 0.215209 (0.204539) | 4.198597 / 2.077655 (2.120942) | 2.338477 / 1.504120 (0.834357) | 2.195741 / 1.541195 (0.654547) | 2.278145 / 1.468490 (0.809655) | 0.502365 / 4.584777 (-4.082412) | 2.987773 / 3.745712 (-0.757939) | 2.896526 / 5.269862 (-2.373336) | 1.841610 / 4.565676 (-2.724067) | 0.058032 / 0.424275 (-0.366243) | 0.006470 / 0.007607 (-0.001137) | 0.496969 / 0.226044 (0.270925) | 4.960984 / 2.268929 (2.692056) | 2.648615 / 55.444624 (-52.796009) | 2.286846 / 6.876477 (-4.589631) | 2.320176 / 2.142072 (0.178104) | 0.600550 / 4.805227 (-4.204678) | 0.125652 / 6.500664 (-6.375012) | 0.062177 / 0.075469 (-0.013292) |\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) | 1.293063 / 1.841788 (-0.548725) | 18.294204 / 8.074308 (10.219896) | 13.720502 / 10.191392 (3.529110) | 0.146480 / 0.680424 (-0.533944) | 0.016965 / 0.534201 (-0.517236) | 0.330137 / 0.579283 (-0.249146) | 0.352051 / 0.434364 (-0.082313) | 0.381754 / 0.540337 (-0.158584) | 0.517935 / 1.386936 (-0.869001) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#269fcd31a2e759c65ffd5952ecef13e6a0d92574 \"CML watermark\")\n" ]

[]

[ "I think you can just add a test `test_dataset_to_csv_multiproc` in `tests/io/test_csv.py` and we'll be good", "Hi @lhoestq, \r\nI've added `test_dataset_to_csv` apart from `test_dataset_to_csv_multiproc` as no test was there to check generated CSV file when `num_proc=1`. Please let me know if anything is also required! " ]

[]

[ "Thanks for your contribution, @YEdenZ.\r\n\r\nPlease note that our old `metrics` module is in the process of being incorporated to a separate library called `evaluate`: https://github.com/huggingface/evaluate\r\n\r\nTherefore, I would ask you to transfer your PR to that repository. Thank you." ]

[]

[ "Hi ! Thanks for reporting :)\r\n\r\nI managed to reproduce the hub using\r\n```python\r\n\r\nfrom datasets import concatenate_datasets, Dataset, load_from_disk\r\n\r\nDataset.from_dict({\"a\": range(9)}).save_to_disk(\"tmp/ds1\")\r\nds1 = load_from_disk(\"tmp/ds1\")\r\nds1 = concatenate_datasets([ds1, ds1])\r\n\r\nDataset.from_dict({\"b\": range(6)}).save_to_disk(\"tmp/ds2\")\r\nds2 = load_from_disk(\"tmp/ds2\")\r\nds2 = concatenate_datasets([ds2, ds2, ds2])\r\n\r\nconcatenate_datasets([ds1, ds2], axis=1)\r\n```\r\nand I get\r\n```python\r\nTraceback (most recent call last): \r\n File \"test.py\", line 98, in <module>\r\n dds = concatenate_datasets([ds1, ds2], axis=1)\r\n File \"/Users/.../datasets/combine.py\", line 182, in concatenate_datasets\r\n return _concatenate_map_style_datasets(dsets, info=info, split=split, axis=axis)\r\n File \"/Users/.../datasets/arrow_dataset.py\", line 5499, in _concatenate_map_style_datasets\r\n table = concat_tables([dset._data for dset in dsets], axis=axis)\r\n File \"/Users/.../datasets/table.py\", line 1778, in concat_tables\r\n return ConcatenationTable.from_tables(tables, axis=axis)\r\n File \"/Users/.../datasets/table.py\", line 1483, in from_tables\r\n blocks = _extend_blocks(blocks, table_blocks, axis=axis)\r\n File \"/Users/.../datasets/table.py\", line 1477, in _extend_blocks\r\n result[i].extend(row_blocks)\r\nIndexError: list index out of range\r\n```\r\n\r\nIt appears to happen when the two datasets have a number of shards that is not the same" ]

[]

[ "Hi ! For future references let me add a link to our discussion here : https://github.com/huggingface/datasets/issues/759#issuecomment-770684693\r\n\r\nLet me know if you manage to fix your proxy issue or if we can do something on our end to help you :)" ]

[]

[ "The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/datasets/pr_4800). All of your documentation changes will be reflected on that endpoint.", "Hi, thanks for working on this! Can you run `make style` at the repo root to fix the code quality error in CI and add a test?", "Hi, I have fixed the code quality error and added a test", "It seems that CI fails due to the lack of memory for allocating a large array, while I pass the test locally.", "Also, the current implementation of the NumPy-to-PyArrow conversion creates a lot of copies, which is not ideal for large arrays.\r\n\r\nWe can improve performance significantly if we rewrite this part:\r\nhttps://github.com/huggingface/datasets/blob/83f695c14507a3a38e9f4d84612cf49e5f50c153/src/datasets/features/features.py#L1322-L1323\r\n\r\nas\r\n```python\r\n values = pa.array(arr.ravel(), type=type) \r\n```", "@xwwwwww Feel free to ignore https://github.com/huggingface/datasets/pull/4800#issuecomment-1212280549 and revert the changes you've made to address it. \r\n\r\nWithout copying the array, this would be possible:\r\n```python\r\narr = np.array([\r\n [1, 2, 3],\r\n [4, 5, 6]\r\n])\r\n\r\ndset = Dataset.from_dict({\"data\": [arr]})\r\n\r\narr[0][0] = 100 # this change would be reflected in dset's PyArrow table -> a breaking change and also probably unexpected by the user \r\n```", "> @xwwwwww Feel free to ignore [#4800 (comment)](https://github.com/huggingface/datasets/pull/4800#issuecomment-1212280549) and revert the changes you've made to address it.\r\n> \r\n> Without copying the array, this would be possible:\r\n> \r\n> ```python\r\n> arr = np.array([\r\n> [1, 2, 3],\r\n> [4, 5, 6]\r\n> ])\r\n> \r\n> dset = Dataset.from_dict({\"data\": [arr]})\r\n> \r\n> arr[0][0] = 100 # this change would be reflected in dset's PyArrow table -> a breaking change and also probably unexpected by the user \r\n> ```\r\n\r\nOh, that makes sense.", "passed tests in ubuntu while failed in windows", "@mariosasko Hi, do you have any clue about this failure in windows?", "Perhaps we can skip the added test on Windows then.\r\n\r\nNot sure if this can help, but the ERR tool available on Windows outputs the following for the returned error code `-1073741819`:\r\n```\r\n# for decimal -1073741819 / hex 0xc0000005\r\n ISCSI_ERR_SETUP_NETWORK_NODE iscsilog.h\r\n# Failed to setup initiator portal. Error status is given in\r\n# the dump data.\r\n STATUS_ACCESS_VIOLATION ntstatus.h\r\n# The instruction at 0x%p referenced memory at 0x%p. The\r\n# memory could not be %s.\r\n USBD_STATUS_DEV_NOT_RESPONDING usb.h\r\n# as an HRESULT: Severity: FAILURE (1), FACILITY_NONE (0x0), Code 0x5\r\n# for decimal 5 / hex 0x5\r\n WINBIO_FP_TOO_FAST winbio_err.h\r\n# Move your finger more slowly on the fingerprint reader.\r\n# as an HRESULT: Severity: FAILURE (1), FACILITY_NULL (0x0), Code 0x5\r\n ERROR_ACCESS_DENIED winerror.h\r\n# Access is denied.\r\n# 5 matches found for \"-1073741819\"\r\n```", "What's the proper way to skip the added test in windows?\r\nI tried `if platform.system() == 'Linux'`, but the CI test seems stuck", "@mariosasko Hi, any idea about this :)", "Hi again! We want to skip the test on Windows but not on Linux. You can use this decorator to do so: \r\n```python\r\[email protected](os.name == \"nt\" and (os.getenv(\"CIRCLECI\") == \"true\" or os.getenv(\"GITHUB_ACTIONS\") == \"true\"), reason=\"The Windows CI runner does not have enough RAM to run this test\")\r\[email protected](...)\r\ndef test_large_array_xd_with_np(...):\r\n ...\r\n```", "> Hi again! We want to skip the test on Windows but not on Linux. You can use this decorator to do so:\r\n> \r\n> ```python\r\n> @pytest.mark.skipif(os.name == \"nt\" and (os.getenv(\"CIRCLECI\") == \"true\" or os.getenv(\"GITHUB_ACTIONS\") == \"true\"), reason=\"The Windows CI runner does not have enough RAM to run this test\")\r\n> @pytest.mark.parametrize(...)\r\n> def test_large_array_xd_with_np(...):\r\n> ...\r\n> ```\r\n\r\nCI on windows still stucks :(", "@mariosasko Hi, could you please take a look at this issue", "@mariosasko Hi, all checks have passed, and we are finally ready to merge this PR :)", "@lhoestq @albertvillanova Perhaps other maintainers can take a look and merge this PR :)" ]

[]

[ "Fix https://github.com/huggingface/datasets/issues/4179", "_The documentation is not available anymore as the PR was closed or merged._", "Just that I understand: With this change, simply doing `load_dataset(\"librispeech_asr\")` is possible and returns the whole dataset?\r\n\r\nAnd to get the subsets, I do sth like:\r\n```python\r\nds = load_dataset(\"librispeech_asr\")\r\ntrain_ds = ds[\"train\"]\r\ndev_clean_ds = ds[\"dev-clean\"]\r\ndev_other_ds = ds[\"dev-other\"]\r\ntest_clean_ds = ds[\"test-clean\"]\r\ntest_other_ds = ds[\"test-other\"]\r\n```\r\n?\r\n", "> Just that I understand: With this change, simply doing `load_dataset(\"librispeech_asr\")` is possible and returns the whole dataset?\r\n> \r\n> And to get the subsets, I do sth like:\r\n> \r\n> ```python\r\n> ds = load_dataset(\"librispeech_asr\")\r\n> train_ds = ds[\"train\"]\r\n> dev_clean_ds = ds[\"dev-clean\"]\r\n> dev_other_ds = ds[\"dev-other\"]\r\n> test_clean_ds = ds[\"test-clean\"]\r\n> test_other_ds = ds[\"test-other\"]\r\n> ```\r\n> \r\n> ?\r\n\r\nYou could do:\r\n\r\n\r\n```python\r\nds = load_dataset(\"librispeech_asr\", \"all\") # <- note that we have to pass a config\r\ntrain_ds = ds[\"train\"]\r\ndev_clean_ds = ds[\"dev-clean\"]\r\ndev_other_ds = ds[\"dev-other\"]\r\ntest_clean_ds = ds[\"test-clean\"]\r\ntest_other_ds = ds[\"test-other\"]\r\n```", "So, `load_dataset(\"librispeech_asr\")` is not possible, it must be `load_dataset(\"librispeech_asr\", \"all\")`?\r\n\r\nWhy is that?\r\n\r\nThe docs say:\r\n```\r\nname: `str` name, optional configuration for the dataset that affects the data generated on disk. Different\r\n `builder_config`s will have their own subdirectories and versions.\r\n If not provided, uses the first configuration in self.BUILDER_CONFIGS\r\n```\r\nhttps://github.com/huggingface/datasets/blob/cd3ce34ab1604118351e1978d26402de57188901/src/datasets/builder.py#L228\r\n\r\nOr maybe you could just define `DEFAULT_CONFIG_NAME`?\r\n", "> If not provided, uses the first configuration in self.BUILDER_CONFIGS\r\n\r\nOh crap this is outdated documentation. No it doesn't take the first config by default.\r\n\r\nEDIT: opened a PR to fix this: https://github.com/huggingface/datasets/pull/4186", "> No it doesn't take the first config by default.\r\n\r\nBut defining `DEFAULT_CONFIG_NAME` would work?\r\n\r\nSo should we define `DEFAULT_CONFIG_NAME = \"all\"` here as well? I think this is a reasonable default config.\r\n\r\nDon't most datasets have some default config?\r\n", "> But defining DEFAULT_CONFIG_NAME would work?\r\n>\r\n> So should we define DEFAULT_CONFIG_NAME = \"all\" here as well? I think this is a reasonable default config.\r\n\r\nYes that would work, and I also find it reasonable to do it :)\r\n\r\n> Don't most datasets have some default config?\r\n\r\nMost datasets only have one configuration, so the single configuration is the default one. Then other datasets gave several configurations, and whether they have a default one is decided case-by-case.\r\n\r\ne.g. `glue` is a benchmark and doesn't have a default task, one must choose which task of `glue` they want to use explicitely.", "Thanks a lot for the feedback! \r\n\r\nUsing `\"all\"` now as the default config. I changed the layout a bit so that there is not a single \"train\", but instead we have multiple \"train.clean.100\", \"train.clean.360\", \"train.other.500\". This way we don't even need to do filtering and it's also cleaner IMO.\r\n\r\n@albertz - you should now be able to do the following:\r\n\r\n```python\r\nload_dataset(\"librispeech_asr\") # <- run this once to download, prepare dataset and cache everything\r\n\r\n# The following operations will be very fast since all the downloading and processing is already cached\r\ntrain_1 = load_dataset(\"librispeech_asr\", split=\"train.clean.100\")\r\nprint(train_1)\r\ntrain_2 = load_dataset(\"librispeech_asr\", split=\"train.clean.100+train.clean.360\")\r\nprint(train_2)\r\ntrain_full = load_dataset(\"librispeech_asr\", split=\"train.clean.100+train.clean.360+train.other.500\")\r\nprint(train_full)\r\ndev_clean_ds = load_dataset(\"librispeech_asr\", split=\"validation.clean\")\r\nprint(dev_clean_ds)\r\ndev_other_ds = load_dataset(\"librispeech_asr\", split=\"validation.other\")\r\nprint(dev_other_ds)\r\ntest_clean_ds = load_dataset(\"librispeech_asr\", split=\"test.clean\")\r\nprint(test_clean_ds)\r\ntest_other_ds = load_dataset(\"librispeech_asr\", split=\"test.other\")\r\nprint(test_other_ds)\r\n```\r\n\r\n\r\n", "Think this way we have the best of both worlds. Also @lhoestq, I think we could highlight better in the docs that it's possible to combine different splits. We do this actually quite a lot for speech. For Common Voice many people include \"validation\" in the training if the data is too small, e.g.: https://github.com/huggingface/transformers/blob/ff06b177917384137af2d9585697d2d76c40cdfc/examples/pytorch/speech-recognition/run_speech_recognition_ctc.py#L147\r\n\r\nShould we maybe add a short section to the loading tutorial here: https://huggingface.co/docs/datasets/v2.1.0/en/loading#hugging-face-hub ? (Happy to do it)", "Is there any advantage or difference in calling `load_dataset` multiple times for each split? Or why not just call `load_dataset` once and then access each split?\r\n\r\nNote in our case, we cannot really use the caching mechanism because we have a recipe pipeline used by multiple users (and I think a common cache dir for all users might end up in problems) and we basically would use `load_dataset(\"librispeech_asr\").save_to_disk(...)` and then later `load_from_disk(...)`. (See here: https://github.com/rwth-i6/i6_core/pull/253)\r\n\r\nSo with `load_from_disk`, we cannot really provide the split this way, so we anyway would do sth like:\r\n```python\r\nds = datasets.load_from_disk(...)\r\ntrain = ds[\"train\"]\r\n```\r\nOr with your latest proposal, it would look like:\r\n```python\r\nds = datasets.load_from_disk(...)\r\ntrain_ds = datasets.concatenate_datasets(\r\n [ds[\"train.clean.100\"], ds[\"train.clean.360\"], ds[\"train.other.500\"]])\r\n```\r\nright?\r\n", "> Is there any advantage or difference in calling `load_dataset` multiple times for each split? Or why not just call `load_dataset` once and then access each split?\r\n> \r\n> Note in our case, we cannot really use the caching mechanism because we have a recipe pipeline used by multiple users (and I think a common cache dir for all users might end up in problems) and we basically would use `load_dataset(\"librispeech_asr\").save_to_disk(...)` and then later `load_from_disk(...)`. (See here: [rwth-i6/i6_core#253](https://github.com/rwth-i6/i6_core/pull/253))\r\n> \r\n> So with `load_from_disk`, we cannot really provide the split this way, so we anyway would do sth like:\r\n> \r\n> ```python\r\n> ds = datasets.load_from_disk(...)\r\n> train = ds[\"train\"]\r\n> ```\r\n> \r\n> Or with your latest proposal, it would look like:\r\n> \r\n> ```python\r\n> ds = datasets.load_from_disk(...)\r\n> train_ds = datasets.concatenate_datasets(\r\n> [ds[\"train.clean.100\"], ds[\"train.clean.360\"], ds[\"train.other.500\"]])\r\n> ```\r\n> \r\n> right?\r\n\r\nI see the use case! The only advantage by calling `datasets` multiple times is that one can easily \"merge\" splits with `\"+\"`, but yeah you can do the exact same with `concatenate`.\r\n\r\n@lhoestq what do you think is the best approach with `load_from_disk`? \r\n\r\n@albertz, you could also define the `cache_dir` when doing `load_dataset(...)` which will then put all the relevant `arrow` files int the cache dir that you defined, e.g.:\r\n\r\n```python\r\nload_dataset(\"librispeech_asr\", cache_dir=\"/easy/to/access/directory\")\r\n```", "@albertz, I took a read through https://github.com/rwth-i6/i6_core/pull/253 . \r\n\r\nI think the best would be the following:\r\n\r\n1. Do `ds = load_dataset(..., cache_dir=\"/dir/that/is/easy/to/access\")` <- having merged this PR, this will save all the original `.flac` files in the `cache_dir`\r\n2. Do `ds.save_to_disk(\"local/path\")` this should then only save the `arrow.format` with a `path` string to the audio files which are located in `cache_dir` <- this won't require a lot of memory after https://github.com/huggingface/datasets/pull/4184#discussion_r854132740 is fixed and can be done for each person individually.\r\n3. `ds = datasets.load_from_disk(\"local/path\")` can the be used. An object of `ds` will then have a `path` variable that links to the original audio files in the `cache_dir`. You can change these audio files then easily to `.mp3. You could do this with the `.map(...)` function, e.g. define a function that maps through all audio files, load them and then save them on disk afterward.", "@lhoestq - I think this one is good to go", "> @albertz, I took a read through [rwth-i6/i6_core#253](https://github.com/rwth-i6/i6_core/pull/253) .\r\n> \r\n> I think the best would be the following:\r\n> \r\n> 1. Do `ds = load_dataset(..., cache_dir=\"/dir/that/is/easy/to/access\")` <- having merged this PR, this will save all the original `.flac` files in the `cache_dir`\r\n> 2. Do `ds.save_to_disk(\"local/path\")` this should then only save the `arrow.format` with a `path` string to the audio files which are located in `cache_dir` <- this won't require a lot of memory after [[Librispeech] Add 'all' config #4184 (comment)](https://github.com/huggingface/datasets/pull/4184#discussion_r854132740) is fixed and can be done for each person individually.\r\n> 3. `ds = datasets.load_from_disk(\"local/path\")` can the be used. An object of `ds` will then have a `path` variable that links to the original audio files in the `cache_dir`. You can change these audio files then easily to `.mp3. You could do this with the `.map(...)` function, e.g. define a function that maps through all audio files, load them and then save them on disk afterward.\r\n\r\nOh, so you say that our current implementation in https://github.com/rwth-i6/i6_core/pull/253 is broken? Because our cache dir is just some temp directory which will be removed afterwards, and we just store what we get out of `save_to_disk`. I think it would be good to clarify that in the doc of `save_to_disk`, that this is not enough and can depend on files from the cache dir. (@dthulke)\r\n\r\nSo, you say we anyway need to share the cache dir among users? But we would want to make sure that after the initial download and preparation of the data, this is set to readonly, because we want to make sure that other people will not modify the data in any way. Right?\r\n\r\nBut then, we don't really need the `save_to_disk` and `load_from_disk` at all, right?\r\n", "@albertz \r\n\r\n> Oh, so you say that our current implementation in https://github.com/rwth-i6/i6_core/pull/253 is broken? Because our cache dir is just some temp directory which will be removed afterwards, and we just store what we get out of save_to_disk. I think it would be good to clarify that in the doc of save_to_disk, that this is not enough and can depend on files from the cache dir. (@dthulke)\r\n\r\nOh, I wasn't aware that audio files are handled this way. Then we should have the cache directory as an additional job output, so that we keep the audio files. \r\n\r\n> So, you say we anyway need to share the cache dir among users?\r\n\r\nNo, the cache dir can still be a directory in the job output folder. Then the audio paths in the corresponding dataset column correspond to the flac files in that directory. This way the \"output\" of the job is contained into the job directory and we don't write files to a global cache directory that is independent of the sisyphus graph.\r\n\r\nIf we want to share the audio data between different users, we can just link to a central instance of the job (similar to how we do it with the `DownloadLibriSpeechCorpusJob`).", "@dthulke - that's a good point actually! So you can do both things:\r\n\r\n1. Convert all audio files to bytes. Bytes can be saved by `arrow` so in this case you can do `save_to_disk(...)`, but then you cannot really inspect the audio files locally as they'll just be saved within a large arrow file (this actually used to be the default case but we're changing this now). The problem of this is summarized here a bit: https://github.com/huggingface/datasets/issues/3663 . You can still do this if you'd like, e.g. you could do:\r\n\r\n```python\r\nds = load_dataset(\"librispeech_asr\")\r\n\r\ndef read_file(batch):\r\n with open(batch[\"file\"], \"r\") as f:\r\n batch[\"bytes\"] = f.read() \r\n return batch\r\n\r\nds = ds.map(read_file)\r\nds.save_to_disk(\"/path\") <- the saved arrow object will now contain everything you need\r\n```\r\n\r\nhowever this is not recommend - it's should be much easier to just save the path to the downloaded audio files.\r\n\r\n2. Not convert audio files to bytes, but just leave them in their original file format. Then only the path to the original files will be save in arrow. This will be the default case. This means that when you do `load_dataset(...)` both the orginal audio data and the arrow file will be saved in the `cache_dir` (which can be saved locally for every user or in a shared cache - we actually use a shared cache quite a bit at Hugging Face). When do you do `save_to_disk(...)` now only the `path` will be saved in `arrow` format (after this PR is merged, you'll see that the `arrow files should be very light weight` meaning that `save_to_disk(...)` can be done for every user, but has a dependency on the `cache_dir` (because the audio files live there).\r\n\r\n=> Now what you could do as well would be to simply move all the audio files to the folder you want (the `save_to_disk(...)` folder) and then change the path of every sample to this folder (maybe with `map(...)`) and then this folder would be self contained. I do however think it's better to just specific a `cache_dir` and re-use `load_dataset(...)` every time instead of `load_from_disk` or `save_to_disk(...)`. Note that you can even pass the relevant cache files to `load_dataset(...)` here: https://huggingface.co/docs/datasets/v2.1.0/en/package_reference/loading_methods#datasets.load_dataset.data_files in which case you can be 100% sure that nothing is redownloaded. \r\n\r\nWe discussed storing audio files quite a bit, e.g. see: https://github.com/huggingface/datasets/issues/3663 and had (too many) changes around this topic recently, but we've come to the conclusion that the best is to leave the audio format in the format it was originally (`.flac` for Librispeech) so that the user can easily inspect it / understand the data. Arrow cannot save data is `.flac` so we'll just save a path to the original data. Curious to hear your guys opinion on this as well.", "So what I would suggest here is to do the following:\r\n\r\n1. Do `load_dataset(..., cache_dir=/a/read-only/folder)`\r\n2. \r\n- Either just re-use `load_dataset(..., cache_dir=...)` which should always re-use the data in the `cache_dir` since the hash of the url matches - so there should never be any duplicated downloading \r\n\r\nor \r\n\r\n- If you want to store the files in MP3 locally, first convert the files to MP3 in the read-only folder, then take do `ds.save_to_disk(/some/path)` which will save the correct path to the read-only folder to MP3 and then you can easily re-use the small arrow dataset that is saved in `/some/path`", "> So what I would suggest here is to do the following:\r\n> \r\n> 1. Do `load_dataset(..., cache_dir=/a/read-only/folder)`\r\n> \r\n> * Either just re-use `load_dataset(..., cache_dir=...)` which should always re-use the data in the `cache_dir` since the hash of the url matches - so there should never be any duplicated downloading\r\n> \r\n> or\r\n> \r\n> * If you want to store the files in MP3 locally, first convert the files to MP3 in the read-only folder, then take do `ds.save_to_disk(/some/path)` which will save the correct path to the read-only folder to MP3 and then you can easily re-use the small arrow dataset that is saved in `/some/path`\r\n\r\nAlso relevant here: https://github.com/huggingface/datasets/issues/3663", "I also added some documentation about how `save_to_disk` handles audio files here: https://github.com/huggingface/datasets/pull/4193", "> > So, you say we anyway need to share the cache dir among users?\r\n> \r\n> No, the cache dir can still be a directory in the job output folder.\r\n\r\n@dthulke But this is what I mean. When we share the job output folder, it means we share the cache dir among users.\r\n\r\nI wonder if `load_dataset(..., cache_dir=job_output_cache_dir)` is always save to do then, that it really would not modify the `job_output_cache_dir`.\r\n\r\nWe could enforce that by making the `job_output_cache_dir` read-only afterwards. We currently don't do this.\r\n\r\n@patrickvonplaten @dthulke But in any case, we actually prefer the data content to be inside the dataset (the arrow files). Lots of small files would be very problematic for our cache manager. We have one main copy of the data on NFS, but accessing the NFS directly by all computing nodes is not feasible, so the cache manager will have copies of the files on the nodes. So it means, whenever we access some file, we query the cache manager DB whether the file is already cached somewhere (some other computing node) and if so, it copies it from the other computing node and not from NFS. This works very well when there are not too many files (but the files can be big). So, we want to have only a few but big files. Even for NFS access this is much better.\r\n\r\nI also commented in #3663.\r\n", "Hey @albertz @dthulke,\r\n\r\nThanks a lot for your input! \r\n\r\nWe've discussed quite a bit with @lhoestq and we think the best approach is the following:\r\n\r\n\r\na)\r\n`load_dataset(...)` will not store both bytes and the files because this would mean that 3x the size of the dataset would often be needed (1. the compressed `tar.gz` file, 2. the extracted file b, 3. the raw bytes in arrow format). \r\n\r\nFor canonical datasets like librispeech and common voice I think we want to keep the dataset filenames because of i) no breaking changes and ii) reasons explained in #3663\r\n\r\nHowever it's also trivial to write your own datasetset downloading script of librispeech and just not extract the folder e.g. this line: https://huggingface.co/datasets/common_voice/blob/main/common_voice.py#L671\r\n\r\nAnd then it'll be allowed to save the bytes and the dataset will be self-contained out-of-the-box when using `load_dataset(...)`\r\n\r\nb) Now, one major problem that you guys uncovered is that `save_to_disk(...)` is currently not necessarily saving a dataset to be self-contained. We will change that asap. This means that after we've corrected this when you do download the canonical librispeech dataset the following will work:\r\n\r\n```python\r\nds = load_dataset(\"....\") # <- here we have a dependency on the filepathes\r\nds[0][\"audio\"][\"bytes\"] # <- will not work\r\n\r\nds.save_to_disk(\"/local/path\") # <- now we want to have a self-contained dataset in arrow format, so we load the files into bytes and save it in arrow format\r\n\r\n# now you can delete everything besides \"/local/path\"\r\n\r\nds = load_from_disk(\"/local/path\") # <- this will work\r\n```\r\n\r\nSo either option a) where you define your own librispeech data downloading script (you guys could just sign up here: https://huggingface.co/join) and upload a dataset loading script in private mode so that no one can see it and you would always store the audio as bytes or b) where you first load then save to disk then delete cache would work. \r\n\r\nHope that fits in your vision :-)\r\n\r\ncc @lhoestq @mariosasko ", "@patrickvonplaten sounds like a good approach to me. For b) this could even be configurable with a parameter like `embed_external_files` as you have for `push_to_hub` (if people prefer to keep separate audio files).\r\n", "> However it's also trivial to write your own datasetset downloading script of librispeech and just not extract the folder\r\n\r\nI don't exactly understand. In all cases, we need to extract it to prepare the dataset, or not? No matter if we want to store the raw bytes inside the dataset or leaving them as local files. Just in the first case, we can safely delete the extracted files after the dataset preparation.\r\n\r\n> `save_to_disk(...)` is currently not necessarily saving a dataset to be self-contained. We will change that asap.\r\n\r\nFor us, this sounds exactly like what we want.\r\n\r\nBut regarding not introducing breaking changes, wouldn't this maybe also break some setups for users who don't expect this new behavior?\r\n", "@albertz I would suggest to move the discussion on implementation details on our side to the following issue: rwth-i6/i6_core/issues/257", "I like the idea of adding `embed_external_files` and set it to True by default to `save_to_disk`.\r\nIt's indeed a kind of breaking change since some users will get bigger Arrow files when updating the lib, but the advantages are nice:\r\n1. I like the idea of having it self contained, in case you want to delete your cache\r\n2. users also upload these Arrow files to cloud storage via the `fs` parameter, and in this case they would expect to upload a self-contained dataset\r\n3. consistency with `push_to_hub`\r\n\r\nIf it sounds good to you I'll open an issue to discuss this and track the advancements", "Closed #4179." ]

[]

[ "_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._", "Thanks for adding this ! You can fix the CI by formatting your code using the `make style` command :)", "[**@lhoestq**](https://github.com/lhoestq) commented on [Oct 27, 2022, 4:08 PM GMT+3:30](https://github.com/huggingface/datasets/pull/5155#issuecomment-1293464680 \"2022-10-27T12:38:04Z - Replied by Github Reply Comments\"):\r\n> Thanks for adding this ! You can fix the CI by formatting your code using the `make style` command :)\r\n\r\nI ran this and force pushed the changes." ]

[]

[ "Also you can ignore the CI failing on `docs`, this has been fixed on master :)", "@lhoestq I need to update other stuff on GEM later today too, so will merge this one and remove columns in the next PR!", "Ok !" ]

[]

[ "Sorry about forgetting flake8.\r\nRather than use up the circleci resources on a new push with only formatting changes, I will wait to push until the results from all tests finish and/or any feedback comes in... probably tomorrow for me.", "\r\nSo... my normal work is with mercurial and seem to have clearly forked this up using git... :(\r\n\r\nWhat I did is after calling:\r\n```\r\ngit fetch upstream\r\ngit rebase upstream/master\r\n```\r\n\r\nI then I attempt to pull in my most recent changes UI commit changes based on @lhoestq's feedback with:\r\n```\r\ngit pull\r\n``` \r\n... which I now suspect undid the above fetch and rebase. Will look into fixing later today when I have more time. Sorry!\r\n", "My dummy data seems quite large as a single row is composed of tokens/labels for an entire youtube video, with at least one row required for each file, which in this case 1 file per 13 youtube channels.\r\n\r\nTo make it smaller I passed `--n_lines 1` to reduce about 5x.\r\n\r\nI then manually reduced size of the particularly long youtube lectures to get the size to about 30KB. However, after recompressing into a zip, and running dummy data test I got the following error:\r\n`FAILED tests/test_dataset_common.py::LocalDatasetTest::test_load_dataset_all_configs_youtube_caption_corrections - OSError: Cannot find data file. `, despite file being there, which I haven't had a chance yet to debug.", "I wrote a small script to generate a smaller json file for the dummy_data, with the hope that I could resolve the pytest error noted above (in case related to a manual typo I could have introduce), however the test contains to fail locally... here's to hoping it can pass on remote!", "Sorry for delayed comments here. Last commit made two changes:\r\n- Increased the valency of the labels from just True/False to more categories to describe the various types of diffs encountered. This required some rewrite of the README\r\n- Reduced the number of remote files to be downloaded from 13 to 4, by combining all 13 of the channel-specific files together, and the splitting them up in a way to meet Github file size requirements. This also reduces size of the dummy-data.", "@lhoestq, thank you for the great feedback, especially given how busy you guys are now! \r\n\r\nI checked out GitHub release tags and looks cool. I have added the version tag to the url, instead of the commit sha as originally suggested, with the hope that it serves the same purpose of pinning the content to this url. Please let me know if I have misunderstood.\r\n\r\nIn regard to dynamically changing the number of files downloaded by first downloading a JSON listing the files, I love that idea. But I am a little confused, as I was thinking that any changes to the dataset itself would require a new PR with an updated `dataset_infos.json`, e.g. `num_examples` would increase. \r\n\r\nIf the purpose of this is not to permit dynamic (without a PR needed) growth of the number of files, but instead to provide stability to the consumers of the dataset, maybe I continued use the release tags, maintaining access to old releases could serve this purpose? I am still learning about these release tags... ", "For dynamic datasets, i.e. datasets that evolve over time, we support custom configurations: they are configurations that are not part of the BUILDER_CONFIGS or in the dataset_infos.json\r\n\r\nFor example for wikipedia, you can use the latest wiki dump by specifying `date=` inside `load_dataset()`. A configuration is created on the fly for this date and is used to build the dataset using the latest data.\r\n\r\nTherefore we don't need to have PRs to update the script for each wikipedia release.\r\n\r\nOne downside though is that we don't have metadata in advance such as the size of the dataset.\r\n\r\nI think this could be a nice addition for the youtube caption dataset in the future to be have a system of releases and be able to load the version we want easily. What do you think ?", "\r\n\r\n\r\n\r\n> For dynamic datasets, i.e. datasets that evolve over time, we support custom configurations: they are configurations that are not part of the BUILDER_CONFIGS or in the dataset_infos.json\r\n> \r\n \r\n> I think this could be a nice addition for the youtube caption dataset in the future to be have a system of releases and be able to load the version we want easily. What do you think ?\r\n\r\nThank you for the suggestion! This sounds great! I will take a look at the some datasets that do this, and would love to give it a try in the future, if I continue to grow the captions dataset in a meaningful way. \r\n\r\nAppreciate all the help on this. It has been a really great experience for me. :)", "Excited to merge! And sorry to be such a github n00b, but from what I've quickly read, I don't 'Close pull request', but rather the next steps are action taken on your end... Please let me know if there is some action to be taken at my end first. :/", "Alright merging this one then :) " ]

[]

[]

[]

[ "Finally, I'm splitting this PR." ]

[]

[ "@lhoestq Thank you for suggesting changes. I fixed all the changes you suggested. Can you please review it again? ", "@lhoestq Thank you for reviewing and suggesting changes. I made the requested changes. Can you please review it again?", "Thanks @lhoestq for reviewing it again. I made the required changes. Can you please have a look ?", "merging since the CI is fixed on master" ]

[]

[]

[]

[ "I am thinking that ideally we should call `fs.glob()` instead...", "Thanks, @lhoestq: the idea of adding the mock filesystem is to avoid network calls and reduce testing time ;) \r\n\r\nI have added `rglob` as well and fixed some bugs." ]

[]

[ "<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) | 0.005969 / 0.011353 (-0.005384) | 0.003687 / 0.011008 (-0.007321) | 0.100843 / 0.038508 (0.062335) | 0.036912 / 0.023109 (0.013803) | 0.312389 / 0.275898 (0.036491) | 0.370335 / 0.323480 (0.046855) | 0.003434 / 0.007986 (-0.004552) | 0.003710 / 0.004328 (-0.000619) | 0.076899 / 0.004250 (0.072648) | 0.053647 / 0.037052 (0.016594) | 0.324825 / 0.258489 (0.066336) | 0.367711 / 0.293841 (0.073870) | 0.028079 / 0.128546 (-0.100467) | 0.008326 / 0.075646 (-0.067320) | 0.312342 / 0.419271 (-0.106930) | 0.047423 / 0.043533 (0.003890) | 0.321063 / 0.255139 (0.065924) | 0.336508 / 0.283200 (0.053308) | 0.019973 / 0.141683 (-0.121710) | 1.529334 / 1.452155 (0.077179) | 1.573746 / 1.492716 (0.081030) |\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) | 0.210849 / 0.018006 (0.192843) | 0.418798 / 0.000490 (0.418309) | 0.007347 / 0.000200 (0.007147) | 0.000070 / 0.000054 (0.000016) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.022718 / 0.037411 (-0.014694) | 0.098400 / 0.014526 (0.083874) | 0.106590 / 0.176557 (-0.069967) | 0.168460 / 0.737135 (-0.568675) | 0.108401 / 0.296338 (-0.187938) |\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) | 0.443066 / 0.215209 (0.227857) | 4.416658 / 2.077655 (2.339003) | 2.088844 / 1.504120 (0.584724) | 1.879564 / 1.541195 (0.338369) | 1.933815 / 1.468490 (0.465325) | 0.565085 / 4.584777 (-4.019692) | 3.412440 / 3.745712 (-0.333273) | 1.754686 / 5.269862 (-3.515175) | 1.024576 / 4.565676 (-3.541100) | 0.067909 / 0.424275 (-0.356366) | 0.011054 / 0.007607 (0.003447) | 0.534748 / 0.226044 (0.308703) | 5.351457 / 2.268929 (3.082529) | 2.517368 / 55.444624 (-52.927256) | 2.182762 / 6.876477 (-4.693715) | 2.238205 / 2.142072 (0.096133) | 0.672962 / 4.805227 (-4.132265) | 0.136098 / 6.500664 (-6.364566) | 0.066534 / 0.075469 (-0.008935) |\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) | 1.281241 / 1.841788 (-0.560547) | 13.872881 / 8.074308 (5.798573) | 13.161023 / 10.191392 (2.969631) | 0.130011 / 0.680424 (-0.550412) | 0.016759 / 0.534201 (-0.517442) | 0.359802 / 0.579283 (-0.219481) | 0.392577 / 0.434364 (-0.041787) | 0.427742 / 0.540337 (-0.112595) | 0.522241 / 1.386936 (-0.864695) |\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) | 0.005985 / 0.011353 (-0.005368) | 0.003705 / 0.011008 (-0.007304) | 0.077699 / 0.038508 (0.039191) | 0.035686 / 0.023109 (0.012577) | 0.420356 / 0.275898 (0.144458) | 0.476753 / 0.323480 (0.153273) | 0.003510 / 0.007986 (-0.004475) | 0.002807 / 0.004328 (-0.001521) | 0.077151 / 0.004250 (0.072901) | 0.046420 / 0.037052 (0.009368) | 0.391781 / 0.258489 (0.133292) | 0.461128 / 0.293841 (0.167287) | 0.027847 / 0.128546 (-0.100699) | 0.008322 / 0.075646 (-0.067324) | 0.082768 / 0.419271 (-0.336503) | 0.042629 / 0.043533 (-0.000904) | 0.405745 / 0.255139 (0.150606) | 0.430797 / 0.283200 (0.147598) | 0.019832 / 0.141683 (-0.121851) | 1.556208 / 1.452155 (0.104054) | 1.612166 / 1.492716 (0.119450) |\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) | 0.230633 / 0.018006 (0.212626) | 0.401667 / 0.000490 (0.401178) | 0.000776 / 0.000200 (0.000576) | 0.000069 / 0.000054 (0.000014) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.024959 / 0.037411 (-0.012452) | 0.100560 / 0.014526 (0.086034) | 0.109175 / 0.176557 (-0.067382) | 0.159919 / 0.737135 (-0.577217) | 0.112810 / 0.296338 (-0.183528) |\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) | 0.460601 / 0.215209 (0.245392) | 4.620039 / 2.077655 (2.542385) | 2.257900 / 1.504120 (0.753780) | 2.039192 / 1.541195 (0.497997) | 2.064451 / 1.468490 (0.595961) | 0.557887 / 4.584777 (-4.026890) | 3.356100 / 3.745712 (-0.389612) | 1.703578 / 5.269862 (-3.566284) | 1.024984 / 4.565676 (-3.540693) | 0.067602 / 0.424275 (-0.356673) | 0.011450 / 0.007607 (0.003842) | 0.563230 / 0.226044 (0.337186) | 5.632150 / 2.268929 (3.363221) | 2.698701 / 55.444624 (-52.745924) | 2.363218 / 6.876477 (-4.513259) | 2.363997 / 2.142072 (0.221925) | 0.671260 / 4.805227 (-4.133967) | 0.136166 / 6.500664 (-6.364499) | 0.067094 / 0.075469 (-0.008375) |\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) | 1.303030 / 1.841788 (-0.538757) | 14.137277 / 8.074308 (6.062969) | 13.937631 / 10.191392 (3.746239) | 0.162626 / 0.680424 (-0.517798) | 0.016687 / 0.534201 (-0.517514) | 0.363657 / 0.579283 (-0.215626) | 0.392021 / 0.434364 (-0.042343) | 0.427275 / 0.540337 (-0.113062) | 0.512192 / 1.386936 (-0.874744) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#42603528d9bd8c3ab287ed0eadc7fa3d1ef4cfd8 \"CML watermark\")\n", "_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) | 0.005974 / 0.011353 (-0.005378) | 0.003947 / 0.011008 (-0.007061) | 0.098604 / 0.038508 (0.060096) | 0.036947 / 0.023109 (0.013838) | 0.311844 / 0.275898 (0.035946) | 0.375243 / 0.323480 (0.051763) | 0.003453 / 0.007986 (-0.004533) | 0.003834 / 0.004328 (-0.000495) | 0.077943 / 0.004250 (0.073692) | 0.052956 / 0.037052 (0.015904) | 0.320812 / 0.258489 (0.062323) | 0.373963 / 0.293841 (0.080122) | 0.028382 / 0.128546 (-0.100164) | 0.008525 / 0.075646 (-0.067121) | 0.311306 / 0.419271 (-0.107965) | 0.047029 / 0.043533 (0.003496) | 0.309933 / 0.255139 (0.054794) | 0.335114 / 0.283200 (0.051915) | 0.019629 / 0.141683 (-0.122054) | 1.569771 / 1.452155 (0.117617) | 1.585899 / 1.492716 (0.093182) |\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) | 0.216565 / 0.018006 (0.198559) | 0.426717 / 0.000490 (0.426228) | 0.003609 / 0.000200 (0.003409) | 0.000077 / 0.000054 (0.000023) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.023079 / 0.037411 (-0.014332) | 0.096954 / 0.014526 (0.082428) | 0.105398 / 0.176557 (-0.071158) | 0.165433 / 0.737135 (-0.571703) | 0.109703 / 0.296338 (-0.186636) |\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) | 0.456227 / 0.215209 (0.241018) | 4.529857 / 2.077655 (2.452202) | 2.214054 / 1.504120 (0.709934) | 2.029716 / 1.541195 (0.488521) | 2.081175 / 1.468490 (0.612685) | 0.563642 / 4.584777 (-4.021135) | 3.355393 / 3.745712 (-0.390320) | 1.765938 / 5.269862 (-3.503924) | 1.039062 / 4.565676 (-3.526615) | 0.067952 / 0.424275 (-0.356323) | 0.011044 / 0.007607 (0.003437) | 0.556935 / 0.226044 (0.330890) | 5.588167 / 2.268929 (3.319239) | 2.667217 / 55.444624 (-52.777407) | 2.337383 / 6.876477 (-4.539094) | 2.429590 / 2.142072 (0.287517) | 0.676972 / 4.805227 (-4.128256) | 0.135782 / 6.500664 (-6.364882) | 0.066323 / 0.075469 (-0.009146) |\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) | 1.237358 / 1.841788 (-0.604429) | 13.910492 / 8.074308 (5.836184) | 13.227275 / 10.191392 (3.035883) | 0.146857 / 0.680424 (-0.533567) | 0.016991 / 0.534201 (-0.517210) | 0.363637 / 0.579283 (-0.215646) | 0.392462 / 0.434364 (-0.041902) | 0.450009 / 0.540337 (-0.090329) | 0.536077 / 1.386936 (-0.850859) |\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) | 0.006067 / 0.011353 (-0.005286) | 0.003851 / 0.011008 (-0.007158) | 0.078462 / 0.038508 (0.039954) | 0.036221 / 0.023109 (0.013112) | 0.389195 / 0.275898 (0.113297) | 0.428710 / 0.323480 (0.105230) | 0.004645 / 0.007986 (-0.003341) | 0.002973 / 0.004328 (-0.001355) | 0.078299 / 0.004250 (0.074048) | 0.047076 / 0.037052 (0.010024) | 0.375673 / 0.258489 (0.117184) | 0.432352 / 0.293841 (0.138511) | 0.028212 / 0.128546 (-0.100334) | 0.008475 / 0.075646 (-0.067172) | 0.083902 / 0.419271 (-0.335369) | 0.046699 / 0.043533 (0.003166) | 0.364502 / 0.255139 (0.109363) | 0.389792 / 0.283200 (0.106592) | 0.025266 / 0.141683 (-0.116417) | 1.517458 / 1.452155 (0.065303) | 1.543634 / 1.492716 (0.050918) |\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) | 0.236479 / 0.018006 (0.218472) | 0.411528 / 0.000490 (0.411038) | 0.005213 / 0.000200 (0.005013) | 0.000091 / 0.000054 (0.000036) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.025764 / 0.037411 (-0.011647) | 0.103174 / 0.014526 (0.088648) | 0.110609 / 0.176557 (-0.065948) | 0.164630 / 0.737135 (-0.572506) | 0.114863 / 0.296338 (-0.181475) |\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) | 0.457155 / 0.215209 (0.241946) | 4.550675 / 2.077655 (2.473021) | 2.350473 / 1.504120 (0.846353) | 2.204919 / 1.541195 (0.663724) | 2.076724 / 1.468490 (0.608234) | 0.563107 / 4.584777 (-4.021670) | 3.390669 / 3.745712 (-0.355043) | 1.741111 / 5.269862 (-3.528751) | 1.033268 / 4.565676 (-3.532408) | 0.068400 / 0.424275 (-0.355875) | 0.011607 / 0.007607 (0.004000) | 0.561944 / 0.226044 (0.335900) | 5.620224 / 2.268929 (3.351296) | 2.705241 / 55.444624 (-52.739384) | 2.344520 / 6.876477 (-4.531957) | 2.386119 / 2.142072 (0.244046) | 0.681583 / 4.805227 (-4.123644) | 0.137272 / 6.500664 (-6.363392) | 0.069217 / 0.075469 (-0.006252) |\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) | 1.322690 / 1.841788 (-0.519098) | 14.464953 / 8.074308 (6.390645) | 14.269350 / 10.191392 (4.077958) | 0.158879 / 0.680424 (-0.521545) | 0.016722 / 0.534201 (-0.517479) | 0.360299 / 0.579283 (-0.218984) | 0.391609 / 0.434364 (-0.042755) | 0.420507 / 0.540337 (-0.119831) | 0.512822 / 1.386936 (-0.874114) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#ca68191900d97b29abb3c2c4ba0502fe30d137d1 \"CML watermark\")\n", "<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) | 0.007106 / 0.011353 (-0.004247) | 0.005224 / 0.011008 (-0.005784) | 0.127563 / 0.038508 (0.089055) | 0.055067 / 0.023109 (0.031958) | 0.418660 / 0.275898 (0.142761) | 0.487891 / 0.323480 (0.164411) | 0.005712 / 0.007986 (-0.002274) | 0.004585 / 0.004328 (0.000256) | 0.090994 / 0.004250 (0.086743) | 0.071837 / 0.037052 (0.034784) | 0.446957 / 0.258489 (0.188468) | 0.475966 / 0.293841 (0.182125) | 0.038062 / 0.128546 (-0.090484) | 0.010056 / 0.075646 (-0.065590) | 0.406796 / 0.419271 (-0.012475) | 0.066542 / 0.043533 (0.023009) | 0.413676 / 0.255139 (0.158537) | 0.448624 / 0.283200 (0.165424) | 0.030332 / 0.141683 (-0.111351) | 1.895307 / 1.452155 (0.443152) | 1.904411 / 1.492716 (0.411694) |\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) | 0.221246 / 0.018006 (0.203240) | 0.461288 / 0.000490 (0.460799) | 0.005957 / 0.000200 (0.005757) | 0.000112 / 0.000054 (0.000058) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.029255 / 0.037411 (-0.008156) | 0.131299 / 0.014526 (0.116773) | 0.135814 / 0.176557 (-0.040742) | 0.201342 / 0.737135 (-0.535793) | 0.141748 / 0.296338 (-0.154591) |\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) | 0.463936 / 0.215209 (0.248727) | 4.709621 / 2.077655 (2.631966) | 2.093844 / 1.504120 (0.589724) | 1.897963 / 1.541195 (0.356768) | 1.927865 / 1.468490 (0.459375) | 0.610879 / 4.584777 (-3.973898) | 4.481370 / 3.745712 (0.735658) | 2.112235 / 5.269862 (-3.157627) | 1.203349 / 4.565676 (-3.362327) | 0.074828 / 0.424275 (-0.349447) | 0.013121 / 0.007607 (0.005514) | 0.580894 / 0.226044 (0.354849) | 5.801872 / 2.268929 (3.532943) | 2.579950 / 55.444624 (-52.864674) | 2.251569 / 6.876477 (-4.624908) | 2.421305 / 2.142072 (0.279232) | 0.760938 / 4.805227 (-4.044289) | 0.169554 / 6.500664 (-6.331110) | 0.077499 / 0.075469 (0.002030) |\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) | 1.410419 / 1.841788 (-0.431368) | 17.442331 / 8.074308 (9.368023) | 15.782183 / 10.191392 (5.590791) | 0.180649 / 0.680424 (-0.499775) | 0.021790 / 0.534201 (-0.512411) | 0.511040 / 0.579283 (-0.068243) | 0.510472 / 0.434364 (0.076108) | 0.607141 / 0.540337 (0.066804) | 0.724794 / 1.386936 (-0.662142) |\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) | 0.007280 / 0.011353 (-0.004073) | 0.004712 / 0.011008 (-0.006296) | 0.089225 / 0.038508 (0.050717) | 0.053157 / 0.023109 (0.030048) | 0.431949 / 0.275898 (0.156051) | 0.478128 / 0.323480 (0.154648) | 0.006181 / 0.007986 (-0.001804) | 0.003387 / 0.004328 (-0.000941) | 0.083741 / 0.004250 (0.079490) | 0.071610 / 0.037052 (0.034557) | 0.414698 / 0.258489 (0.156209) | 0.484422 / 0.293841 (0.190581) | 0.034988 / 0.128546 (-0.093558) | 0.009831 / 0.075646 (-0.065816) | 0.089644 / 0.419271 (-0.329628) | 0.057053 / 0.043533 (0.013520) | 0.413144 / 0.255139 (0.158005) | 0.445464 / 0.283200 (0.162264) | 0.026109 / 0.141683 (-0.115574) | 1.842899 / 1.452155 (0.390745) | 1.923774 / 1.492716 (0.431057) |\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) | 0.245051 / 0.018006 (0.227045) | 0.460444 / 0.000490 (0.459954) | 0.000444 / 0.000200 (0.000244) | 0.000067 / 0.000054 (0.000012) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.034835 / 0.037411 (-0.002577) | 0.130078 / 0.014526 (0.115553) | 0.147012 / 0.176557 (-0.029544) | 0.203097 / 0.737135 (-0.534038) | 0.149636 / 0.296338 (-0.146702) |\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) | 0.521664 / 0.215209 (0.306455) | 5.283865 / 2.077655 (3.206210) | 2.456701 / 1.504120 (0.952581) | 2.266059 / 1.541195 (0.724864) | 2.295387 / 1.468490 (0.826897) | 0.613200 / 4.584777 (-3.971577) | 4.526107 / 3.745712 (0.780394) | 2.047327 / 5.269862 (-3.222535) | 1.261063 / 4.565676 (-3.304614) | 0.070402 / 0.424275 (-0.353873) | 0.014128 / 0.007607 (0.006521) | 0.620929 / 0.226044 (0.394884) | 6.109127 / 2.268929 (3.840198) | 3.081406 / 55.444624 (-52.363218) | 2.658224 / 6.876477 (-4.218253) | 2.671974 / 2.142072 (0.529902) | 0.744081 / 4.805227 (-4.061146) | 0.161498 / 6.500664 (-6.339166) | 0.075148 / 0.075469 (-0.000321) |\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) | 1.585640 / 1.841788 (-0.256148) | 17.884321 / 8.074308 (9.810013) | 15.938937 / 10.191392 (5.747545) | 0.220818 / 0.680424 (-0.459605) | 0.021452 / 0.534201 (-0.512749) | 0.499747 / 0.579283 (-0.079536) | 0.512318 / 0.434364 (0.077954) | 0.562853 / 0.540337 (0.022515) | 0.678512 / 1.386936 (-0.708424) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#aa50937d82256827aee3dbd749c7a23555e05e38 \"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" } ]

[ "Hi @zijwang, thanks for reporting this issue.\r\n\r\nYou did not mention which `datasets` version you are using, but looking at the code in the stack trace, it seems you are using an old version.\r\n\r\nCould you please update `datasets` (`pip install -U datasets`) and check if the problem persists?" ]

[ { "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" } ]

[ "cc @patrickvonplaten?", "Does anybody know if there is a bundled link, which would allow direct data download instead of manual? \r\nSomething similar to: `https://voice-prod-bundler-ee1969a6ce8178826482b88e843c335139bd3fb4.s3.amazonaws.com/cv-corpus-6.1-2020-12-11/ab.tar.gz` ? cc @patil-suraj \r\n", "Also see: https://github.com/common-voice/common-voice-bundler/issues/15" ]

[]

[ "In case it's relevant for this PR, I'm finding that I cannot stream the `bookcorpus` dataset (using the `master` branch of `datasets`), which is a `.tar.bz2` file:\r\n\r\n```python\r\nfrom datasets import load_dataset\r\n\r\nbooks_dataset_streamed = load_dataset(\"bookcorpus\", split=\"train\", streaming=True)\r\n# Throws a 404 HTTP error\r\nnext(iter(books_dataset_streamed))\r\n```\r\n\r\nThe full stack trace is:\r\n\r\n```\r\n---------------------------------------------------------------------------\r\nClientResponseError Traceback (most recent call last)\r\n<ipython-input-11-5ebbbe110b13> in <module>()\r\n----> 1 next(iter(books_dataset_streamed))\r\n\r\n11 frames\r\n/usr/local/lib/python3.7/dist-packages/datasets/iterable_dataset.py in __iter__(self)\r\n 339 \r\n 340 def __iter__(self):\r\n--> 341 for key, example in self._iter():\r\n 342 if self.features:\r\n 343 # we encode the example for ClassLabel feature types for example\r\n\r\n/usr/local/lib/python3.7/dist-packages/datasets/iterable_dataset.py in _iter(self)\r\n 336 else:\r\n 337 ex_iterable = self._ex_iterable\r\n--> 338 yield from ex_iterable\r\n 339 \r\n 340 def __iter__(self):\r\n\r\n/usr/local/lib/python3.7/dist-packages/datasets/iterable_dataset.py in __iter__(self)\r\n 76 \r\n 77 def __iter__(self):\r\n---> 78 for key, example in self.generate_examples_fn(**self.kwargs):\r\n 79 yield key, example\r\n 80 \r\n\r\n/root/.cache/huggingface/modules/datasets_modules/datasets/bookcorpus/44662c4a114441c35200992bea923b170e6f13f2f0beb7c14e43759cec498700/bookcorpus.py in _generate_examples(self, directory)\r\n 98 for txt_file in files:\r\n 99 with open(txt_file, mode=\"r\", encoding=\"utf-8\") as f:\r\n--> 100 for line in f:\r\n 101 yield _id, {\"text\": line.strip()}\r\n 102 _id += 1\r\n\r\n/usr/local/lib/python3.7/dist-packages/fsspec/implementations/http.py in read(self, length)\r\n 496 else:\r\n 497 length = min(self.size - self.loc, length)\r\n--> 498 return super().read(length)\r\n 499 \r\n 500 async def async_fetch_all(self):\r\n\r\n/usr/local/lib/python3.7/dist-packages/fsspec/spec.py in read(self, length)\r\n 1481 # don't even bother calling fetch\r\n 1482 return b\"\"\r\n-> 1483 out = self.cache._fetch(self.loc, self.loc + length)\r\n 1484 self.loc += len(out)\r\n 1485 return out\r\n\r\n/usr/local/lib/python3.7/dist-packages/fsspec/caching.py in _fetch(self, start, end)\r\n 374 ):\r\n 375 # First read, or extending both before and after\r\n--> 376 self.cache = self.fetcher(start, bend)\r\n 377 self.start = start\r\n 378 elif start < self.start:\r\n\r\n/usr/local/lib/python3.7/dist-packages/fsspec/asyn.py in wrapper(*args, **kwargs)\r\n 86 def wrapper(*args, **kwargs):\r\n 87 self = obj or args[0]\r\n---> 88 return sync(self.loop, func, *args, **kwargs)\r\n 89 \r\n 90 return wrapper\r\n\r\n/usr/local/lib/python3.7/dist-packages/fsspec/asyn.py in sync(loop, func, timeout, *args, **kwargs)\r\n 67 raise FSTimeoutError\r\n 68 if isinstance(result[0], BaseException):\r\n---> 69 raise result[0]\r\n 70 return result[0]\r\n 71 \r\n\r\n/usr/local/lib/python3.7/dist-packages/fsspec/asyn.py in _runner(event, coro, result, timeout)\r\n 23 coro = asyncio.wait_for(coro, timeout=timeout)\r\n 24 try:\r\n---> 25 result[0] = await coro\r\n 26 except Exception as ex:\r\n 27 result[0] = ex\r\n\r\n/usr/local/lib/python3.7/dist-packages/fsspec/implementations/http.py in async_fetch_range(self, start, end)\r\n 535 # range request outside file\r\n 536 return b\"\"\r\n--> 537 r.raise_for_status()\r\n 538 if r.status == 206:\r\n 539 # partial content, as expected\r\n\r\n/usr/local/lib/python3.7/dist-packages/aiohttp/client_reqrep.py in raise_for_status(self)\r\n 1003 status=self.status,\r\n 1004 message=self.reason,\r\n-> 1005 headers=self.headers,\r\n 1006 )\r\n 1007 \r\n\r\nClientResponseError: 404, message='Not Found', url=URL('https://storage.googleapis.com/huggingface-nlp/datasets/bookcorpus/bookcorpus.tar.bz2/books_large_p1.txt')\r\n```\r\n\r\nLet me know if this is unrelated and I'll open a separate issue :)\r\n\r\nEnvironment info:\r\n\r\n```\r\n- `datasets` version: 1.11.1.dev0\r\n- Platform: Linux-5.4.104+-x86_64-with-Ubuntu-18.04-bionic\r\n- Python version: 3.7.11\r\n- PyArrow version: 3.0.0\r\n```", "@lewtun: `.tar.compression-extension` files are not supported yet. That is the objective of this PR.", "> @lewtun: `.tar.compression-extension` files are not supported yet. That is the objective of this PR.\r\n\r\nthanks for the context and the great work on the streaming features (right now i'm writing the streaming section of the HF course, so am acting like a beta tester 😄)", "@lewtun this PR fixes previous issue with xjoin:\r\n\r\nGiven:\r\n```python\r\nxjoin(\r\n \"https://storage.googleapis.com/huggingface-nlp/datasets/bookcorpus/bookcorpus.tar.bz2\",\r\n \"books_large_p1.txt\"\r\n)\r\n```\r\n\r\n- Before it gave: \r\n `\"https://storage.googleapis.com/huggingface-nlp/datasets/bookcorpus/bookcorpus.tar.bz2/books_large_p1.txt\"`\r\n thus raising the 404 error\r\n\r\n- Now it gives:\r\n `tar://books_large_p1.txt::https://storage.googleapis.com/huggingface-nlp/datasets/bookcorpus/bookcorpus.tar.bz2`\r\n (this is the expected format for `fsspec`) and additionally passes the parameter `compression=\"bz2\"`.\r\n See: https://github.com/huggingface/datasets/pull/2806/files#diff-97bb2d08db65ce3b679aefc43cadad76d053c1e58ecc315e49b80873d0fbdabeR15", "closing in favor of #3066 " ]

[]

[ "hi @lhoestq, I wanted to push another change to this branch b/c I found a bug in the parsing. I need to swap arg1 and arg2. I tried to git push -u origin nell but it didn't work. So I tried to do git push --force -u origin nell which seems to work, but nothing is happening to this branch. I think this is because it's closed. Do I need to open another PR?\r\n\r\nThe change should be below in _generate_examples:\r\n best_arg1 = row[9]\r\n best_arg2 = row[8]\r\n", "Hi @ontocord !\r\n\r\nYup the easiest thing to do is to open a new PR with a title like \"Fix NELL dataset argument order\". If it's a simple fix we can look at it pretty fast :) " ]

[]

[]

[]

[]

[]

[]

[ { "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" } ]

[ "cc @aconneau ;) ", "These dataset files are no longer available. https://data.statmt.org/cc-100/ files provided in this link are no longer available. Can anybody fix that issue?\r\n@abhishekkrthakur @yjernite ", "Hi ! Can you open an issue to report this problem ? This will help keep track of the fix :)", "Ok" ]

[ { "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 the suggestion, I didn't know about this decorator.\r\n\r\nIf you are interesting in contributing, feel free to open a pull request to add the overload methods for each typing combination :) To assign you to this issue, you can comment `#self-assign` in this thread.\r\n\r\n`Dataset.__getitem__` is defined right here: https://github.com/huggingface/datasets/blob/e6f1352fe19679de897f3d962e616936a17094f5/src/datasets/arrow_dataset.py#L1840", "#self-assign" ]

[]

[ "Yes looks good to me.\r\nNote that we may refactor quite strongly the `info.py` to make it a lot simpler (it's very complicated for basically a dictionary of info I think)", "No, problem ^^ It might just be a temporary fix :)", "Fixed." ]

[]

[]

[]

[ "The pytest suite doesn't allow manual downloads so we just make sure that the `datasets-cli test` command to run without errors instead", "@lhoestq Changes made. Thank you for the review. I've made some same mistakes for https://github.com/huggingface/datasets/pull/1253 too. Will fix them before review." ]

[]

[ "_The documentation is not available anymore as the PR was closed or merged._", "To showcase the current issue, here's a Colab Gist, that shows that the `imdb` dataset cannot be read/iterated, since one or more samples contain a non-ascii character that is being converted to `numpy.bytes_`, and so on fails.\r\n\r\nColab Gist at https://gist.github.com/alvarobartt/1746959d1abb9a33e0c593f3bd82a2fb\r\n\r\nAlso, here's a quick sample of what's happening:\r\n\r\n```python\r\nfrom datasets import load_dataset\r\n\r\nds = load_dataset(\"imdb\", split=\"train\")\r\ntfds = ds.to_tf_dataset(batch_size=16)\r\nfor batch in tfds:\r\n print(batch)\r\n>>> UnicodeEncodeError: 'ascii' codec can't encode character '\\xe9' in position 0: ordinal not in range(128)\r\n```\r\n\r\nA more detailed version of it:\r\n\r\n```python\r\nfrom datasets import Dataset\r\n\r\nds = Dataset.from_dict(\r\n {\r\n \"a\": [1],\r\n \"b\": [\"é\"],\r\n }\r\n)\r\ntfds = ds.to_tf_dataset(batch_size=1)\r\nfor batch in tfds:\r\n print(batch)\r\n>>> UnicodeEncodeError: 'ascii' codec can't encode character '\\xe9' in position 0: ordinal not in range(128)\r\n```\r\n\r\nThe original issue comes from https://github.com/tensorflow/tensorflow/blob/388d952114e59a1aeda440ed4737b29f8b7c6e8a/tensorflow/python/ops/script_ops.py#LL234C4-L234C4, which could easily be solved by replacing that line with `return result.astype(np.unicode_)` but they are mentioning that it may lead to issues.\r\n\r\nEven the following fails in `numpy`:\r\n\r\n```python\r\nimport numpy as np\r\n\r\nx = np.array([\"é\"]).astype(np.bytes_)\r\n```", "cc. @lhoestq :hugs:", "cc @Rocketknight1 ", "> Nice ! Could you add some tests to make sure that batch_size=None works as expected ?\r\n\r\nSure, I'll add the tests for everything, including the string-encoding issue to make sure it's solved!", "Thanks for the review @lhoestq and @Rocketknight1! I do understand that processing it in batches is always more efficient than processing it one-by-one, it was just to make `batch_size` optional. What we can do is default it to a certain batch size e.g. 16 as before, and that's it, but I think it can still remain optional.", "@Rocketknight1 then I'll add the integration tests for the optional `batch_size` as well as for the encoding of non-ASCII compatible characters 😄 Do we set the default `batch_size` to 16 instead of `None`?", "@alvarobartt I think 16 is a reasonable default, yep!", "I think default should be None, not 16.\r\nUsers won't expect to have it batched by default.", "Then I'll leave it as is, and add the unit/integration tests, thanks @Rocketknight1 and @lhoestq ", "Hi @Rocketknight1 @lhoestq! So the string-encoding issue is already solved, but I've got one doubt about the `batch_size` being optional in the multiprocessing approach, since in that case I assume the `batch_size` should be mandatory, for the moment I'm assuming it is/should be mandatory, but let me know if you want me to add a check to disallow `batch_size=None` when `num_workers>1`. Thanks!", "> To showcase the current issue, here's a Colab Gist, that shows that the `imdb` dataset cannot be read/iterated, since one or more samples contain a non-ascii character that is being converted to `numpy.bytes_`, and so on fails.\r\n> \r\n> Colab Gist at https://gist.github.com/alvarobartt/1746959d1abb9a33e0c593f3bd82a2fb\r\n\r\nI've used the Colab shared above for testing purposes, and it works fine, plus the unit/integration tests are passing. I've also trained a `KerasNLP` model with incoming data from 🤗`datasets` with no issue at all!", "> in the multiprocessing approach, since in that case I assume the batch_size should be mandatory,\r\n\r\nNo I think they're quite orthogonal, no need to have it mandatory", "> No I think they're quite orthogonal, no need to have it mandatory\r\n\r\nBut it will break if `batch_size=None` as the multiprocessing approach will aim to prepare batches and distribute those to every worker, and assuming `batch_size=1` when `batch_size=None` I guess is not a good assumption, right?", "Ah I see. Multiprocessing should support batch_size=None indeed. If you have ideas you can do it in this PR, or raise a NotImplementedError and we can see later", "Sure @lhoestq, I can add a `NotImplementedError` for the moment, and prepare the next PR straight-away to tackle the multiprocessing approach with `batch_size=None`, but not sure if that may eventually collide with @Rocketknight1 PR at https://github.com/huggingface/datasets/pull/5863", "Yes, let me merge the PR at #5863 after this one, and then we can open another to improve the behaviour with multiprocessing and `batch_size=None`!", "Sure @Rocketknight1 makes complete sense to me! Do you want me to add the `raise NotImplementedError` and then we merge this PR? Or you prefer to directly merge the current?", "`raise NotImplementedError` for now with an error telling the user that multiprocessing needs them to specify a batch size, I think!", "Since you recently approved @Rocketknight1, are we ready to merge? Thanks 🤗", "Ah actually it looks like `minimal_tf_collate_fn` doesn't support batch_size=None", "Hi @lhoestq so I didn't include the call to `collate_fn`, as we won't need to collate the incoming data e.g. \"str\" should remain a \"str\" not a [\"str\"], and the `minimal_collate_fn` was indeed putting everything into a list, so the output was not un-batched, but batched with size 1", "What if the user passes a collate_fn ? The torch DataLoader still applies it if batch_size=None for example.\r\n\r\nDoes my last change look of to you ? If so I think we can merge", "> What if the user passes a collate_fn ? The torch DataLoader still applies it if batch_size=None for example.\r\n> \r\n> Does my last change look of to you ? If so I think we can merge\r\n\r\nI think we're good, since it won't batch it under the scenario of `str` being provided instead of `List[str]`, and the unit/integration tests are passing, so I'm OK to merge. Maybe we can double check with Matt? cc @Rocketknight1 ", "Yes, and sorry for the delay! I'm happy to merge.", "<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) | 0.006555 / 0.011353 (-0.004798) | 0.004521 / 0.011008 (-0.006487) | 0.096633 / 0.038508 (0.058125) | 0.032859 / 0.023109 (0.009750) | 0.294632 / 0.275898 (0.018734) | 0.325140 / 0.323480 (0.001660) | 0.005676 / 0.007986 (-0.002310) | 0.005252 / 0.004328 (0.000924) | 0.074349 / 0.004250 (0.070099) | 0.045836 / 0.037052 (0.008784) | 0.302919 / 0.258489 (0.044430) | 0.340686 / 0.293841 (0.046845) | 0.028398 / 0.128546 (-0.100148) | 0.008942 / 0.075646 (-0.066704) | 0.326994 / 0.419271 (-0.092278) | 0.049556 / 0.043533 (0.006023) | 0.293883 / 0.255139 (0.038744) | 0.316522 / 0.283200 (0.033322) | 0.097385 / 0.141683 (-0.044298) | 1.405334 / 1.452155 (-0.046821) | 1.521529 / 1.492716 (0.028812) |\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) | 0.212269 / 0.018006 (0.194263) | 0.445692 / 0.000490 (0.445203) | 0.004930 / 0.000200 (0.004730) | 0.000093 / 0.000054 (0.000039) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.026907 / 0.037411 (-0.010504) | 0.108607 / 0.014526 (0.094081) | 0.116806 / 0.176557 (-0.059751) | 0.178428 / 0.737135 (-0.558707) | 0.122326 / 0.296338 (-0.174012) |\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) | 0.404211 / 0.215209 (0.189002) | 4.045374 / 2.077655 (1.967719) | 1.877237 / 1.504120 (0.373117) | 1.706276 / 1.541195 (0.165081) | 1.750610 / 1.468490 (0.282120) | 0.522331 / 4.584777 (-4.062446) | 3.742286 / 3.745712 (-0.003426) | 1.791285 / 5.269862 (-3.478577) | 1.043872 / 4.565676 (-3.521805) | 0.065176 / 0.424275 (-0.359099) | 0.011821 / 0.007607 (0.004214) | 0.507374 / 0.226044 (0.281329) | 5.088803 / 2.268929 (2.819875) | 2.282742 / 55.444624 (-53.161882) | 1.950737 / 6.876477 (-4.925740) | 2.042262 / 2.142072 (-0.099810) | 0.636525 / 4.805227 (-4.168702) | 0.140837 / 6.500664 (-6.359827) | 0.063223 / 0.075469 (-0.012246) |\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) | 1.188070 / 1.841788 (-0.653718) | 14.622681 / 8.074308 (6.548372) | 13.247988 / 10.191392 (3.056596) | 0.165858 / 0.680424 (-0.514566) | 0.017476 / 0.534201 (-0.516725) | 0.391973 / 0.579283 (-0.187310) | 0.433326 / 0.434364 (-0.001038) | 0.467163 / 0.540337 (-0.073175) | 0.568359 / 1.386936 (-0.818577) |\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) | 0.006076 / 0.011353 (-0.005276) | 0.004439 / 0.011008 (-0.006570) | 0.074496 / 0.038508 (0.035988) | 0.031396 / 0.023109 (0.008287) | 0.372237 / 0.275898 (0.096339) | 0.403412 / 0.323480 (0.079932) | 0.005430 / 0.007986 (-0.002555) | 0.003846 / 0.004328 (-0.000483) | 0.074403 / 0.004250 (0.070153) | 0.045398 / 0.037052 (0.008346) | 0.394133 / 0.258489 (0.135644) | 0.421769 / 0.293841 (0.127928) | 0.027936 / 0.128546 (-0.100610) | 0.008962 / 0.075646 (-0.066685) | 0.083158 / 0.419271 (-0.336113) | 0.044863 / 0.043533 (0.001331) | 0.393834 / 0.255139 (0.138695) | 0.391537 / 0.283200 (0.108337) | 0.097971 / 0.141683 (-0.043712) | 1.496632 / 1.452155 (0.044477) | 1.585511 / 1.492716 (0.092795) |\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) | 0.010094 / 0.018006 (-0.007913) | 0.437811 / 0.000490 (0.437321) | 0.000963 / 0.000200 (0.000763) | 0.000084 / 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) | 0.028864 / 0.037411 (-0.008547) | 0.112480 / 0.014526 (0.097954) | 0.120938 / 0.176557 (-0.055619) | 0.170888 / 0.737135 (-0.566247) | 0.125903 / 0.296338 (-0.170435) |\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) | 0.426716 / 0.215209 (0.211507) | 4.238380 / 2.077655 (2.160725) | 2.052889 / 1.504120 (0.548769) | 1.871043 / 1.541195 (0.329848) | 1.890405 / 1.468490 (0.421915) | 0.522059 / 4.584777 (-4.062718) | 3.813331 / 3.745712 (0.067619) | 2.891651 / 5.269862 (-2.378210) | 1.323836 / 4.565676 (-3.241841) | 0.065124 / 0.424275 (-0.359151) | 0.011498 / 0.007607 (0.003891) | 0.525102 / 0.226044 (0.299057) | 5.245190 / 2.268929 (2.976261) | 2.531149 / 55.444624 (-52.913476) | 2.197323 / 6.876477 (-4.679153) | 2.197314 / 2.142072 (0.055241) | 0.633423 / 4.805227 (-4.171804) | 0.140248 / 6.500664 (-6.360416) | 0.064432 / 0.075469 (-0.011037) |\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) | 1.270639 / 1.841788 (-0.571149) | 14.856678 / 8.074308 (6.782369) | 14.337631 / 10.191392 (4.146239) | 0.195319 / 0.680424 (-0.485105) | 0.017628 / 0.534201 (-0.516573) | 0.393984 / 0.579283 (-0.185299) | 0.421987 / 0.434364 (-0.012376) | 0.459245 / 0.540337 (-0.081092) | 0.557786 / 1.386936 (-0.829150) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#a129219a48c1b07c06d4bc1db32c317bf513089d \"CML watermark\")\n", "Will you eventually need help with your PR @Rocketknight1? I'll be happy to help if needed 😄 ", "<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) | 0.007577 / 0.011353 (-0.003776) | 0.004960 / 0.011008 (-0.006048) | 0.113622 / 0.038508 (0.075114) | 0.037981 / 0.023109 (0.014872) | 0.355312 / 0.275898 (0.079414) | 0.393384 / 0.323480 (0.069904) | 0.006575 / 0.007986 (-0.001411) | 0.005941 / 0.004328 (0.001612) | 0.085976 / 0.004250 (0.081726) | 0.053784 / 0.037052 (0.016732) | 0.369358 / 0.258489 (0.110869) | 0.399402 / 0.293841 (0.105561) | 0.032155 / 0.128546 (-0.096391) | 0.010448 / 0.075646 (-0.065199) | 0.389009 / 0.419271 (-0.030263) | 0.057377 / 0.043533 (0.013844) | 0.354968 / 0.255139 (0.099829) | 0.382404 / 0.283200 (0.099204) | 0.111056 / 0.141683 (-0.030627) | 1.807986 / 1.452155 (0.355832) | 1.866070 / 1.492716 (0.373354) |\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) | 0.244439 / 0.018006 (0.226432) | 0.491942 / 0.000490 (0.491452) | 0.001910 / 0.000200 (0.001710) | 0.000112 / 0.000054 (0.000058) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.031024 / 0.037411 (-0.006387) | 0.129674 / 0.014526 (0.115148) | 0.142974 / 0.176557 (-0.033583) | 0.213568 / 0.737135 (-0.523568) | 0.147794 / 0.296338 (-0.148545) |\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) | 0.480333 / 0.215209 (0.265124) | 4.792901 / 2.077655 (2.715246) | 2.233145 / 1.504120 (0.729025) | 2.036291 / 1.541195 (0.495096) | 2.109631 / 1.468490 (0.641140) | 0.624546 / 4.584777 (-3.960231) | 4.543511 / 3.745712 (0.797799) | 3.961345 / 5.269862 (-1.308517) | 1.903634 / 4.565676 (-2.662042) | 0.076584 / 0.424275 (-0.347691) | 0.014590 / 0.007607 (0.006983) | 0.593195 / 0.226044 (0.367151) | 5.928740 / 2.268929 (3.659811) | 2.781164 / 55.444624 (-52.663460) | 2.364303 / 6.876477 (-4.512173) | 2.510139 / 2.142072 (0.368067) | 0.770886 / 4.805227 (-4.034341) | 0.167995 / 6.500664 (-6.332669) | 0.076622 / 0.075469 (0.001153) |\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) | 1.402398 / 1.841788 (-0.439390) | 17.921233 / 8.074308 (9.846925) | 17.036738 / 10.191392 (6.845346) | 0.168997 / 0.680424 (-0.511427) | 0.020259 / 0.534201 (-0.513941) | 0.465322 / 0.579283 (-0.113962) | 0.500435 / 0.434364 (0.066071) | 0.546846 / 0.540337 (0.006509) | 0.658130 / 1.386936 (-0.728806) |\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) | 0.007624 / 0.011353 (-0.003729) | 0.005265 / 0.011008 (-0.005744) | 0.086886 / 0.038508 (0.048377) | 0.038235 / 0.023109 (0.015126) | 0.463969 / 0.275898 (0.188071) | 0.502451 / 0.323480 (0.178971) | 0.006285 / 0.007986 (-0.001701) | 0.004525 / 0.004328 (0.000197) | 0.086557 / 0.004250 (0.082307) | 0.052414 / 0.037052 (0.015362) | 0.482167 / 0.258489 (0.223678) | 0.513684 / 0.293841 (0.219843) | 0.032929 / 0.128546 (-0.095618) | 0.010249 / 0.075646 (-0.065397) | 0.093377 / 0.419271 (-0.325895) | 0.054114 / 0.043533 (0.010582) | 0.466116 / 0.255139 (0.210977) | 0.488977 / 0.283200 (0.205777) | 0.115446 / 0.141683 (-0.026237) | 1.762912 / 1.452155 (0.310757) | 1.874191 / 1.492716 (0.381475) |\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) | 0.012666 / 0.018006 (-0.005341) | 0.485962 / 0.000490 (0.485473) | 0.002621 / 0.000200 (0.002421) | 0.000128 / 0.000054 (0.000074) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.033661 / 0.037411 (-0.003751) | 0.135395 / 0.014526 (0.120869) | 0.147230 / 0.176557 (-0.029326) | 0.205847 / 0.737135 (-0.531288) | 0.151496 / 0.296338 (-0.144842) |\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) | 0.514097 / 0.215209 (0.298887) | 5.134093 / 2.077655 (3.056438) | 2.496775 / 1.504120 (0.992655) | 2.268078 / 1.541195 (0.726883) | 2.342153 / 1.468490 (0.873663) | 0.623130 / 4.584777 (-3.961647) | 4.601787 / 3.745712 (0.856075) | 3.414249 / 5.269862 (-1.855613) | 1.849603 / 4.565676 (-2.716073) | 0.078350 / 0.424275 (-0.345925) | 0.013785 / 0.007607 (0.006178) | 0.638783 / 0.226044 (0.412739) | 6.378356 / 2.268929 (4.109427) | 3.072867 / 55.444624 (-52.371757) | 2.668123 / 6.876477 (-4.208354) | 2.693905 / 2.142072 (0.551833) | 0.764583 / 4.805227 (-4.040644) | 0.166854 / 6.500664 (-6.333810) | 0.076883 / 0.075469 (0.001414) |\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) | 1.502003 / 1.841788 (-0.339784) | 18.674205 / 8.074308 (10.599897) | 16.837759 / 10.191392 (6.646367) | 0.176995 / 0.680424 (-0.503428) | 0.020126 / 0.534201 (-0.514075) | 0.464480 / 0.579283 (-0.114803) | 0.516477 / 0.434364 (0.082113) | 0.549818 / 0.540337 (0.009481) | 0.659927 / 1.386936 (-0.727009) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#a129219a48c1b07c06d4bc1db32c317bf513089d \"CML watermark\")\n", "@alvarobartt Yes, I'll ping you for a review once it's ready!" ]

[]

[]

[]

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

[ { "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" } ]

[]

[]

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

[]

[]

[]

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

[ { "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, \r\nI want to work on this issue and I was thinking why don't we use `imap` [in this loop](https://github.com/huggingface/datasets/blob/440b14d0dd428ae1b25881aa72ba7bbb8ad9ff84/src/datasets/io/json.py#L99)? This way, using offset (which is being used to slice the pyarrow table) we can convert pyarrow table to `json` using multiprocessing. I've a small code snippet for some clarity:\r\n```\r\nresult = list(\r\n pool.imap(self._apply_df, [(offset, batch_size) for offset in range(0, len(self.dataset), batch_size)])\r\n )\r\n```\r\n`_apply_df` is a function which will return `batch.to_pandas().to_json(path_or_buf=None, orient=\"records\", lines=True)` which is basically json version of the batched pyarrow table. Later on we can concatenate it to form json file? \r\n\r\nI think the only downside here is to write file from `imap` output (output would be a list and we'll need to iterate over it and write in a file) which might add a little overhead cost. What do you think about this?", "Followed up in https://github.com/huggingface/datasets/pull/2747" ]

[]

[]

[]

[ "Can you run `make style` to fix the code format ?\r\n\r\nAlso it looks like the file `file_downloaded/TWNERTC_TC_Coarse Grained NER_DomainIndependent_NoiseReduction.zip/TWNERTC_TC_Coarse Grained NER_DomainIndependent_NoiseReduction.DUMP` is missing inside the dummy_data.zip\r\n\r\n\r\n(note that `TWNERTC_TC_Coarse Grained NER_DomainIndependent_NoiseReduction.zip` is a directory name, not an actual zip file)", "Hi Quentin, thank you for your patience with me. I've fixed the preprocessing pipeline, got this very weird error that Yacine told me to push. I've pushed it and after I'll find out that it will work, I will have my final pr on styling.", "looks like you removed the dataset script file in your latest commit, is it expected ?" ]

[]

[]

[]

[]

[]