Datasets:

---

Francesco-A

/

github-issues_huggingface-datasets

like 0

[]

[ "Sounds great! It'll make things easier for the user while accessing the dataset. I'll make some changes to the current file then.", "I've updated with the suggested changes. Updated the README, YAML tags as well (not sure of Size category tag as I couldn't pass the path of `dataset_infos.json` for this dataset)\r\n" ]

[]

[]

[]

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

[]

[ "Hi @haonan-li , thank you for the report! It seems to be a bug on the [`huggingface_hub`](https://github.com/huggingface/huggingface_hub) site, there is even no such dataset as `mbzuai/bactrian-x` on the Hub. I opened and [issue](https://github.com/huggingface/huggingface_hub/issues/1453) there.", "I think `load_dataset(\"mbzuai/bactrian-x\")` shouldn't be loaded at all and raise an error but because of [this fallback](https://github.com/huggingface/datasets/blob/main/src/datasets/load.py#L1194) to packaged loaders when no other options are applicable, it loads the dataset with standard `json` loader instead of the custom loading script." ]

[]

[ "It seems that selecting a subset of colums directly from the dataset, i.e., dataset[\"column\"], is slow.", "I was wrong, ```dataset[\"column\"]``` is fast." ]

[]

[ "I guess,test failing due to connection error?", "We just fixed the other dataset on master. Could you rebase from master and push to rerun the CI ?", "If I'm not wrong this function returns None if no main class was found.\r\nI think it makes things less clear not to have a return at the end of the function.\r\nI guess we can have one return in the for loop instead of the break statement, AND one return at the end to explicitly return None.\r\nWhat do you think ?" ]

[]

[ "@lhoestq @albertvillanova Iwas trying to fix the failing tests in circleCI but is there a test elasticsearch instance somewhere? If not, can I launch a docker container to have one?" ]

[]

[ "_The documentation is not available anymore as the PR was closed or merged._", "WDYT @albertvillanova ?", "The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/datasets/pr_5211). All of your documentation changes will be reflected on that endpoint." ]

[]

[]

[]

[ "<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) | 0.007003 / 0.011353 (-0.004350) | 0.004854 / 0.011008 (-0.006154) | 0.096982 / 0.038508 (0.058474) | 0.033218 / 0.023109 (0.010109) | 0.314088 / 0.275898 (0.038190) | 0.351315 / 0.323480 (0.027835) | 0.005679 / 0.007986 (-0.002307) | 0.005404 / 0.004328 (0.001075) | 0.071773 / 0.004250 (0.067522) | 0.044593 / 0.037052 (0.007540) | 0.323643 / 0.258489 (0.065154) | 0.357172 / 0.293841 (0.063331) | 0.036782 / 0.128546 (-0.091764) | 0.012146 / 0.075646 (-0.063501) | 0.334874 / 0.419271 (-0.084397) | 0.051475 / 0.043533 (0.007942) | 0.305949 / 0.255139 (0.050810) | 0.339326 / 0.283200 (0.056126) | 0.101509 / 0.141683 (-0.040174) | 1.458254 / 1.452155 (0.006099) | 1.535252 / 1.492716 (0.042535) |\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) | 0.264837 / 0.018006 (0.246831) | 0.441444 / 0.000490 (0.440955) | 0.003331 / 0.000200 (0.003131) | 0.000084 / 0.000054 (0.000030) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.026529 / 0.037411 (-0.010882) | 0.105924 / 0.014526 (0.091398) | 0.117191 / 0.176557 (-0.059365) | 0.176606 / 0.737135 (-0.560529) | 0.123452 / 0.296338 (-0.172887) |\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) | 0.412351 / 0.215209 (0.197142) | 4.135468 / 2.077655 (2.057813) | 1.912820 / 1.504120 (0.408700) | 1.738993 / 1.541195 (0.197798) | 1.754228 / 1.468490 (0.285738) | 0.692239 / 4.584777 (-3.892538) | 3.765672 / 3.745712 (0.019959) | 2.081141 / 5.269862 (-3.188720) | 1.425153 / 4.565676 (-3.140523) | 0.085055 / 0.424275 (-0.339220) | 0.011918 / 0.007607 (0.004311) | 0.517573 / 0.226044 (0.291529) | 5.179809 / 2.268929 (2.910881) | 2.471620 / 55.444624 (-52.973005) | 2.140634 / 6.876477 (-4.735843) | 2.200150 / 2.142072 (0.058077) | 0.831662 / 4.805227 (-3.973566) | 0.168828 / 6.500664 (-6.331836) | 0.062755 / 0.075469 (-0.012714) |\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) | 1.196890 / 1.841788 (-0.644898) | 14.826423 / 8.074308 (6.752114) | 14.020782 / 10.191392 (3.829390) | 0.161275 / 0.680424 (-0.519149) | 0.017467 / 0.534201 (-0.516734) | 0.422278 / 0.579283 (-0.157005) | 0.424053 / 0.434364 (-0.010311) | 0.490768 / 0.540337 (-0.049570) | 0.584490 / 1.386936 (-0.802446) |\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) | 0.007102 / 0.011353 (-0.004250) | 0.005145 / 0.011008 (-0.005863) | 0.073823 / 0.038508 (0.035315) | 0.032947 / 0.023109 (0.009838) | 0.336978 / 0.275898 (0.061080) | 0.368961 / 0.323480 (0.045481) | 0.006052 / 0.007986 (-0.001934) | 0.003970 / 0.004328 (-0.000358) | 0.072925 / 0.004250 (0.068674) | 0.044502 / 0.037052 (0.007450) | 0.340849 / 0.258489 (0.082360) | 0.381487 / 0.293841 (0.087646) | 0.037207 / 0.128546 (-0.091339) | 0.012095 / 0.075646 (-0.063551) | 0.085206 / 0.419271 (-0.334065) | 0.056236 / 0.043533 (0.012703) | 0.334048 / 0.255139 (0.078909) | 0.360442 / 0.283200 (0.077242) | 0.104402 / 0.141683 (-0.037281) | 1.446907 / 1.452155 (-0.005248) | 1.542430 / 1.492716 (0.049713) |\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) | 0.238720 / 0.018006 (0.220714) | 0.445857 / 0.000490 (0.445367) | 0.009280 / 0.000200 (0.009080) | 0.000150 / 0.000054 (0.000095) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.028414 / 0.037411 (-0.008998) | 0.110506 / 0.014526 (0.095981) | 0.124593 / 0.176557 (-0.051964) | 0.170951 / 0.737135 (-0.566184) | 0.128033 / 0.296338 (-0.168305) |\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) | 0.426206 / 0.215209 (0.210997) | 4.267289 / 2.077655 (2.189634) | 2.026880 / 1.504120 (0.522760) | 1.844052 / 1.541195 (0.302858) | 1.897697 / 1.468490 (0.429207) | 0.713545 / 4.584777 (-3.871232) | 3.815052 / 3.745712 (0.069339) | 3.217091 / 5.269862 (-2.052770) | 1.790546 / 4.565676 (-2.775130) | 0.087501 / 0.424275 (-0.336774) | 0.012136 / 0.007607 (0.004529) | 0.534495 / 0.226044 (0.308451) | 5.325913 / 2.268929 (3.056984) | 2.484309 / 55.444624 (-52.960315) | 2.149721 / 6.876477 (-4.726756) | 2.158764 / 2.142072 (0.016692) | 0.855273 / 4.805227 (-3.949954) | 0.170374 / 6.500664 (-6.330290) | 0.064053 / 0.075469 (-0.011416) |\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) | 1.253171 / 1.841788 (-0.588617) | 15.254562 / 8.074308 (7.180254) | 14.242119 / 10.191392 (4.050727) | 0.159298 / 0.680424 (-0.521126) | 0.017504 / 0.534201 (-0.516696) | 0.419710 / 0.579283 (-0.159574) | 0.417879 / 0.434364 (-0.016485) | 0.486328 / 0.540337 (-0.054009) | 0.578933 / 1.386936 (-0.808003) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#bc38663c8e2c2b0b246791c3ed8bddbff163dd64 \"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) | 0.008476 / 0.011353 (-0.002877) | 0.005745 / 0.011008 (-0.005263) | 0.115307 / 0.038508 (0.076799) | 0.039356 / 0.023109 (0.016247) | 0.367155 / 0.275898 (0.091257) | 0.422147 / 0.323480 (0.098667) | 0.006817 / 0.007986 (-0.001168) | 0.004652 / 0.004328 (0.000323) | 0.084045 / 0.004250 (0.079795) | 0.055483 / 0.037052 (0.018431) | 0.364249 / 0.258489 (0.105760) | 0.415975 / 0.293841 (0.122134) | 0.041322 / 0.128546 (-0.087224) | 0.014178 / 0.075646 (-0.061469) | 0.392658 / 0.419271 (-0.026614) | 0.060156 / 0.043533 (0.016623) | 0.373938 / 0.255139 (0.118799) | 0.397494 / 0.283200 (0.114294) | 0.113811 / 0.141683 (-0.027872) | 1.688581 / 1.452155 (0.236427) | 1.790374 / 1.492716 (0.297658) |\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) | 0.222203 / 0.018006 (0.204196) | 0.471109 / 0.000490 (0.470619) | 0.007071 / 0.000200 (0.006871) | 0.000156 / 0.000054 (0.000102) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.032112 / 0.037411 (-0.005299) | 0.118726 / 0.014526 (0.104200) | 0.134918 / 0.176557 (-0.041639) | 0.207766 / 0.737135 (-0.529369) | 0.139756 / 0.296338 (-0.156582) |\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) | 0.479858 / 0.215209 (0.264649) | 4.798428 / 2.077655 (2.720773) | 2.221573 / 1.504120 (0.717453) | 1.964956 / 1.541195 (0.423761) | 2.021763 / 1.468490 (0.553273) | 0.820401 / 4.584777 (-3.764376) | 4.533887 / 3.745712 (0.788175) | 4.121332 / 5.269862 (-1.148529) | 2.195807 / 4.565676 (-2.369869) | 0.103133 / 0.424275 (-0.321142) | 0.014620 / 0.007607 (0.007013) | 0.605012 / 0.226044 (0.378967) | 5.966623 / 2.268929 (3.697694) | 2.844118 / 55.444624 (-52.600506) | 2.463569 / 6.876477 (-4.412907) | 2.597177 / 2.142072 (0.455105) | 0.983201 / 4.805227 (-3.822026) | 0.199500 / 6.500664 (-6.301164) | 0.078387 / 0.075469 (0.002918) |\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) | 1.401083 / 1.841788 (-0.440705) | 17.258725 / 8.074308 (9.184417) | 16.825992 / 10.191392 (6.634600) | 0.216762 / 0.680424 (-0.463662) | 0.021135 / 0.534201 (-0.513066) | 0.513688 / 0.579283 (-0.065595) | 0.488892 / 0.434364 (0.054529) | 0.566745 / 0.540337 (0.026408) | 0.688958 / 1.386936 (-0.697978) |\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) | 0.007948 / 0.011353 (-0.003405) | 0.005981 / 0.011008 (-0.005027) | 0.084474 / 0.038508 (0.045966) | 0.037952 / 0.023109 (0.014843) | 0.383359 / 0.275898 (0.107461) | 0.409324 / 0.323480 (0.085844) | 0.006641 / 0.007986 (-0.001344) | 0.004785 / 0.004328 (0.000456) | 0.083214 / 0.004250 (0.078964) | 0.053177 / 0.037052 (0.016125) | 0.393147 / 0.258489 (0.134658) | 0.438496 / 0.293841 (0.144655) | 0.042090 / 0.128546 (-0.086456) | 0.013373 / 0.075646 (-0.062273) | 0.097585 / 0.419271 (-0.321686) | 0.056359 / 0.043533 (0.012826) | 0.378113 / 0.255139 (0.122974) | 0.403874 / 0.283200 (0.120674) | 0.123503 / 0.141683 (-0.018180) | 1.639557 / 1.452155 (0.187403) | 1.759787 / 1.492716 (0.267071) |\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) | 0.242534 / 0.018006 (0.224528) | 0.459040 / 0.000490 (0.458550) | 0.000454 / 0.000200 (0.000254) | 0.000066 / 0.000054 (0.000011) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.031747 / 0.037411 (-0.005664) | 0.125823 / 0.014526 (0.111297) | 0.138985 / 0.176557 (-0.037571) | 0.194371 / 0.737135 (-0.542764) | 0.148905 / 0.296338 (-0.147433) |\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) | 0.508201 / 0.215209 (0.292992) | 5.007519 / 2.077655 (2.929865) | 2.412956 / 1.504120 (0.908836) | 2.143378 / 1.541195 (0.602183) | 2.192966 / 1.468490 (0.724476) | 0.828497 / 4.584777 (-3.756280) | 4.496457 / 3.745712 (0.750745) | 2.397546 / 5.269862 (-2.872315) | 1.522889 / 4.565676 (-3.042787) | 0.099904 / 0.424275 (-0.324371) | 0.014561 / 0.007607 (0.006954) | 0.627417 / 0.226044 (0.401373) | 6.296441 / 2.268929 (4.027512) | 2.962858 / 55.444624 (-52.481767) | 2.543083 / 6.876477 (-4.333394) | 2.711884 / 2.142072 (0.569811) | 0.997969 / 4.805227 (-3.807259) | 0.200283 / 6.500664 (-6.300382) | 0.075934 / 0.075469 (0.000465) |\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) | 1.541707 / 1.841788 (-0.300081) | 17.791559 / 8.074308 (9.717251) | 16.782877 / 10.191392 (6.591485) | 0.171954 / 0.680424 (-0.508470) | 0.020506 / 0.534201 (-0.513695) | 0.504189 / 0.579283 (-0.075094) | 0.501655 / 0.434364 (0.067291) | 0.583120 / 0.540337 (0.042782) | 0.694931 / 1.386936 (-0.692005) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#53355f308f4ffb9b4071f5d420b5c6767799ef1c \"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) | 0.007613 / 0.011353 (-0.003740) | 0.005057 / 0.011008 (-0.005951) | 0.099147 / 0.038508 (0.060639) | 0.035358 / 0.023109 (0.012249) | 0.303442 / 0.275898 (0.027544) | 0.336898 / 0.323480 (0.013418) | 0.006216 / 0.007986 (-0.001770) | 0.004085 / 0.004328 (-0.000244) | 0.074567 / 0.004250 (0.070317) | 0.050917 / 0.037052 (0.013865) | 0.301786 / 0.258489 (0.043297) | 0.341362 / 0.293841 (0.047521) | 0.037019 / 0.128546 (-0.091528) | 0.011977 / 0.075646 (-0.063669) | 0.334688 / 0.419271 (-0.084583) | 0.051326 / 0.043533 (0.007793) | 0.299878 / 0.255139 (0.044739) | 0.325571 / 0.283200 (0.042371) | 0.110744 / 0.141683 (-0.030939) | 1.480898 / 1.452155 (0.028743) | 1.566917 / 1.492716 (0.074201) |\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) | 0.253249 / 0.018006 (0.235242) | 0.558576 / 0.000490 (0.558086) | 0.003838 / 0.000200 (0.003638) | 0.000085 / 0.000054 (0.000030) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.028731 / 0.037411 (-0.008681) | 0.110643 / 0.014526 (0.096117) | 0.119560 / 0.176557 (-0.056996) | 0.178010 / 0.737135 (-0.559126) | 0.130286 / 0.296338 (-0.166053) |\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) | 0.400190 / 0.215209 (0.184981) | 3.999326 / 2.077655 (1.921672) | 1.797332 / 1.504120 (0.293212) | 1.610808 / 1.541195 (0.069613) | 1.679949 / 1.468490 (0.211459) | 0.696539 / 4.584777 (-3.888238) | 3.784766 / 3.745712 (0.039054) | 2.205008 / 5.269862 (-3.064854) | 1.501697 / 4.565676 (-3.063979) | 0.085553 / 0.424275 (-0.338723) | 0.012223 / 0.007607 (0.004616) | 0.494858 / 0.226044 (0.268813) | 4.968535 / 2.268929 (2.699606) | 2.258759 / 55.444624 (-53.185865) | 1.926236 / 6.876477 (-4.950241) | 2.072155 / 2.142072 (-0.069917) | 0.838354 / 4.805227 (-3.966873) | 0.168810 / 6.500664 (-6.331854) | 0.064347 / 0.075469 (-0.011122) |\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) | 1.166696 / 1.841788 (-0.675091) | 14.721287 / 8.074308 (6.646979) | 14.319272 / 10.191392 (4.127880) | 0.144534 / 0.680424 (-0.535890) | 0.017502 / 0.534201 (-0.516699) | 0.422682 / 0.579283 (-0.156601) | 0.424426 / 0.434364 (-0.009938) | 0.493561 / 0.540337 (-0.046777) | 0.586765 / 1.386936 (-0.800171) |\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) | 0.007764 / 0.011353 (-0.003589) | 0.005516 / 0.011008 (-0.005492) | 0.074745 / 0.038508 (0.036237) | 0.034364 / 0.023109 (0.011255) | 0.344318 / 0.275898 (0.068420) | 0.374779 / 0.323480 (0.051299) | 0.005904 / 0.007986 (-0.002082) | 0.004323 / 0.004328 (-0.000005) | 0.073191 / 0.004250 (0.068941) | 0.051549 / 0.037052 (0.014496) | 0.341792 / 0.258489 (0.083303) | 0.387576 / 0.293841 (0.093735) | 0.037483 / 0.128546 (-0.091063) | 0.012410 / 0.075646 (-0.063237) | 0.086480 / 0.419271 (-0.332791) | 0.050035 / 0.043533 (0.006502) | 0.335475 / 0.255139 (0.080336) | 0.361436 / 0.283200 (0.078236) | 0.106890 / 0.141683 (-0.034792) | 1.464032 / 1.452155 (0.011877) | 1.563490 / 1.492716 (0.070774) |\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) | 0.268765 / 0.018006 (0.250758) | 0.563811 / 0.000490 (0.563321) | 0.004904 / 0.000200 (0.004704) | 0.000096 / 0.000054 (0.000041) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.029885 / 0.037411 (-0.007526) | 0.113885 / 0.014526 (0.099359) | 0.124283 / 0.176557 (-0.052274) | 0.173619 / 0.737135 (-0.563517) | 0.131781 / 0.296338 (-0.164557) |\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) | 0.420296 / 0.215209 (0.205087) | 4.167656 / 2.077655 (2.090001) | 1.982356 / 1.504120 (0.478237) | 1.792181 / 1.541195 (0.250986) | 1.871459 / 1.468490 (0.402969) | 0.707066 / 4.584777 (-3.877711) | 3.835922 / 3.745712 (0.090210) | 3.506796 / 5.269862 (-1.763066) | 1.857172 / 4.565676 (-2.708505) | 0.086219 / 0.424275 (-0.338056) | 0.012404 / 0.007607 (0.004796) | 0.512393 / 0.226044 (0.286348) | 5.111623 / 2.268929 (2.842695) | 2.493523 / 55.444624 (-52.951101) | 2.188220 / 6.876477 (-4.688257) | 2.319096 / 2.142072 (0.177024) | 0.844084 / 4.805227 (-3.961144) | 0.171130 / 6.500664 (-6.329534) | 0.065913 / 0.075469 (-0.009556) |\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) | 1.284768 / 1.841788 (-0.557020) | 15.334610 / 8.074308 (7.260301) | 14.724436 / 10.191392 (4.533044) | 0.188425 / 0.680424 (-0.491999) | 0.017984 / 0.534201 (-0.516217) | 0.428150 / 0.579283 (-0.151133) | 0.429013 / 0.434364 (-0.005351) | 0.500818 / 0.540337 (-0.039519) | 0.592879 / 1.386936 (-0.794057) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#ee68da958c2fab3a26d9f0efb1e207ecbcf7ce15 \"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) | 0.006870 / 0.011353 (-0.004483) | 0.004702 / 0.011008 (-0.006306) | 0.099258 / 0.038508 (0.060750) | 0.029008 / 0.023109 (0.005899) | 0.330599 / 0.275898 (0.054701) | 0.361163 / 0.323480 (0.037683) | 0.005020 / 0.007986 (-0.002965) | 0.003474 / 0.004328 (-0.000855) | 0.075902 / 0.004250 (0.071651) | 0.037462 / 0.037052 (0.000410) | 0.336213 / 0.258489 (0.077724) | 0.370645 / 0.293841 (0.076804) | 0.032435 / 0.128546 (-0.096111) | 0.011686 / 0.075646 (-0.063960) | 0.326040 / 0.419271 (-0.093232) | 0.043750 / 0.043533 (0.000217) | 0.332629 / 0.255139 (0.077490) | 0.353302 / 0.283200 (0.070102) | 0.090421 / 0.141683 (-0.051262) | 1.470097 / 1.452155 (0.017942) | 1.544908 / 1.492716 (0.052191) |\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) | 0.213418 / 0.018006 (0.195411) | 0.434808 / 0.000490 (0.434319) | 0.005949 / 0.000200 (0.005749) | 0.000072 / 0.000054 (0.000018) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.023085 / 0.037411 (-0.014327) | 0.098222 / 0.014526 (0.083696) | 0.104543 / 0.176557 (-0.072013) | 0.165423 / 0.737135 (-0.571713) | 0.108732 / 0.296338 (-0.187606) |\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) | 0.433933 / 0.215209 (0.218724) | 4.334358 / 2.077655 (2.256704) | 2.013984 / 1.504120 (0.509864) | 1.862981 / 1.541195 (0.321787) | 1.873936 / 1.468490 (0.405446) | 0.699857 / 4.584777 (-3.884920) | 3.417815 / 3.745712 (-0.327897) | 1.946403 / 5.269862 (-3.323459) | 1.308683 / 4.565676 (-3.256994) | 0.083297 / 0.424275 (-0.340978) | 0.012610 / 0.007607 (0.005003) | 0.540877 / 0.226044 (0.314832) | 5.408293 / 2.268929 (3.139365) | 2.529574 / 55.444624 (-52.915050) | 2.201047 / 6.876477 (-4.675429) | 2.392966 / 2.142072 (0.250894) | 0.812719 / 4.805227 (-3.992509) | 0.154013 / 6.500664 (-6.346651) | 0.067614 / 0.075469 (-0.007855) |\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) | 1.228150 / 1.841788 (-0.613638) | 14.037090 / 8.074308 (5.962782) | 14.259416 / 10.191392 (4.068024) | 0.155554 / 0.680424 (-0.524870) | 0.016521 / 0.534201 (-0.517680) | 0.379615 / 0.579283 (-0.199668) | 0.421352 / 0.434364 (-0.013012) | 0.446512 / 0.540337 (-0.093825) | 0.531802 / 1.386936 (-0.855134) |\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) | 0.006629 / 0.011353 (-0.004724) | 0.004432 / 0.011008 (-0.006577) | 0.076662 / 0.038508 (0.038154) | 0.027674 / 0.023109 (0.004565) | 0.341667 / 0.275898 (0.065769) | 0.376493 / 0.323480 (0.053014) | 0.005076 / 0.007986 (-0.002910) | 0.004655 / 0.004328 (0.000326) | 0.075698 / 0.004250 (0.071448) | 0.036905 / 0.037052 (-0.000147) | 0.342394 / 0.258489 (0.083905) | 0.383330 / 0.293841 (0.089489) | 0.031729 / 0.128546 (-0.096817) | 0.011582 / 0.075646 (-0.064064) | 0.085721 / 0.419271 (-0.333551) | 0.042012 / 0.043533 (-0.001521) | 0.342063 / 0.255139 (0.086924) | 0.367335 / 0.283200 (0.084136) | 0.089641 / 0.141683 (-0.052042) | 1.520353 / 1.452155 (0.068198) | 1.643653 / 1.492716 (0.150937) |\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) | 0.178995 / 0.018006 (0.160989) | 0.436544 / 0.000490 (0.436055) | 0.002311 / 0.000200 (0.002111) | 0.000081 / 0.000054 (0.000026) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.025386 / 0.037411 (-0.012026) | 0.099717 / 0.014526 (0.085192) | 0.110809 / 0.176557 (-0.065747) | 0.162931 / 0.737135 (-0.574204) | 0.110430 / 0.296338 (-0.185909) |\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) | 0.438592 / 0.215209 (0.223382) | 4.372560 / 2.077655 (2.294905) | 2.069686 / 1.504120 (0.565567) | 1.860576 / 1.541195 (0.319382) | 1.898161 / 1.468490 (0.429671) | 0.698353 / 4.584777 (-3.886424) | 3.462440 / 3.745712 (-0.283272) | 1.868602 / 5.269862 (-3.401260) | 1.160498 / 4.565676 (-3.405179) | 0.082869 / 0.424275 (-0.341406) | 0.012690 / 0.007607 (0.005083) | 0.533278 / 0.226044 (0.307233) | 5.386214 / 2.268929 (3.117285) | 2.519243 / 55.444624 (-52.925382) | 2.171109 / 6.876477 (-4.705368) | 2.272617 / 2.142072 (0.130544) | 0.805843 / 4.805227 (-3.999384) | 0.152275 / 6.500664 (-6.348389) | 0.068038 / 0.075469 (-0.007431) |\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) | 1.291967 / 1.841788 (-0.549821) | 14.386474 / 8.074308 (6.312166) | 14.180693 / 10.191392 (3.989301) | 0.131714 / 0.680424 (-0.548710) | 0.016596 / 0.534201 (-0.517605) | 0.384293 / 0.579283 (-0.194990) | 0.404051 / 0.434364 (-0.030313) | 0.452167 / 0.540337 (-0.088170) | 0.542718 / 1.386936 (-0.844218) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#f9c770bb1a43fa7fe390286d7535266d3964d067 \"CML watermark\")\n" ]

[]

[ "cc'ing @mapmeld ", "Full language list is now deployed to https://huggingface.co/datasets ! Recommend close", "Cool @mapmeld ! My 2 cents (for a next iteration), it would be cool to have a small search widget in the filter dropdown as you have a ton of languages now here! Closing this in the meantime." ]

[]

[]

[]

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

[]

[]

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

[ "Hi,\r\n\r\nThe `use_fast_tokenizer` argument has been recently added to the bert-score lib. I've opened a PR with the fix. In the meantime, you can try to downgrade the version of bert-score with the following command to make the code work:\r\n```\r\npip uninstall bert-score\r\npip install \"bert-score<0.3.10\"\r\n```" ]

[]

[ "There is an exception when running `pip install .[tests]`:\r\n```\r\nProcessing /home/circleci/datasets\r\nCollecting numpy>=1.17 (from datasets==1.12.2.dev0)\r\n Downloading https://files.pythonhosted.org/packages/45/b2/6c7545bb7a38754d63048c7696804a0d947328125d81bf12beaa692c3ae3/numpy-1.19.5-cp36-cp36m-manylinux1_x86_64.whl (13.4MB)\r\n 100% |████████████████████████████████| 13.4MB 3.9MB/s eta 0:00:011\r\n\r\n...\r\n\r\nCollecting faiss-cpu (from datasets==1.12.2.dev0)\r\n Downloading https://files.pythonhosted.org/packages/87/91/bf8ea0d42733cbb04f98d3bf27808e4919ceb5ec71102e21119398a97237/faiss-cpu-1.7.1.post2.tar.gz (41kB)\r\n 100% |████████████████████████████████| 51kB 30.9MB/s ta 0:00:01\r\n Complete output from command python setup.py egg_info:\r\n Traceback (most recent call last):\r\n File \"/home/circleci/.pyenv/versions/3.6.14/lib/python3.6/site-packages/setuptools/sandbox.py\", line 154, in save_modules\r\n yield saved\r\n File \"/home/circleci/.pyenv/versions/3.6.14/lib/python3.6/site-packages/setuptools/sandbox.py\", line 195, in setup_context\r\n yield\r\n File \"/home/circleci/.pyenv/versions/3.6.14/lib/python3.6/site-packages/setuptools/sandbox.py\", line 250, in run_setup\r\n _execfile(setup_script, ns)\r\n File \"/home/circleci/.pyenv/versions/3.6.14/lib/python3.6/site-packages/setuptools/sandbox.py\", line 45, in _execfile\r\n exec(code, globals, locals)\r\n File \"/tmp/easy_install-1pop4blm/numpy-1.21.2/setup.py\", line 34, in <module>\r\n method can be invoked.\r\n RuntimeError: Python version >= 3.7 required.\r\n```\r\n\r\nApparently, `numpy-1.21.2` tries to be installed in the temporary directory `/tmp/easy_install-1pop4blm` instead of the downloaded `numpy-1.19.5` (requirement of `datasets`).\r\n\r\nThis is caused because `pip` downloads the `.tar.gz` (instead of the `.whl`) and tries to build it in a tmp dir." ]

[]

[ "You can just run: \r\n`val = nlp.load_dataset('squad')` \r\n\r\nif you want to have just the validation script you can also do:\r\n\r\n`val = nlp.load_dataset('squad', split=\"validation\")`", "If you want to load a local dataset, make sure you include a `./` before the folder name. ", "This happens by just doing run all cells on colab ... I assumed the colab example is broken. ", "Oh I see you might have a wrong version of pyarrow install on the colab -> could you try the following. Add these lines to the beginning of your notebook, restart the runtime and run it again:\r\n```\r\n!pip uninstall -y -qq pyarrow\r\n!pip uninstall -y -qq nlp\r\n!pip install -qq git+https://github.com/huggingface/nlp.git\r\n```", "> Oh I see you might have a wrong version of pyarrow install on the colab -> could you try the following. Add these lines to the beginning of your notebook, restart the runtime and run it again:\r\n> \r\n> ```\r\n> !pip uninstall -y -qq pyarrow\r\n> !pip uninstall -y -qq nlp\r\n> !pip install -qq git+https://github.com/huggingface/nlp.git\r\n> ```\r\n\r\nTried, having the same error.", "Can you post a link here of your colab? I'll make a copy of it and see what's wrong", "This should be fixed in the current version of the notebook. You can try it again", "Also see: https://github.com/huggingface/nlp/issues/222", "I am getting this error when running this command\r\n```\r\nval = nlp.load_dataset('squad', split=\"validation\")\r\n```\r\n\r\nFileNotFoundError: [Errno 2] No such file or directory: '/root/.cache/huggingface/datasets/squad/plain_text/1.0.0/dataset_info.json'\r\n\r\nCan anybody help?", "It seems like your download was corrupted :-/ Can you run the following command: \r\n\r\n```\r\nrm -r /root/.cache/huggingface/datasets\r\n```\r\n\r\nto delete the cache completely and rerun the download? ", "I tried the notebook again today and it worked without barfing. 👌 " ]

[]

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

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

[ "I'm running into this as well. (Thank you so much for reporting @jorgeecardona — was staring at this massive stack trace and unsure what exactly was wrong!)", "Hi ! Thanks for reporting :)\r\n\r\nUntil this is fixed on `dill`'s side, we could implement a custom saving in our Pickler indefined in utils.py_utils.py\r\nThere is already a suggestion in this message about how to do it:\r\nhttps://github.com/uqfoundation/dill/issues/250#issuecomment-852566284\r\n\r\nLet me know if such a workaround could help, and feel free to open a PR if you want to contribute !", "I have the same bug.\r\nthe code is as follows:\r\n\r\nthe error is: \r\n\r\n\r\nLook for the solution for this bug.", "Hi ! I think your RecursionError comes from a different issue @BitcoinNLPer , could you open a separate issue please ?\r\n\r\nAlso which dataset are you using ? I tried loading `CodedotAI/code_clippy` but I get a different error\r\n```python\r\nTraceback (most recent call last):\r\n File \"<stdin>\", line 1, in <module>\r\n File \"/Users/quentinlhoest/Desktop/hf/datasets/src/datasets/load.py\", line 1615, in load_dataset\r\n **config_kwargs,\r\n File \"/Users/quentinlhoest/Desktop/hf/datasets/src/datasets/load.py\", line 1446, in load_dataset_builder\r\n builder_cls = import_main_class(dataset_module.module_path)\r\n File \"/Users/quentinlhoest/Desktop/hf/datasets/src/datasets/load.py\", line 101, in import_main_class\r\n module = importlib.import_module(module_path)\r\n File \"/Users/quentinlhoest/.virtualenvs/hf-datasets/lib/python3.7/importlib/__init__.py\", line 127, in import_module\r\n return _bootstrap._gcd_import(name[level:], package, level)\r\n File \"<frozen importlib._bootstrap>\", line 1006, in _gcd_import\r\n File \"<frozen importlib._bootstrap>\", line 983, in _find_and_load\r\n File \"<frozen importlib._bootstrap>\", line 967, in _find_and_load_unlocked\r\n File \"<frozen importlib._bootstrap>\", line 677, in _load_unlocked\r\n File \"<frozen importlib._bootstrap_external>\", line 728, in exec_module\r\n File \"<frozen importlib._bootstrap>\", line 219, in _call_with_frames_removed\r\n File \"/Users/quentinlhoest/.cache/huggingface/modules/datasets_modules/datasets/CodedotAI___code_clippy/d332f69d036e8c80f47bc9a96d676c3fa30cb50af7bb81e2d4d12e80b83efc4d/code_clippy.py\", line 66, in <module>\r\n url_elements = results.find_all(\"a\")\r\nAttributeError: 'NoneType' object has no attribute 'find_all'\r\n```" ]

[]

[ "@lhoestq Thank you for your code review! I think I could do the necessary corrections. Could you please check it again when you have time?", "Thank you for merging!" ]

[]

[]

[]

[]

[]

[]

[]

[]

[]

[ "_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) | 0.007220 / 0.011353 (-0.004133) | 0.004558 / 0.011008 (-0.006451) | 0.116647 / 0.038508 (0.078139) | 0.046845 / 0.023109 (0.023736) | 0.352429 / 0.275898 (0.076531) | 0.429739 / 0.323480 (0.106259) | 0.006620 / 0.007986 (-0.001366) | 0.003731 / 0.004328 (-0.000597) | 0.088683 / 0.004250 (0.084433) | 0.070583 / 0.037052 (0.033530) | 0.366699 / 0.258489 (0.108210) | 0.420730 / 0.293841 (0.126889) | 0.037342 / 0.128546 (-0.091204) | 0.010041 / 0.075646 (-0.065605) | 0.383477 / 0.419271 (-0.035795) | 0.060279 / 0.043533 (0.016746) | 0.349988 / 0.255139 (0.094849) | 0.371423 / 0.283200 (0.088224) | 0.026725 / 0.141683 (-0.114958) | 1.736886 / 1.452155 (0.284731) | 1.812874 / 1.492716 (0.320157) |\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) | 0.253256 / 0.018006 (0.235250) | 0.563470 / 0.000490 (0.562980) | 0.010475 / 0.000200 (0.010275) | 0.000164 / 0.000054 (0.000110) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.030518 / 0.037411 (-0.006893) | 0.133324 / 0.014526 (0.118798) | 0.137095 / 0.176557 (-0.039461) | 0.202227 / 0.737135 (-0.534909) | 0.144195 / 0.296338 (-0.152143) |\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) | 0.480870 / 0.215209 (0.265661) | 4.822713 / 2.077655 (2.745058) | 2.124183 / 1.504120 (0.620064) | 1.910733 / 1.541195 (0.369538) | 1.970266 / 1.468490 (0.501776) | 0.624695 / 4.584777 (-3.960082) | 4.459659 / 3.745712 (0.713947) | 2.210123 / 5.269862 (-3.059739) | 1.300520 / 4.565676 (-3.265157) | 0.077096 / 0.424275 (-0.347180) | 0.013333 / 0.007607 (0.005726) | 0.596841 / 0.226044 (0.370797) | 5.917397 / 2.268929 (3.648469) | 2.699397 / 55.444624 (-52.745228) | 2.274833 / 6.876477 (-4.601644) | 2.525376 / 2.142072 (0.383304) | 0.755718 / 4.805227 (-4.049510) | 0.163587 / 6.500664 (-6.337077) | 0.072817 / 0.075469 (-0.002653) |\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) | 1.524306 / 1.841788 (-0.317481) | 18.843312 / 8.074308 (10.769004) | 15.694644 / 10.191392 (5.503252) | 0.177400 / 0.680424 (-0.503024) | 0.020104 / 0.534201 (-0.514097) | 0.466421 / 0.579283 (-0.112862) | 0.537274 / 0.434364 (0.102910) | 0.576920 / 0.540337 (0.036583) | 0.718889 / 1.386936 (-0.668047) |\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) | 0.007671 / 0.011353 (-0.003682) | 0.004850 / 0.011008 (-0.006158) | 0.090085 / 0.038508 (0.051576) | 0.052023 / 0.023109 (0.028914) | 0.508575 / 0.275898 (0.232677) | 0.590024 / 0.323480 (0.266544) | 0.004564 / 0.007986 (-0.003422) | 0.005345 / 0.004328 (0.001017) | 0.087904 / 0.004250 (0.083653) | 0.064446 / 0.037052 (0.027394) | 0.525625 / 0.258489 (0.267136) | 0.584307 / 0.293841 (0.290466) | 0.037221 / 0.128546 (-0.091325) | 0.010588 / 0.075646 (-0.065059) | 0.098612 / 0.419271 (-0.320659) | 0.059597 / 0.043533 (0.016064) | 0.488064 / 0.255139 (0.232925) | 0.522330 / 0.283200 (0.239131) | 0.030004 / 0.141683 (-0.111679) | 1.732512 / 1.452155 (0.280357) | 1.809027 / 1.492716 (0.316310) |\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) | 0.218741 / 0.018006 (0.200735) | 0.494946 / 0.000490 (0.494456) | 0.004580 / 0.000200 (0.004380) | 0.000104 / 0.000054 (0.000049) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.034916 / 0.037411 (-0.002495) | 0.133695 / 0.014526 (0.119169) | 0.147964 / 0.176557 (-0.028592) | 0.213210 / 0.737135 (-0.523926) | 0.148850 / 0.296338 (-0.147488) |\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) | 0.508855 / 0.215209 (0.293646) | 5.065088 / 2.077655 (2.987433) | 2.473110 / 1.504120 (0.968990) | 2.259765 / 1.541195 (0.718570) | 2.359189 / 1.468490 (0.890699) | 0.639082 / 4.584777 (-3.945695) | 4.768195 / 3.745712 (1.022482) | 2.253803 / 5.269862 (-3.016059) | 1.442996 / 4.565676 (-3.122680) | 0.078761 / 0.424275 (-0.345514) | 0.013936 / 0.007607 (0.006329) | 0.625977 / 0.226044 (0.399933) | 6.260817 / 2.268929 (3.991888) | 3.149640 / 55.444624 (-52.294985) | 2.753555 / 6.876477 (-4.122921) | 2.831872 / 2.142072 (0.689799) | 0.781294 / 4.805227 (-4.023933) | 0.169109 / 6.500664 (-6.331555) | 0.075810 / 0.075469 (0.000341) |\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) | 1.533282 / 1.841788 (-0.308506) | 19.460579 / 8.074308 (11.386271) | 17.250424 / 10.191392 (7.059032) | 0.193485 / 0.680424 (-0.486939) | 0.020650 / 0.534201 (-0.513551) | 0.472110 / 0.579283 (-0.107173) | 0.532276 / 0.434364 (0.097912) | 0.613152 / 0.540337 (0.072814) | 0.684684 / 1.386936 (-0.702252) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#650a86ee122209d4a8c8e8068c01ebfd3ba553f5 \"CML watermark\")\n" ]

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

[ "Hi,\r\n\r\nyou can define your own custom formatting with `Dataset.set_transform()` and then run the tokenizer with the batches of augmented data as follows:\r\n```python\r\ndset = load_dataset(\"imdb\", split=\"train\") # Let's say we are working with the IMDB dataset\r\ndset.set_transform(lambda ex: {\"text\": augly_text_augmentation(ex[\"text\"])}, columns=\"text\", output_all_columns=True)\r\ndataloader = torch.utils.data.DataLoader(dset, batch_size=32)\r\nfor epoch in range(5):\r\n for batch in dataloader:\r\n tokenizer_output = tokenizer(batch.pop(\"text\"), padding=True, truncation=True, return_tensors=\"pt\")\r\n batch.update(tokenizer_output)\r\n output = model(**batch)\r\n ...\r\n```", "Preprocessing functions/augmentations, unless super generic, should be defined in separate libraries, so I'm closing this issue." ]

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

[]

[ { "color": "E5583E", "default": false, "description": "Related to the dataset viewer on huggingface.co", "id": 3470211881, "name": "dataset-viewer", "node_id": "LA_kwDODunzps7O1zsp", "url": "https://api.github.com/repos/huggingface/datasets/labels/dataset-viewer" } ]

[ "Hi! A dataset has to be streamable to work with the viewer. I did a quick test, and yours is, so this might be a bug in the viewer. cc @severo \r\n", "Looking at it. The message is now:\r\n\r\n```\r\nMessage: cannot cache function '__shear_dense': no locator available for file '/src/services/worker/.venv/lib/python3.9/site-packages/librosa/util/utils.py'\r\n```\r\n\r\nso possibly it's related to the libraries versions?\r\n", "Maybe this SO thread can help: https://stackoverflow.com/questions/59290386/runtimeerror-at-cannot-cache-function-shear-dense-no-locator-available-fo", "Same error for https://huggingface.co/datasets/LIUM/tedlium/viewer/release1/test. cc @sanchit-gandhi . I'm on it", "Fixed in the datasets viewer, by setting the `NUMBA_CACHE_DIR` env var to a writable directory.", "https://huggingface.co/datasets/Roh/ryanspeech/viewer/male/train\r\n\r\n<img width=\"1538\" alt=\"Capture d’écran 2022-06-08 à 11 30 08\" src=\"https://user-images.githubusercontent.com/1676121/172583285-4cd49a0f-5715-423b-95dd-5f6ace3b2416.png\">\r\n", "https://huggingface.co/datasets/LIUM/tedlium/viewer/\r\n\r\n<img width=\"1538\" alt=\"Capture d’écran 2022-06-08 à 14 31 52\" src=\"https://user-images.githubusercontent.com/1676121/172616897-fbcb7df7-0308-4d09-a17d-48826bc91374.png\">\r\n" ]

[]

[ "@lhoestq thank you very much for the quick review and useful comments! \r\n\r\nI have tried to address them all, and a few comments that you left for ro_sts I have applied to the ro_sts_parallel as well (in read-me: fixed source_datasets, links to homepage, repository, leaderboard, thanks to me message, in ro_sts_parallel.py changed to camel case as well). In the ro_sts_parallel I have changed the order on the languages, also in the example, as you said order doesn't matter, but just to have them listed in the readme in the same order.\r\n\r\nI have commented above on why we would like to keep them as separate datasets, hope it makes sense.\r\n\r\nIf there is anything else I should change please let me know.\r\n\r\nThanks again!", "@lhoestq I tried to adjust the ro_sts_parallel, locally when I run the tests they are passing, but somewhere it has the old name of rosts-parallel-ro-en which I am trying to change to ro_sts_parallel. I don't think I have left anything related to rosts-parallel-ro-en, but when the dataset_infos.json is regenerated it adds it. Could you please help me out, how can I fix this? Thanks in advance!", "Great, thanks for all your help! " ]

[]

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

[ { "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 ! :) cc @andersjohanandreassen can you take a look at this ?\r\n\r\nAlso @cceyda feel free to open an issue at [BIG-Bench](https://github.com/google/BIG-bench) as well regarding the `AttributeError`", "I'm on vacation for the next week, so won't be able to do much debugging at the moment. Sorry for the inconvenience.\r\nBut I did quickly take a look:\r\n\r\n**pypi**:\r\nI managed to reproduce the above error with the pypi version begin out of date. \r\nThe version on `https://storage.googleapis.com/public_research_data/bigbench/bigbench-0.0.1.tar.gz` should be up to date, but it was my understanding that there was some issue with the pypi upload, so I don't even understand why there is a version [on pypi from April 1](https://pypi.org/project/bigbench/0.0.1/). Perhaps @ethansdyer, who's handling the pypi upload, knows the answer to that?\r\n\r\n**OOM error**:\r\nBut, I'm unable to reproduce the OOM error in a google colab with GPU enabled.\r\nThis is what I ran:\r\n```\r\n!pip install bigbench@https://storage.googleapis.com/public_research_data/bigbench/bigbench-0.0.1.tar.gz\r\n!pip install datasets\r\n\r\nfrom datasets import load_dataset\r\ndataset = load_dataset(\"bigbench\",\"swedish_to_german_proverbs\")\r\n``` \r\nThe `swedish_to_german_proverbs`task is only 72 examples, so I don't understand what could be causing the OOM error. Loading the task has no effect on the RAM for me. @cceyda Can you confirm that this does not occur in a [colab](https://colab.research.google.com/)?\r\nIf the GPU is somehow causing issues on your system, disabling the GPU from TF might be an option too\r\n```\r\nimport os\r\nos.environ[\"CUDA_VISIBLE_DEVICES\"] = \"-1\"\r\n```\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n", "Solved.\r\nYes it works on colab, and somehow magically on my machine too now. hmm not sure what was wrong before I had used a fresh venv both times with just the dataloading code, and tried multiple times. (maybe just a wrong tensorflow version got mixed up somehow) The tensorflow call seems to come from the bigbench side anyway.\r\n\r\nabout bigbench pypi version update, I opened an issue over there https://github.com/google/BIG-bench/issues/846\r\n\r\nanyway closing this now. If anyone else has the same problem can re-open." ]

[]

[]

[]

[]

[]

[ "_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" } ]

[ "Thanks for reporting, @LZY-the-boys.\r\n\r\nCould you please give more details about what is your intended dataset structure? What are the names of the columns and the value of each row?\r\n\r\nCurrently, the JSON-Lines format is supported:\r\n- Each line correspond to one row of the dataset\r\n- Each line is composed of one JSON object, where the names are the names of the columns, and the values are the values for the row-column pair." ]

[]

[ "_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) | 0.007617 / 0.011353 (-0.003736) | 0.004580 / 0.011008 (-0.006428) | 0.100913 / 0.038508 (0.062405) | 0.087703 / 0.023109 (0.064594) | 0.424159 / 0.275898 (0.148261) | 0.467195 / 0.323480 (0.143715) | 0.006890 / 0.007986 (-0.001096) | 0.003765 / 0.004328 (-0.000564) | 0.077513 / 0.004250 (0.073262) | 0.064889 / 0.037052 (0.027837) | 0.422349 / 0.258489 (0.163860) | 0.477391 / 0.293841 (0.183550) | 0.036025 / 0.128546 (-0.092522) | 0.009939 / 0.075646 (-0.065707) | 0.342409 / 0.419271 (-0.076862) | 0.061568 / 0.043533 (0.018035) | 0.431070 / 0.255139 (0.175931) | 0.462008 / 0.283200 (0.178809) | 0.027480 / 0.141683 (-0.114203) | 1.802271 / 1.452155 (0.350116) | 1.861336 / 1.492716 (0.368620) |\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) | 0.255806 / 0.018006 (0.237800) | 0.507969 / 0.000490 (0.507479) | 0.010060 / 0.000200 (0.009860) | 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) | 0.032286 / 0.037411 (-0.005125) | 0.104468 / 0.014526 (0.089942) | 0.112707 / 0.176557 (-0.063850) | 0.181285 / 0.737135 (-0.555850) | 0.113180 / 0.296338 (-0.183158) |\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) | 0.449265 / 0.215209 (0.234056) | 4.465941 / 2.077655 (2.388287) | 2.177889 / 1.504120 (0.673769) | 1.969864 / 1.541195 (0.428669) | 2.077502 / 1.468490 (0.609011) | 0.561607 / 4.584777 (-4.023170) | 4.281873 / 3.745712 (0.536161) | 4.975352 / 5.269862 (-0.294510) | 2.907121 / 4.565676 (-1.658555) | 0.070205 / 0.424275 (-0.354070) | 0.009164 / 0.007607 (0.001557) | 0.581921 / 0.226044 (0.355876) | 5.538667 / 2.268929 (3.269739) | 2.798853 / 55.444624 (-52.645771) | 2.314015 / 6.876477 (-4.562462) | 2.584836 / 2.142072 (0.442763) | 0.672333 / 4.805227 (-4.132894) | 0.153828 / 6.500664 (-6.346836) | 0.069757 / 0.075469 (-0.005712) |\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) | 1.559670 / 1.841788 (-0.282118) | 23.994639 / 8.074308 (15.920331) | 16.856160 / 10.191392 (6.664768) | 0.195555 / 0.680424 (-0.484869) | 0.021586 / 0.534201 (-0.512615) | 0.469295 / 0.579283 (-0.109989) | 0.481582 / 0.434364 (0.047218) | 0.588667 / 0.540337 (0.048329) | 0.734347 / 1.386936 (-0.652589) |\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) | 0.009614 / 0.011353 (-0.001739) | 0.004616 / 0.011008 (-0.006392) | 0.077223 / 0.038508 (0.038715) | 0.103074 / 0.023109 (0.079965) | 0.447834 / 0.275898 (0.171936) | 0.524696 / 0.323480 (0.201216) | 0.007120 / 0.007986 (-0.000866) | 0.003890 / 0.004328 (-0.000438) | 0.076406 / 0.004250 (0.072156) | 0.073488 / 0.037052 (0.036436) | 0.466221 / 0.258489 (0.207732) | 0.532206 / 0.293841 (0.238365) | 0.037596 / 0.128546 (-0.090950) | 0.010029 / 0.075646 (-0.065617) | 0.084313 / 0.419271 (-0.334959) | 0.060088 / 0.043533 (0.016555) | 0.437792 / 0.255139 (0.182653) | 0.512850 / 0.283200 (0.229650) | 0.032424 / 0.141683 (-0.109259) | 1.762130 / 1.452155 (0.309975) | 1.946097 / 1.492716 (0.453381) |\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) | 0.250774 / 0.018006 (0.232768) | 0.506869 / 0.000490 (0.506379) | 0.008232 / 0.000200 (0.008032) | 0.000164 / 0.000054 (0.000110) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.037779 / 0.037411 (0.000368) | 0.111933 / 0.014526 (0.097407) | 0.122385 / 0.176557 (-0.054172) | 0.190372 / 0.737135 (-0.546763) | 0.122472 / 0.296338 (-0.173866) |\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) | 0.488502 / 0.215209 (0.273293) | 4.878114 / 2.077655 (2.800459) | 2.504144 / 1.504120 (1.000024) | 2.321077 / 1.541195 (0.779883) | 2.416797 / 1.468490 (0.948307) | 0.583582 / 4.584777 (-4.001195) | 4.277896 / 3.745712 (0.532184) | 3.874780 / 5.269862 (-1.395082) | 2.540099 / 4.565676 (-2.025577) | 0.068734 / 0.424275 (-0.355541) | 0.009158 / 0.007607 (0.001550) | 0.578401 / 0.226044 (0.352357) | 5.763354 / 2.268929 (3.494426) | 3.167771 / 55.444624 (-52.276853) | 2.675220 / 6.876477 (-4.201257) | 2.920927 / 2.142072 (0.778855) | 0.673948 / 4.805227 (-4.131280) | 0.157908 / 6.500664 (-6.342756) | 0.071672 / 0.075469 (-0.003797) |\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) | 1.635120 / 1.841788 (-0.206668) | 24.853480 / 8.074308 (16.779172) | 17.162978 / 10.191392 (6.971586) | 0.209577 / 0.680424 (-0.470847) | 0.030110 / 0.534201 (-0.504091) | 0.546970 / 0.579283 (-0.032313) | 0.581912 / 0.434364 (0.147548) | 0.571460 / 0.540337 (0.031123) | 0.823411 / 1.386936 (-0.563525) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#83b792dddd074ccd007c407f942f6870aac7ee84 \"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) | 0.006674 / 0.011353 (-0.004679) | 0.004198 / 0.011008 (-0.006810) | 0.084859 / 0.038508 (0.046351) | 0.076065 / 0.023109 (0.052955) | 0.316065 / 0.275898 (0.040167) | 0.352097 / 0.323480 (0.028617) | 0.005610 / 0.007986 (-0.002376) | 0.003600 / 0.004328 (-0.000729) | 0.064921 / 0.004250 (0.060671) | 0.054493 / 0.037052 (0.017441) | 0.318125 / 0.258489 (0.059636) | 0.370183 / 0.293841 (0.076342) | 0.031141 / 0.128546 (-0.097405) | 0.008755 / 0.075646 (-0.066891) | 0.288241 / 0.419271 (-0.131030) | 0.052379 / 0.043533 (0.008846) | 0.328147 / 0.255139 (0.073008) | 0.347548 / 0.283200 (0.064348) | 0.024393 / 0.141683 (-0.117290) | 1.480646 / 1.452155 (0.028492) | 1.575867 / 1.492716 (0.083151) |\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) | 0.268978 / 0.018006 (0.250971) | 0.586470 / 0.000490 (0.585980) | 0.003190 / 0.000200 (0.002990) | 0.000081 / 0.000054 (0.000026) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.030595 / 0.037411 (-0.006816) | 0.083037 / 0.014526 (0.068511) | 0.103706 / 0.176557 (-0.072850) | 0.164104 / 0.737135 (-0.573031) | 0.104536 / 0.296338 (-0.191802) |\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) | 0.382274 / 0.215209 (0.167065) | 3.811878 / 2.077655 (1.734223) | 1.840098 / 1.504120 (0.335978) | 1.670949 / 1.541195 (0.129754) | 1.763755 / 1.468490 (0.295264) | 0.479526 / 4.584777 (-4.105251) | 3.544443 / 3.745712 (-0.201269) | 3.263004 / 5.269862 (-2.006858) | 2.092801 / 4.565676 (-2.472875) | 0.057167 / 0.424275 (-0.367108) | 0.007450 / 0.007607 (-0.000157) | 0.463731 / 0.226044 (0.237686) | 4.624630 / 2.268929 (2.355701) | 2.327078 / 55.444624 (-53.117546) | 1.977734 / 6.876477 (-4.898743) | 2.237152 / 2.142072 (0.095079) | 0.573210 / 4.805227 (-4.232018) | 0.132095 / 6.500664 (-6.368569) | 0.060283 / 0.075469 (-0.015186) |\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) | 1.243404 / 1.841788 (-0.598384) | 20.306778 / 8.074308 (12.232470) | 14.561660 / 10.191392 (4.370268) | 0.170826 / 0.680424 (-0.509598) | 0.018574 / 0.534201 (-0.515627) | 0.392367 / 0.579283 (-0.186916) | 0.402918 / 0.434364 (-0.031446) | 0.476629 / 0.540337 (-0.063708) | 0.653709 / 1.386936 (-0.733227) |\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) | 0.006562 / 0.011353 (-0.004791) | 0.004092 / 0.011008 (-0.006916) | 0.065951 / 0.038508 (0.027443) | 0.078090 / 0.023109 (0.054981) | 0.369679 / 0.275898 (0.093781) | 0.411442 / 0.323480 (0.087962) | 0.005646 / 0.007986 (-0.002339) | 0.003537 / 0.004328 (-0.000791) | 0.066024 / 0.004250 (0.061773) | 0.058947 / 0.037052 (0.021895) | 0.389219 / 0.258489 (0.130730) | 0.414200 / 0.293841 (0.120359) | 0.030372 / 0.128546 (-0.098174) | 0.008631 / 0.075646 (-0.067015) | 0.071692 / 0.419271 (-0.347580) | 0.048035 / 0.043533 (0.004502) | 0.376960 / 0.255139 (0.121821) | 0.389847 / 0.283200 (0.106648) | 0.023940 / 0.141683 (-0.117743) | 1.487633 / 1.452155 (0.035479) | 1.561680 / 1.492716 (0.068964) |\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) | 0.301467 / 0.018006 (0.283461) | 0.544159 / 0.000490 (0.543669) | 0.000408 / 0.000200 (0.000208) | 0.000055 / 0.000054 (0.000001) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.030939 / 0.037411 (-0.006472) | 0.087432 / 0.014526 (0.072906) | 0.103263 / 0.176557 (-0.073293) | 0.154551 / 0.737135 (-0.582585) | 0.104631 / 0.296338 (-0.191707) |\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) | 0.422348 / 0.215209 (0.207139) | 4.206003 / 2.077655 (2.128348) | 2.212619 / 1.504120 (0.708499) | 2.049616 / 1.541195 (0.508421) | 2.139093 / 1.468490 (0.670603) | 0.489647 / 4.584777 (-4.095130) | 3.523291 / 3.745712 (-0.222422) | 3.277657 / 5.269862 (-1.992205) | 2.111353 / 4.565676 (-2.454324) | 0.057597 / 0.424275 (-0.366679) | 0.007675 / 0.007607 (0.000068) | 0.493068 / 0.226044 (0.267023) | 4.939493 / 2.268929 (2.670565) | 2.695995 / 55.444624 (-52.748630) | 2.374904 / 6.876477 (-4.501573) | 2.600110 / 2.142072 (0.458038) | 0.586306 / 4.805227 (-4.218921) | 0.134137 / 6.500664 (-6.366527) | 0.061897 / 0.075469 (-0.013572) |\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) | 1.330628 / 1.841788 (-0.511160) | 20.557964 / 8.074308 (12.483656) | 14.251632 / 10.191392 (4.060240) | 0.148772 / 0.680424 (-0.531652) | 0.018383 / 0.534201 (-0.515817) | 0.392552 / 0.579283 (-0.186731) | 0.403959 / 0.434364 (-0.030405) | 0.462154 / 0.540337 (-0.078184) | 0.608832 / 1.386936 (-0.778104) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#7a291b2b659a356199dff0ab004ad3845459034b \"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) | 0.007659 / 0.011353 (-0.003694) | 0.004500 / 0.011008 (-0.006508) | 0.100379 / 0.038508 (0.061871) | 0.079731 / 0.023109 (0.056622) | 0.381788 / 0.275898 (0.105890) | 0.416524 / 0.323480 (0.093044) | 0.004446 / 0.007986 (-0.003539) | 0.003752 / 0.004328 (-0.000577) | 0.074956 / 0.004250 (0.070706) | 0.062885 / 0.037052 (0.025832) | 0.383849 / 0.258489 (0.125360) | 0.433906 / 0.293841 (0.140065) | 0.036079 / 0.128546 (-0.092468) | 0.009927 / 0.075646 (-0.065719) | 0.343879 / 0.419271 (-0.075393) | 0.061055 / 0.043533 (0.017523) | 0.376703 / 0.255139 (0.121564) | 0.428111 / 0.283200 (0.144911) | 0.028667 / 0.141683 (-0.113016) | 1.777755 / 1.452155 (0.325600) | 1.878283 / 1.492716 (0.385567) |\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) | 0.220829 / 0.018006 (0.202823) | 0.506406 / 0.000490 (0.505916) | 0.005550 / 0.000200 (0.005350) | 0.000123 / 0.000054 (0.000069) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.034928 / 0.037411 (-0.002483) | 0.103873 / 0.014526 (0.089347) | 0.114352 / 0.176557 (-0.062204) | 0.188218 / 0.737135 (-0.548918) | 0.117343 / 0.296338 (-0.178995) |\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) | 0.459148 / 0.215209 (0.243939) | 4.582092 / 2.077655 (2.504437) | 2.275603 / 1.504120 (0.771483) | 2.058155 / 1.541195 (0.516960) | 2.163886 / 1.468490 (0.695396) | 0.573033 / 4.584777 (-4.011744) | 4.414891 / 3.745712 (0.669178) | 7.280433 / 5.269862 (2.010572) | 4.119414 / 4.565676 (-0.446262) | 0.067432 / 0.424275 (-0.356843) | 0.008687 / 0.007607 (0.001080) | 0.556029 / 0.226044 (0.329984) | 5.557192 / 2.268929 (3.288264) | 2.921596 / 55.444624 (-52.523028) | 2.520249 / 6.876477 (-4.356228) | 2.778965 / 2.142072 (0.636893) | 0.684765 / 4.805227 (-4.120462) | 0.159228 / 6.500664 (-6.341436) | 0.074015 / 0.075469 (-0.001454) |\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) | 1.534470 / 1.841788 (-0.307318) | 23.630693 / 8.074308 (15.556385) | 17.058142 / 10.191392 (6.866750) | 0.200909 / 0.680424 (-0.479515) | 0.021637 / 0.534201 (-0.512564) | 0.467417 / 0.579283 (-0.111866) | 0.460456 / 0.434364 (0.026092) | 0.541131 / 0.540337 (0.000793) | 0.728560 / 1.386936 (-0.658376) |\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) | 0.007625 / 0.011353 (-0.003727) | 0.004495 / 0.011008 (-0.006513) | 0.076373 / 0.038508 (0.037865) | 0.085260 / 0.023109 (0.062151) | 0.475778 / 0.275898 (0.199880) | 0.504604 / 0.323480 (0.181124) | 0.006733 / 0.007986 (-0.001253) | 0.003751 / 0.004328 (-0.000578) | 0.074993 / 0.004250 (0.070743) | 0.064704 / 0.037052 (0.027652) | 0.490072 / 0.258489 (0.231583) | 0.507560 / 0.293841 (0.213719) | 0.036765 / 0.128546 (-0.091781) | 0.009955 / 0.075646 (-0.065692) | 0.082452 / 0.419271 (-0.336820) | 0.057131 / 0.043533 (0.013598) | 0.467664 / 0.255139 (0.212525) | 0.482143 / 0.283200 (0.198943) | 0.025396 / 0.141683 (-0.116287) | 1.807587 / 1.452155 (0.355433) | 1.853355 / 1.492716 (0.360639) |\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) | 0.250543 / 0.018006 (0.232537) | 0.495685 / 0.000490 (0.495196) | 0.000415 / 0.000200 (0.000215) | 0.000063 / 0.000054 (0.000008) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.035795 / 0.037411 (-0.001616) | 0.105954 / 0.014526 (0.091428) | 0.120158 / 0.176557 (-0.056399) | 0.181714 / 0.737135 (-0.555422) | 0.121242 / 0.296338 (-0.175097) |\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) | 0.488241 / 0.215209 (0.273032) | 4.866916 / 2.077655 (2.789262) | 2.531530 / 1.504120 (1.027410) | 2.360642 / 1.541195 (0.819448) | 2.457320 / 1.468490 (0.988830) | 0.571224 / 4.584777 (-4.013553) | 4.339042 / 3.745712 (0.593330) | 3.672812 / 5.269862 (-1.597050) | 2.364535 / 4.565676 (-2.201142) | 0.067004 / 0.424275 (-0.357271) | 0.009019 / 0.007607 (0.001412) | 0.563751 / 0.226044 (0.337707) | 5.664917 / 2.268929 (3.395989) | 3.043316 / 55.444624 (-52.401308) | 2.682722 / 6.876477 (-4.193755) | 2.863482 / 2.142072 (0.721409) | 0.666171 / 4.805227 (-4.139056) | 0.151862 / 6.500664 (-6.348802) | 0.071199 / 0.075469 (-0.004271) |\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) | 1.601880 / 1.841788 (-0.239907) | 23.069073 / 8.074308 (14.994765) | 16.918377 / 10.191392 (6.726985) | 0.173614 / 0.680424 (-0.506810) | 0.021843 / 0.534201 (-0.512358) | 0.470531 / 0.579283 (-0.108753) | 0.471152 / 0.434364 (0.036788) | 0.550968 / 0.540337 (0.010631) | 0.718869 / 1.386936 (-0.668067) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#f9e6eea46fc9503765c125395e30e26c1ae2e084 \"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) | 0.007530 / 0.011353 (-0.003823) | 0.004151 / 0.011008 (-0.006858) | 0.098490 / 0.038508 (0.059982) | 0.086955 / 0.023109 (0.063846) | 0.362133 / 0.275898 (0.086235) | 0.391402 / 0.323480 (0.067922) | 0.006274 / 0.007986 (-0.001712) | 0.003711 / 0.004328 (-0.000618) | 0.073519 / 0.004250 (0.069269) | 0.066170 / 0.037052 (0.029118) | 0.379057 / 0.258489 (0.120568) | 0.398132 / 0.293841 (0.104291) | 0.033936 / 0.128546 (-0.094610) | 0.009977 / 0.075646 (-0.065670) | 0.323766 / 0.419271 (-0.095505) | 0.078615 / 0.043533 (0.035082) | 0.352403 / 0.255139 (0.097264) | 0.386607 / 0.283200 (0.103407) | 0.036579 / 0.141683 (-0.105103) | 1.691899 / 1.452155 (0.239745) | 1.819396 / 1.492716 (0.326680) |\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) | 0.216888 / 0.018006 (0.198882) | 0.465781 / 0.000490 (0.465291) | 0.006197 / 0.000200 (0.005997) | 0.000086 / 0.000054 (0.000031) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.032870 / 0.037411 (-0.004542) | 0.096026 / 0.014526 (0.081500) | 0.111093 / 0.176557 (-0.065464) | 0.185982 / 0.737135 (-0.551154) | 0.106967 / 0.296338 (-0.189371) |\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) | 0.441567 / 0.215209 (0.226358) | 4.353813 / 2.077655 (2.276158) | 2.176034 / 1.504120 (0.671914) | 1.969631 / 1.541195 (0.428437) | 2.048821 / 1.468490 (0.580330) | 0.549144 / 4.584777 (-4.035633) | 4.016166 / 3.745712 (0.270453) | 3.764249 / 5.269862 (-1.505613) | 2.293995 / 4.565676 (-2.271681) | 0.065227 / 0.424275 (-0.359048) | 0.008303 / 0.007607 (0.000695) | 0.513783 / 0.226044 (0.287738) | 5.247617 / 2.268929 (2.978689) | 2.782114 / 55.444624 (-52.662510) | 2.342776 / 6.876477 (-4.533701) | 2.621569 / 2.142072 (0.479497) | 0.679336 / 4.805227 (-4.125891) | 0.152061 / 6.500664 (-6.348603) | 0.070294 / 0.075469 (-0.005175) |\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) | 1.471778 / 1.841788 (-0.370010) | 22.714904 / 8.074308 (14.640596) | 15.607991 / 10.191392 (5.416599) | 0.172592 / 0.680424 (-0.507832) | 0.021799 / 0.534201 (-0.512402) | 0.462740 / 0.579283 (-0.116543) | 0.490885 / 0.434364 (0.056521) | 0.552997 / 0.540337 (0.012660) | 0.763784 / 1.386936 (-0.623152) |\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) | 0.007466 / 0.011353 (-0.003886) | 0.004322 / 0.011008 (-0.006686) | 0.074331 / 0.038508 (0.035823) | 0.085315 / 0.023109 (0.062206) | 0.409284 / 0.275898 (0.133386) | 0.464584 / 0.323480 (0.141104) | 0.005651 / 0.007986 (-0.002335) | 0.003577 / 0.004328 (-0.000751) | 0.070250 / 0.004250 (0.066000) | 0.059780 / 0.037052 (0.022727) | 0.419668 / 0.258489 (0.161179) | 0.462984 / 0.293841 (0.169143) | 0.034159 / 0.128546 (-0.094387) | 0.008999 / 0.075646 (-0.066647) | 0.076302 / 0.419271 (-0.342969) | 0.052274 / 0.043533 (0.008741) | 0.425938 / 0.255139 (0.170799) | 0.430399 / 0.283200 (0.147200) | 0.025017 / 0.141683 (-0.116666) | 1.680697 / 1.452155 (0.228542) | 1.774677 / 1.492716 (0.281960) |\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) | 0.291514 / 0.018006 (0.273508) | 0.461175 / 0.000490 (0.460685) | 0.023061 / 0.000200 (0.022861) | 0.000120 / 0.000054 (0.000065) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.033950 / 0.037411 (-0.003462) | 0.100032 / 0.014526 (0.085506) | 0.118308 / 0.176557 (-0.058249) | 0.183601 / 0.737135 (-0.553535) | 0.116936 / 0.296338 (-0.179402) |\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) | 0.478779 / 0.215209 (0.263570) | 4.709505 / 2.077655 (2.631850) | 2.457442 / 1.504120 (0.953322) | 2.213737 / 1.541195 (0.672542) | 2.340642 / 1.468490 (0.872152) | 0.567187 / 4.584777 (-4.017590) | 3.923061 / 3.745712 (0.177349) | 3.752989 / 5.269862 (-1.516873) | 2.324028 / 4.565676 (-2.241649) | 0.064471 / 0.424275 (-0.359804) | 0.008845 / 0.007607 (0.001238) | 0.547447 / 0.226044 (0.321402) | 5.599435 / 2.268929 (3.330506) | 2.980547 / 55.444624 (-52.464077) | 2.754908 / 6.876477 (-4.121569) | 2.832978 / 2.142072 (0.690906) | 0.635059 / 4.805227 (-4.170168) | 0.153478 / 6.500664 (-6.347187) | 0.067146 / 0.075469 (-0.008323) |\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) | 1.555588 / 1.841788 (-0.286200) | 22.828906 / 8.074308 (14.754597) | 16.211008 / 10.191392 (6.019616) | 0.168009 / 0.680424 (-0.512415) | 0.021966 / 0.534201 (-0.512235) | 0.464872 / 0.579283 (-0.114411) | 0.460429 / 0.434364 (0.026065) | 0.530498 / 0.540337 (-0.009839) | 0.705020 / 1.386936 (-0.681916) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#deb9e703237c8310c5a6db04f54d54368e951edd \"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) | 0.005964 / 0.011353 (-0.005389) | 0.003644 / 0.011008 (-0.007364) | 0.079607 / 0.038508 (0.041099) | 0.058387 / 0.023109 (0.035278) | 0.312226 / 0.275898 (0.036328) | 0.349206 / 0.323480 (0.025726) | 0.004715 / 0.007986 (-0.003271) | 0.002869 / 0.004328 (-0.001460) | 0.061668 / 0.004250 (0.057417) | 0.045694 / 0.037052 (0.008642) | 0.313516 / 0.258489 (0.055027) | 0.357543 / 0.293841 (0.063702) | 0.027179 / 0.128546 (-0.101367) | 0.007961 / 0.075646 (-0.067686) | 0.262473 / 0.419271 (-0.156798) | 0.045588 / 0.043533 (0.002055) | 0.313102 / 0.255139 (0.057963) | 0.368686 / 0.283200 (0.085486) | 0.020556 / 0.141683 (-0.121127) | 1.447258 / 1.452155 (-0.004897) | 1.527319 / 1.492716 (0.034602) |\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) | 0.199417 / 0.018006 (0.181411) | 0.422155 / 0.000490 (0.421665) | 0.004972 / 0.000200 (0.004772) | 0.000073 / 0.000054 (0.000018) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.023539 / 0.037411 (-0.013872) | 0.073055 / 0.014526 (0.058529) | 0.083631 / 0.176557 (-0.092926) | 0.145923 / 0.737135 (-0.591212) | 0.083820 / 0.296338 (-0.212518) |\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) | 0.396305 / 0.215209 (0.181096) | 3.967065 / 2.077655 (1.889410) | 2.101109 / 1.504120 (0.596989) | 1.958817 / 1.541195 (0.417622) | 2.037894 / 1.468490 (0.569404) | 0.496955 / 4.584777 (-4.087822) | 3.078948 / 3.745712 (-0.666764) | 3.363655 / 5.269862 (-1.906207) | 2.087659 / 4.565676 (-2.478018) | 0.057171 / 0.424275 (-0.367104) | 0.006410 / 0.007607 (-0.001197) | 0.470535 / 0.226044 (0.244491) | 4.715259 / 2.268929 (2.446330) | 2.355510 / 55.444624 (-53.089114) | 2.025270 / 6.876477 (-4.851207) | 2.210401 / 2.142072 (0.068329) | 0.580538 / 4.805227 (-4.224689) | 0.125068 / 6.500664 (-6.375596) | 0.059871 / 0.075469 (-0.015598) |\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) | 1.245468 / 1.841788 (-0.596320) | 18.322042 / 8.074308 (10.247734) | 13.609726 / 10.191392 (3.418334) | 0.143623 / 0.680424 (-0.536801) | 0.017068 / 0.534201 (-0.517133) | 0.330758 / 0.579283 (-0.248525) | 0.339946 / 0.434364 (-0.094418) | 0.377861 / 0.540337 (-0.162476) | 0.524593 / 1.386936 (-0.862343) |\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) | 0.006049 / 0.011353 (-0.005304) | 0.003737 / 0.011008 (-0.007271) | 0.062816 / 0.038508 (0.024308) | 0.063768 / 0.023109 (0.040658) | 0.362001 / 0.275898 (0.086103) | 0.395251 / 0.323480 (0.071772) | 0.004823 / 0.007986 (-0.003163) | 0.002881 / 0.004328 (-0.001447) | 0.061987 / 0.004250 (0.057737) | 0.049950 / 0.037052 (0.012898) | 0.362442 / 0.258489 (0.103953) | 0.399321 / 0.293841 (0.105480) | 0.027616 / 0.128546 (-0.100930) | 0.007965 / 0.075646 (-0.067681) | 0.068584 / 0.419271 (-0.350687) | 0.044700 / 0.043533 (0.001168) | 0.361011 / 0.255139 (0.105872) | 0.386007 / 0.283200 (0.102807) | 0.024621 / 0.141683 (-0.117061) | 1.441497 / 1.452155 (-0.010657) | 1.533145 / 1.492716 (0.040429) |\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) | 0.223446 / 0.018006 (0.205440) | 0.411147 / 0.000490 (0.410657) | 0.001821 / 0.000200 (0.001621) | 0.000081 / 0.000054 (0.000027) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.025661 / 0.037411 (-0.011751) | 0.077838 / 0.014526 (0.063313) | 0.086148 / 0.176557 (-0.090408) | 0.140386 / 0.737135 (-0.596750) | 0.088793 / 0.296338 (-0.207546) |\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) | 0.425209 / 0.215209 (0.210000) | 4.250723 / 2.077655 (2.173068) | 2.403437 / 1.504120 (0.899317) | 2.283584 / 1.541195 (0.742390) | 2.326870 / 1.468490 (0.858380) | 0.504781 / 4.584777 (-4.079996) | 3.017042 / 3.745712 (-0.728670) | 4.643068 / 5.269862 (-0.626794) | 2.535710 / 4.565676 (-2.029967) | 0.058520 / 0.424275 (-0.365755) | 0.006766 / 0.007607 (-0.000841) | 0.500664 / 0.226044 (0.274620) | 5.017073 / 2.268929 (2.748145) | 2.668661 / 55.444624 (-52.775963) | 2.335486 / 6.876477 (-4.540991) | 2.486518 / 2.142072 (0.344445) | 0.598795 / 4.805227 (-4.206432) | 0.126395 / 6.500664 (-6.374269) | 0.063154 / 0.075469 (-0.012315) |\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) | 1.358059 / 1.841788 (-0.483728) | 18.615724 / 8.074308 (10.541416) | 13.670934 / 10.191392 (3.479542) | 0.134650 / 0.680424 (-0.545774) | 0.016941 / 0.534201 (-0.517260) | 0.335215 / 0.579283 (-0.244068) | 0.356118 / 0.434364 (-0.078246) | 0.393109 / 0.540337 (-0.147228) | 0.534165 / 1.386936 (-0.852771) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#da7d3b557665f34e84cd151ffe9d80b45a19fe33 \"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) | 0.007343 / 0.011353 (-0.004010) | 0.005145 / 0.011008 (-0.005863) | 0.099820 / 0.038508 (0.061312) | 0.033487 / 0.023109 (0.010378) | 0.313069 / 0.275898 (0.037171) | 0.335420 / 0.323480 (0.011940) | 0.005959 / 0.007986 (-0.002027) | 0.005373 / 0.004328 (0.001044) | 0.076568 / 0.004250 (0.072317) | 0.048702 / 0.037052 (0.011650) | 0.322957 / 0.258489 (0.064468) | 0.363044 / 0.293841 (0.069203) | 0.035070 / 0.128546 (-0.093476) | 0.012029 / 0.075646 (-0.063618) | 0.334664 / 0.419271 (-0.084607) | 0.050549 / 0.043533 (0.007017) | 0.310113 / 0.255139 (0.054974) | 0.324405 / 0.283200 (0.041205) | 0.097596 / 0.141683 (-0.044087) | 1.440741 / 1.452155 (-0.011414) | 1.531194 / 1.492716 (0.038478) |\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) | 0.220799 / 0.018006 (0.202793) | 0.438158 / 0.000490 (0.437668) | 0.007737 / 0.000200 (0.007537) | 0.000082 / 0.000054 (0.000027) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.026888 / 0.037411 (-0.010523) | 0.106281 / 0.014526 (0.091755) | 0.117419 / 0.176557 (-0.059138) | 0.179144 / 0.737135 (-0.557992) | 0.122477 / 0.296338 (-0.173861) |\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) | 0.412667 / 0.215209 (0.197458) | 4.108784 / 2.077655 (2.031129) | 1.834300 / 1.504120 (0.330180) | 1.627256 / 1.541195 (0.086061) | 1.691036 / 1.468490 (0.222546) | 0.713405 / 4.584777 (-3.871372) | 3.839262 / 3.745712 (0.093550) | 2.108453 / 5.269862 (-3.161408) | 1.340740 / 4.565676 (-3.224936) | 0.087776 / 0.424275 (-0.336499) | 0.012730 / 0.007607 (0.005123) | 0.505323 / 0.226044 (0.279279) | 5.085176 / 2.268929 (2.816247) | 2.307165 / 55.444624 (-53.137459) | 1.936771 / 6.876477 (-4.939706) | 2.097391 / 2.142072 (-0.044681) | 0.856215 / 4.805227 (-3.949012) | 0.171826 / 6.500664 (-6.328838) | 0.066603 / 0.075469 (-0.008866) |\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) | 1.202126 / 1.841788 (-0.639661) | 15.173598 / 8.074308 (7.099290) | 15.012645 / 10.191392 (4.821253) | 0.162187 / 0.680424 (-0.518237) | 0.017462 / 0.534201 (-0.516739) | 0.423895 / 0.579283 (-0.155388) | 0.432010 / 0.434364 (-0.002354) | 0.503234 / 0.540337 (-0.037104) | 0.598948 / 1.386936 (-0.787988) |\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) | 0.007099 / 0.011353 (-0.004254) | 0.005167 / 0.011008 (-0.005841) | 0.075551 / 0.038508 (0.037043) | 0.033050 / 0.023109 (0.009940) | 0.339629 / 0.275898 (0.063731) | 0.380486 / 0.323480 (0.057006) | 0.005776 / 0.007986 (-0.002209) | 0.004029 / 0.004328 (-0.000299) | 0.075074 / 0.004250 (0.070823) | 0.046709 / 0.037052 (0.009656) | 0.340203 / 0.258489 (0.081714) | 0.380849 / 0.293841 (0.087008) | 0.035027 / 0.128546 (-0.093519) | 0.012226 / 0.075646 (-0.063420) | 0.087525 / 0.419271 (-0.331747) | 0.049361 / 0.043533 (0.005828) | 0.341854 / 0.255139 (0.086715) | 0.359590 / 0.283200 (0.076390) | 0.100102 / 0.141683 (-0.041581) | 1.482759 / 1.452155 (0.030605) | 1.569905 / 1.492716 (0.077189) |\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) | 0.213615 / 0.018006 (0.195609) | 0.441117 / 0.000490 (0.440628) | 0.004932 / 0.000200 (0.004732) | 0.000093 / 0.000054 (0.000038) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.031313 / 0.037411 (-0.006098) | 0.110191 / 0.014526 (0.095665) | 0.125320 / 0.176557 (-0.051237) | 0.177658 / 0.737135 (-0.559477) | 0.127928 / 0.296338 (-0.168410) |\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) | 0.426952 / 0.215209 (0.211743) | 4.247731 / 2.077655 (2.170076) | 2.107318 / 1.504120 (0.603198) | 1.843845 / 1.541195 (0.302650) | 1.894822 / 1.468490 (0.426332) | 0.696232 / 4.584777 (-3.888545) | 3.826516 / 3.745712 (0.080804) | 2.126688 / 5.269862 (-3.143174) | 1.327062 / 4.565676 (-3.238615) | 0.085693 / 0.424275 (-0.338582) | 0.012226 / 0.007607 (0.004619) | 0.521904 / 0.226044 (0.295859) | 5.219798 / 2.268929 (2.950869) | 2.524908 / 55.444624 (-52.919716) | 2.212078 / 6.876477 (-4.664399) | 2.373944 / 2.142072 (0.231871) | 0.833846 / 4.805227 (-3.971381) | 0.169639 / 6.500664 (-6.331025) | 0.064538 / 0.075469 (-0.010931) |\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) | 1.254930 / 1.841788 (-0.586858) | 15.585277 / 8.074308 (7.510969) | 14.762857 / 10.191392 (4.571465) | 0.146959 / 0.680424 (-0.533465) | 0.017451 / 0.534201 (-0.516750) | 0.424469 / 0.579283 (-0.154814) | 0.422359 / 0.434364 (-0.012004) | 0.489930 / 0.540337 (-0.050408) | 0.595856 / 1.386936 (-0.791080) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#213c72f52ae52b662f967d3218f66c70a3043048 \"CML watermark\")\n", "@albertvillanova thanks for the review. As you prefer for the github CI config. I just took it from @lhoestq's branch when testing hfh==0.14.0. I think it's still relevant for next releases. In any case, I let you handle merging the PR :)", "<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) | 0.008371 / 0.011353 (-0.002982) | 0.005210 / 0.011008 (-0.005798) | 0.105639 / 0.038508 (0.067131) | 0.045903 / 0.023109 (0.022794) | 0.391231 / 0.275898 (0.115333) | 0.438824 / 0.323480 (0.115345) | 0.006270 / 0.007986 (-0.001715) | 0.005950 / 0.004328 (0.001621) | 0.079685 / 0.004250 (0.075434) | 0.052121 / 0.037052 (0.015069) | 0.387787 / 0.258489 (0.129298) | 0.434322 / 0.293841 (0.140481) | 0.032598 / 0.128546 (-0.095948) | 0.012126 / 0.075646 (-0.063520) | 0.359658 / 0.419271 (-0.059613) | 0.046686 / 0.043533 (0.003154) | 0.391973 / 0.255139 (0.136834) | 0.421149 / 0.283200 (0.137949) | 0.105920 / 0.141683 (-0.035763) | 1.483008 / 1.452155 (0.030854) | 1.617010 / 1.492716 (0.124294) |\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) | 0.199111 / 0.018006 (0.181105) | 0.407995 / 0.000490 (0.407505) | 0.006706 / 0.000200 (0.006506) | 0.000229 / 0.000054 (0.000175) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.030247 / 0.037411 (-0.007164) | 0.115977 / 0.014526 (0.101451) | 0.118112 / 0.176557 (-0.058444) | 0.182710 / 0.737135 (-0.554426) | 0.122483 / 0.296338 (-0.173855) |\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) | 0.430455 / 0.215209 (0.215246) | 4.314298 / 2.077655 (2.236643) | 1.898124 / 1.504120 (0.394005) | 1.734909 / 1.541195 (0.193715) | 1.802400 / 1.468490 (0.333910) | 0.717237 / 4.584777 (-3.867539) | 4.004705 / 3.745712 (0.258993) | 2.138901 / 5.269862 (-3.130960) | 1.254037 / 4.565676 (-3.311640) | 0.085594 / 0.424275 (-0.338681) | 0.013774 / 0.007607 (0.006166) | 0.535218 / 0.226044 (0.309174) | 5.373730 / 2.268929 (3.104801) | 2.371194 / 55.444624 (-53.073430) | 2.111206 / 6.876477 (-4.765270) | 2.225137 / 2.142072 (0.083064) | 0.838325 / 4.805227 (-3.966902) | 0.159176 / 6.500664 (-6.341488) | 0.072285 / 0.075469 (-0.003184) |\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) | 1.352232 / 1.841788 (-0.489555) | 16.926722 / 8.074308 (8.852414) | 16.709531 / 10.191392 (6.518139) | 0.159249 / 0.680424 (-0.521175) | 0.017667 / 0.534201 (-0.516534) | 0.426894 / 0.579283 (-0.152390) | 0.539903 / 0.434364 (0.105539) | 0.537471 / 0.540337 (-0.002866) | 0.619592 / 1.386936 (-0.767344) |\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) | 0.008354 / 0.011353 (-0.002999) | 0.005366 / 0.011008 (-0.005642) | 0.080961 / 0.038508 (0.042453) | 0.046574 / 0.023109 (0.023465) | 0.345949 / 0.275898 (0.070051) | 0.394041 / 0.323480 (0.070562) | 0.006209 / 0.007986 (-0.001777) | 0.005980 / 0.004328 (0.001651) | 0.076235 / 0.004250 (0.071984) | 0.051833 / 0.037052 (0.014780) | 0.348786 / 0.258489 (0.090297) | 0.397421 / 0.293841 (0.103580) | 0.033026 / 0.128546 (-0.095520) | 0.012217 / 0.075646 (-0.063429) | 0.087439 / 0.419271 (-0.331832) | 0.045488 / 0.043533 (0.001955) | 0.352160 / 0.255139 (0.097021) | 0.379079 / 0.283200 (0.095879) | 0.116111 / 0.141683 (-0.025572) | 1.470177 / 1.452155 (0.018022) | 1.587499 / 1.492716 (0.094783) |\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) | 0.296149 / 0.018006 (0.278143) | 0.592362 / 0.000490 (0.591872) | 0.000492 / 0.000200 (0.000292) | 0.000064 / 0.000054 (0.000010) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.036599 / 0.037411 (-0.000813) | 0.113768 / 0.014526 (0.099242) | 0.116198 / 0.176557 (-0.060358) | 0.180329 / 0.737135 (-0.556806) | 0.123942 / 0.296338 (-0.172396) |\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) | 0.452445 / 0.215209 (0.237236) | 4.504330 / 2.077655 (2.426675) | 2.275645 / 1.504120 (0.771525) | 2.107765 / 1.541195 (0.566571) | 2.086363 / 1.468490 (0.617873) | 0.723721 / 4.584777 (-3.861056) | 3.825330 / 3.745712 (0.079618) | 2.162743 / 5.269862 (-3.107119) | 1.255953 / 4.565676 (-3.309724) | 0.085860 / 0.424275 (-0.338415) | 0.013790 / 0.007607 (0.006183) | 0.560257 / 0.226044 (0.334213) | 5.618180 / 2.268929 (3.349251) | 2.625423 / 55.444624 (-52.819202) | 2.374381 / 6.876477 (-4.502095) | 2.496560 / 2.142072 (0.354488) | 0.841120 / 4.805227 (-3.964107) | 0.161541 / 6.500664 (-6.339123) | 0.075270 / 0.075469 (-0.000199) |\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) | 1.432916 / 1.841788 (-0.408872) | 14.858534 / 8.074308 (6.784226) | 14.973521 / 10.191392 (4.782129) | 0.148312 / 0.680424 (-0.532112) | 0.016811 / 0.534201 (-0.517390) | 0.382623 / 0.579283 (-0.196660) | 0.389767 / 0.434364 (-0.044596) | 0.449657 / 0.540337 (-0.090680) | 0.533723 / 1.386936 (-0.853214) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#f8344350f15265a585188ac986ae49a8ed8289fe \"CML watermark\")\n", "I agree it is good to have a way to run the CI on push, without needing to open a PR.\r\n\r\nBut I think the branch name should be more generic (and this is not specific to this PR). See:\r\n- #5790 ", "<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) | 0.007208 / 0.011353 (-0.004145) | 0.005600 / 0.011008 (-0.005408) | 0.096129 / 0.038508 (0.057621) | 0.027834 / 0.023109 (0.004725) | 0.295106 / 0.275898 (0.019208) | 0.323983 / 0.323480 (0.000503) | 0.005164 / 0.007986 (-0.002822) | 0.003962 / 0.004328 (-0.000366) | 0.078339 / 0.004250 (0.074089) | 0.036974 / 0.037052 (-0.000078) | 0.310315 / 0.258489 (0.051826) | 0.338036 / 0.293841 (0.044195) | 0.042124 / 0.128546 (-0.086422) | 0.015886 / 0.075646 (-0.059760) | 0.337961 / 0.419271 (-0.081310) | 0.051507 / 0.043533 (0.007974) | 0.297505 / 0.255139 (0.042366) | 0.310728 / 0.283200 (0.027528) | 0.086312 / 0.141683 (-0.055371) | 1.356923 / 1.452155 (-0.095232) | 1.429366 / 1.492716 (-0.063350) |\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) | 0.205495 / 0.018006 (0.187489) | 0.460639 / 0.000490 (0.460149) | 0.003996 / 0.000200 (0.003796) | 0.000093 / 0.000054 (0.000038) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.021970 / 0.037411 (-0.015442) | 0.090283 / 0.014526 (0.075757) | 0.098579 / 0.176557 (-0.077978) | 0.160437 / 0.737135 (-0.576699) | 0.102738 / 0.296338 (-0.193600) |\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) | 0.494474 / 0.215209 (0.279265) | 4.967453 / 2.077655 (2.889799) | 2.045852 / 1.504120 (0.541732) | 1.858022 / 1.541195 (0.316827) | 1.771874 / 1.468490 (0.303384) | 1.186368 / 4.584777 (-3.398408) | 4.974762 / 3.745712 (1.229050) | 2.616225 / 5.269862 (-2.653636) | 1.702971 / 4.565676 (-2.862705) | 0.124929 / 0.424275 (-0.299346) | 0.011774 / 0.007607 (0.004167) | 0.569643 / 0.226044 (0.343598) | 5.793114 / 2.268929 (3.524186) | 2.441561 / 55.444624 (-53.003064) | 1.862233 / 6.876477 (-5.014243) | 1.931142 / 2.142072 (-0.210931) | 1.148915 / 4.805227 (-3.656313) | 0.203914 / 6.500664 (-6.296750) | 0.062468 / 0.075469 (-0.013001) |\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) | 1.188708 / 1.841788 (-0.653080) | 13.710830 / 8.074308 (5.636522) | 15.695153 / 10.191392 (5.503761) | 0.171467 / 0.680424 (-0.508957) | 0.024509 / 0.534201 (-0.509692) | 0.450270 / 0.579283 (-0.129014) | 0.500712 / 0.434364 (0.066348) | 0.488632 / 0.540337 (-0.051706) | 0.574893 / 1.386936 (-0.812043) |\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) | 0.007254 / 0.011353 (-0.004099) | 0.006199 / 0.011008 (-0.004809) | 0.072079 / 0.038508 (0.033571) | 0.026909 / 0.023109 (0.003800) | 0.355538 / 0.275898 (0.079640) | 0.358625 / 0.323480 (0.035145) | 0.005564 / 0.007986 (-0.002421) | 0.005278 / 0.004328 (0.000950) | 0.076469 / 0.004250 (0.072219) | 0.038269 / 0.037052 (0.001216) | 0.355214 / 0.258489 (0.096725) | 0.383219 / 0.293841 (0.089378) | 0.046516 / 0.128546 (-0.082030) | 0.015393 / 0.075646 (-0.060254) | 0.088506 / 0.419271 (-0.330765) | 0.050326 / 0.043533 (0.006793) | 0.327265 / 0.255139 (0.072126) | 0.370176 / 0.283200 (0.086976) | 0.102438 / 0.141683 (-0.039245) | 1.378969 / 1.452155 (-0.073186) | 1.441998 / 1.492716 (-0.050719) |\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) | 0.209044 / 0.018006 (0.191038) | 0.455733 / 0.000490 (0.455243) | 0.005856 / 0.000200 (0.005656) | 0.000116 / 0.000054 (0.000061) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.025336 / 0.037411 (-0.012075) | 0.097449 / 0.014526 (0.082923) | 0.106301 / 0.176557 (-0.070255) | 0.153053 / 0.737135 (-0.584082) | 0.107938 / 0.296338 (-0.188401) |\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) | 0.491070 / 0.215209 (0.275861) | 5.049637 / 2.077655 (2.971982) | 2.064709 / 1.504120 (0.560589) | 1.782266 / 1.541195 (0.241072) | 1.798570 / 1.468490 (0.330080) | 0.988886 / 4.584777 (-3.595891) | 4.690324 / 3.745712 (0.944612) | 4.317355 / 5.269862 (-0.952507) | 2.347596 / 4.565676 (-2.218081) | 0.117249 / 0.424275 (-0.307026) | 0.011614 / 0.007607 (0.004007) | 0.630033 / 0.226044 (0.403988) | 6.140108 / 2.268929 (3.871180) | 2.638080 / 55.444624 (-52.806545) | 2.133017 / 6.876477 (-4.743459) | 2.123392 / 2.142072 (-0.018680) | 1.178056 / 4.805227 (-3.627171) | 0.209465 / 6.500664 (-6.291199) | 0.063234 / 0.075469 (-0.012235) |\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) | 1.238089 / 1.841788 (-0.603699) | 14.066866 / 8.074308 (5.992558) | 16.225480 / 10.191392 (6.034088) | 0.206466 / 0.680424 (-0.473958) | 0.027279 / 0.534201 (-0.506922) | 0.443006 / 0.579283 (-0.136277) | 0.509512 / 0.434364 (0.075148) | 0.479075 / 0.540337 (-0.061263) | 0.573546 / 1.386936 (-0.813390) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#c6015a070c66a5bbd84603d415ccc57cb668b44b \"CML watermark\")\n" ]

[]

[]

[]

[ "Possible duplicate\r\n\r\n#1776 https://github.com/huggingface/datasets/issues/\r\n\r\nreally looking PR for this feature", "Hi @acul3 \r\n\r\nIssue #1776 talks about doing on-the-fly data pre-processing, which I think is solved in the next release as mentioned in the issue #1825. I also look forward to using this feature, though :)\r\n\r\nI wanted to ask about on-the-fly data loading from the cache (before pre-processing).", "Hi ! Currently when you load a dataset via `load_dataset` for example, then the dataset is memory-mapped from an Arrow file on disk. Therefore there's almost no RAM usage even if your dataset contains TB of data.\r\nUsually at training time only one batch of data at a time is loaded in memory.\r\n\r\nDoes that answer your question or were you thinking about something else ?", "Hi @lhoestq,\r\n\r\nI apologize for the late response. This answers my question. Thanks a lot." ]

[]

[ "The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/datasets/pr_5126). All of your documentation changes will be reflected on that endpoint.", "I have removed the reference to the Issue in the PR title, so that we avoid to have both references (to the issue and to the PR) in the merge commit to the main branch.\r\n\r\nInstead, it should be commented in the PR description, so that the PR is appropriately linked by GitHub to its corresponding Issue:\r\n\r\n> Fix #5098.", "@albertvillanova What should I test in your opinion? Also, where should I save the test file and how should I name it? Thanks for your support", "The regression test to be implemented should test what your PR fixes: that is, that `_resolve_single_pattern_locally` function does not resolve any symbolic link when passed a directory that does contain any.\r\n\r\nAs you are testing a function in `data_files.py`, the corresponding test should be in `tests/test_data_files.py`.\r\n\r\nYou could name the test something lilke: `test_resolve_single_pattern_locally_does_not_resolve_symbolic_links`\r\n\r\nYou could take inspiration from other tests there in that file." ]

[]

[ "_The documentation is not available anymore as the PR was closed or merged._", "thank you @lhoestq!! :hugs: " ]

[ { "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 and thanks @mariosasko for fixing ! We just did a patch release `2.10.1` with the fix" ]

[]

[ "Hi! Let me know if any more comments! Will fix it! :-)" ]

[]

[ "We should try to regenerate the data using the official script.\r\nBut iirc that's what we used in the first place, so not sure why it didn't match in the first place.\r\n\r\nI'll let you know when the dataset is updated", "Thanks, looking forward to hearing your update on this thread. \r\n\r\nThis is a blocking issue for us; would appreciate any progress on this front. We can also help with the fix, if you deem it appropriately. ", "I just started the generation on my side, I'll let you know how it goes :) ", "Hmm after a first run I'm still missing 136668/226711 urls.\r\nI'll relaunch it tomorrow to try to get the remaining ones.", "Update: I'm missing 36/226711 urls but I haven't managed to download them yet", "Thanks! That sounds like a reasonable number! ", "So I managed to download them all but when parsing only 226,181/226,711 worked.\r\nNot sure if it's worth digging and debugging parsing at this point :/ ", "Maybe @sshleifer can help, I think he's already played with xsum at one point", "Thanks @lhoestq\r\nIt would be great to improve coverage, but IDs are the really crucial part for us. We'd really appreciate an update to the dataset with IDs either way!", "I gave up at an even earlier point. The dataset I use has 204,017 train examples.", "@lhoestq @sshleifer like @jbragg said earlier, the main issue for us is that the current XSUM dataset (in your package) does not have IDs suggested by the original dataset ([here is the file](https://raw.githubusercontent.com/EdinburghNLP/XSum/master/XSum-Dataset/XSum-TRAINING-DEV-TEST-SPLIT-90-5-5.json).) Would appreciate if you update the XSUM dataset to include the instance IDs. \r\n\r\nThe missing instances is also a problem, but likely not worth pursuing given its relatively small scale. ", ">So I managed to download them all but when parsing only 226,181/226,711 worked.\r\n\r\n@lhoestq any chance we could update the HF-hosted dataset with the IDs in your new version? Happy to help if there's something I can do.", "Well I couldn't parse what I downloaded.\r\nUnfortunately I think I won't be able to take a look at it this week.\r\nI can try to send you what I got if you want to give it a shot @jbragg \r\nOtherwise feel free to re-run the xsum download script, maybe you'll be luckier than me", "Resolved via #754" ]

[ { "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" } ]

[ "uploaded dataset [here](https://huggingface.co/datasets/embedding-data/WikiAnswers)" ]

[]

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

[]

[ "Yes I agree. Maybe something that lets specify different format depending on the column ? Especially to better control dtype and shape (and ragged for tf)\r\n\r\nOh and I forgot: this one should also fix the second issue found in #477 for the next release", "I am running into the same issue with the error message on my local windows machine -\r\nAttributeError: 'tensorflow.python.framework.ops.EagerTensor' object has no attribute 'to_tensor'. Tensorflow version is 2.6. Anything that I can do to fix it?\r\ntrain_features = {x: tf_train_dataset[x].to_tensor() for x in tokenizer.model_input_names}\r\ntrain_tf_dataset = tf.data.Dataset.from_tensor_slices((train_features, tf_train_dataset[\"label\"]))\r\ntrain_tf_dataset = train_tf_dataset.shuffle(len(tf_train_dataset)).batch(8)\r\n\r\neval_features = {x: tf_eval_dataset[x].to_tensor() for x in tokenizer.model_input_names}\r\neval_tf_dataset = tf.data.Dataset.from_tensor_slices((eval_features, tf_eval_dataset[\"label\"]))\r\neval_tf_dataset = eval_tf_dataset.batch(8)\r\n\r\nttributeError Traceback (most recent call last)\r\n<ipython-input-59-f50e45c2c0dc> in <module>\r\n----> 1 train_features = {x: tf_train_dataset[x].convert_to_tensor() for x in tokenizer.model_input_names}\r\n 2 train_tf_dataset = tf.data.Dataset.from_tensor_slices((train_features, tf_train_dataset[\"label\"]))\r\n 3 train_tf_dataset = train_tf_dataset.shuffle(len(tf_train_dataset)).batch(8)\r\n 4 \r\n 5 eval_features = {x: tf_eval_dataset[x].to_tensor() for x in tokenizer.model_input_names}\r\n\r\n<ipython-input-59-f50e45c2c0dc> in <dictcomp>(.0)\r\n----> 1 train_features = {x: tf_train_dataset[x].convert_to_tensor() for x in tokenizer.model_input_names}\r\n 2 train_tf_dataset = tf.data.Dataset.from_tensor_slices((train_features, tf_train_dataset[\"label\"]))\r\n 3 train_tf_dataset = train_tf_dataset.shuffle(len(tf_train_dataset)).batch(8)\r\n 4 \r\n 5 eval_features = {x: tf_eval_dataset[x].to_tensor() for x in tokenizer.model_input_names}\r\n\r\n~\\AppData\\Roaming\\Python\\Python38\\site-packages\\tensorflow\\python\\framework\\ops.py in __getattr__(self, name)\r\n 399 from tensorflow.python.ops.numpy_ops import np_config\r\n 400 np_config.enable_numpy_behavior()\"\"\".format(type(self).__name__, name))\r\n--> 401 self.__getattribute__(name)\r\n 402 \r\n 403 @staticmethod\r\n\r\nAttributeError: 'tensorflow.python.framework.ops.EagerTensor' object has no attribute 'convert_to_tensor'\r\n\r\n", "Hi ! Before calling `to_tensor`, make sure that your object is a RaggedTensor, because it may already be a regular Tensor if the shapes of your examples are all the same", "Okay. i am not familiar with how to check the difference between the two. I will research on this." ]

[]

[]

[ { "color": "d73a4a", "default": true, "description": "Something isn't working", "id": 1935892857, "name": "bug", "node_id": "MDU6TGFiZWwxOTM1ODkyODU3", "url": "https://api.github.com/repos/huggingface/datasets/labels/bug" }, { "color": "7057ff", "default": true, "description": "Good for newcomers", "id": 1935892877, "name": "good first issue", "node_id": "MDU6TGFiZWwxOTM1ODkyODc3", "url": "https://api.github.com/repos/huggingface/datasets/labels/good%20first%20issue" } ]

[ "Are you able to reproduce this? My example is small enough that it should be easy to try.", "Hi! Thanks for reporting and providing a reproducible example. Indeed, by default, `datasets` performs an expensive cast on the values returned by `map` to convert them to one of the types supported by PyArrow (the underlying storage format used by `datasets`). This cast is not needed on NumPy arrays as PyArrow supports them natively, so one way to make this transform faster is to add `return_tensors=\"np\"` to the tokenizer call. \r\n\r\nI think we should mention this in the docs (cc @stevhliu)", "I tested this tokenize function and indeed noticed a casting. However it seems to only concerns the `offset_mapping` field, which contains a list of tuples, that is converted to a list of lists. Since `pyarrow` also supports tuples, we actually don't need to convert the tuples to lists. \r\n\r\nI think this can be changed here: \r\n\r\nhttps://github.com/huggingface/datasets/blob/ede72d3f9796339701ec59899c7c31d2427046fb/src/datasets/features/features.py#L382-L383\r\n\r\n```diff\r\n- if isinstance(obj, list): \r\n+ if isinstance(obj, (list, tuple)): \r\n```\r\n\r\nand here: \r\n\r\nhttps://github.com/huggingface/datasets/blob/ede72d3f9796339701ec59899c7c31d2427046fb/src/datasets/features/features.py#L386-L387\r\n\r\n```diff\r\n- return obj if isinstance(obj, list) else [], isinstance(obj, tuple)\r\n+ return obj, False\r\n```\r\n\r\n@srobertjames can you try applying these changes and let us know if it helps ? If so, feel free to open a Pull Request to contribute this improvement if you want :)", "Wow, adding `return_tensors=\"np\"` sped up my example by a **factor 17x** of and completely eliminated the casting! I'd recommend not only to document it, but to make that the default.\r\n\r\nThe code at https://github.com/huggingface/notebooks/blob/main/examples/question_answering.ipynb does not specify `return_tensors=\"np\"` but yet avoids the casting penalty. How does it do that? (The ntbk seems to do `return_overflowing_tokens=True, return_offsets_mapping=True,`).\r\n\r\nAlso, surprisingly enough, using `return_tensors=\"pt\"` (which is my eventual application) yields this error:\r\n```\r\nTypeError: Provided `function` which is applied to all elements of table returns a `dict` of types \r\n[<class 'torch.Tensor'>, <class 'torch.Tensor'>, <class 'torch.Tensor'>, <class 'torch.Tensor'>]. \r\nWhen using `batched=True`, make sure provided `function` returns a `dict` of types like \r\n`(<class 'list'>, <class 'numpy.ndarray'>)`.\r\n```", "Setting the output to `\"np\"` makes the whole pipeline fast because it moves the data buffers from rust to python to arrow using zero-copy, and also because it does eliminate the casting completely ;)\r\n\r\nHave you had a chance to try eliminating the tuple casting using the trick above ?", "@lhoestq I just benchmarked the two edits to `features.py` above, and they appear to solve the problem, bringing my original example to within 20% the speed of the output `\"np\"` example. Nice!\r\n\r\nFor a pull request, do you suggest simply following https://github.com/huggingface/datasets/blob/main/CONTRIBUTING.md ?", "Cool ! Sure feel free to follow these instructions to open a PR :) thanks !", "#take", "Resolved via https://github.com/huggingface/datasets/pull/4993." ]

[]

[ "I'm also fine with the removal in 1.15" ]

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

[ "I just take a look at all the outputs produced by `datasets` using the different log-levels.\r\nAs far as i can tell using `datasets==1.17.0` they overall issue seems to be fixed.\r\n\r\nHowever, I noticed that there is one tqdm based progress indicator appearing on STDERR that I can simply not suppress.\r\n```\r\nResolving data files: 100%|██████████| 652/652 [00:00<00:00, 1604.52it/s]\r\n```\r\n\r\nAccording to _get_origin_metadata_locally_or_by_urls it shold be supressable by using the `NOTSET` log-level\r\nhttps://github.com/huggingface/datasets/blob/1406a04c3e911cec2680d8bc513653e0cafcaaa4/src/datasets/data_files.py#L491-L501\r\nSadly when specifiing the log-level `NOTSET` it seems to has no effect.\r\n\r\nBut appart from it not having any effect I must admit that it seems unintuitive to me.\r\nI would suggest changing this such that it is only shown when the log-level is greater or equal to INFO.\r\n\r\nThis would conform better to INFO according to the [documentation](https://huggingface.co/docs/datasets/v1.0.0/package_reference/logging_methods.html#datasets.logging.set_verbosity_info).\r\n> This will display most of the logging information and tqdm bars.\r\n\r\nAny inputs on this?\r\nI will be happy to supply a PR if desired 👍 ", "Hi! This should disable the tqdm output:\r\n```python\r\nimport datasets\r\ndatasets.set_progress_bar_enabled(False)\r\n```\r\n\r\nOn a side note: I believe the issue with logging (not tqdm) is still relevant on master." ]

[]

[ "the CI error was just a connection error due to all the activity on the repo this week ^^'\r\nI re-ran it so it should be good now", "I hadn't realized the problem with the dummies since it had passed without errors.\r\nSuggestion: maybe we can show the user a warning based on the generated dummy size.", "Thanks for changing to configs ! Looks all good now :) \r\n\r\nBefore we merge, can you re-lighten the dummy data please if you don't mind ? The idea is to have them weigh only a few KB (currently it's 50KB each). Feel free to remove any unnecessary files or chunk of text", "(also you can ignore the `RemoteDatasetTest ` CI errors, they're fixed on master )", "merging since the CI is fixed on master" ]

[]

[ "You're right, thanks for pointing it out. We will update it " ]

[]

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

[]

[ "Thanks for creating the issue :)\r\nThe download link for wmt-en-de raw looks like a mirror. We should use that instead of the current url.\r\nIs this mirror official ?\r\n\r\nAlso it looks like for `ro-en` it tried to download other languages. If we manage to only download the one that is asked it'd be cool\r\n\r\nAlso cc @patrickvonplaten ", "Mirror is not official.", "Shall we host the files ourselves or it is fine to use this mirror in your opinion ?", "Should we add an argument in `load_dataset` to override some URL with a custom URL (e.g. mirror) or a local path?\r\n\r\nThis could also be used to provide local files instead of the original files as requested by some users (e.g. when you made a dataset with the same format than SQuAD and what to use it instead of the official dataset files).", "@lhoestq I think we should host it ourselves. I'll put the subset of wmt (without preprocessed files) that we need on s3 and post a link over the weekend.", "Is there a solution yet? The download speed is still too slow. 60-70kbps download for wmt16 and around 100kbps for wmt19. @sshleifer ", "I'm working on mirror links which will provide high download speed :)\r\nSee https://github.com/huggingface/datasets/issues/1892", "Resolved via https://github.com/huggingface/datasets/pull/1912" ]

[ { "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 @eladsegal, thanks for reporting.\r\n\r\nI am sorry, but I can't reproduce the bug:\r\n```\r\nIn [1]: from datasets import load_dataset\r\n\r\nIn [2]: ds = load_dataset(\"qasper\")\r\nDownloading: 5.11kB [00:00, ?B/s]\r\nDownloading and preparing dataset qasper/qasper (download: 9.88 MiB, generated: 35.11 MiB, post-processed: Unknown size, total: 44.99 MiB) to .cache\\qasper\\qasper\\0.1.0\\b99154d2a15aa54bfc669f82b2eda715a2e342e81023d39613b0e2920fdb3ad8...\r\nDataset qasper downloaded and prepared to .cache\\qasper\\qasper\\0.1.0\\b99154d2a15aa54bfc669f82b2eda715a2e342e81023d39613b0e2920fdb3ad8. Subsequent calls will reuse this data.\r\n100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<?, ?it/s]\r\n\r\nIn [3]: ds\r\nOut[3]:\r\nDatasetDict({\r\n train: Dataset({\r\n features: ['id', 'title', 'abstract', 'full_text', 'qas'],\r\n num_rows: 888\r\n })\r\n validation: Dataset({\r\n features: ['id', 'title', 'abstract', 'full_text', 'qas'],\r\n num_rows: 281\r\n })\r\n})\r\n``` \r\n\r\nThis makes me suspect that the origin of the problem might be the cache: I didn't have this dataset in my cache, although I guess you already had it, before the code change introduced by #3110.\r\n\r\n@lhoestq might it be possible that the code change introduced by #3110 makes \"inaccessible\" all previously cached TAR-based datasets?\r\n- Before the caching system downloaded and extracted the tar dataset\r\n- Now it only downloads the tar dataset (no extraction is done)", "I can't reproduce either in my environment (macos, python 3.7).\r\n\r\nIn your case it generates zero examples. This can only happen if the extraction of the TAR archive doesn't output the right filenames. Indeed if the `qasper` script can't find the right file to load, it's currently ignored and it returns zero examples. This case was not even considered when #3110 was developed since we considered the file names to be deterministic - and not depend on your environment.\r\n\r\nTherefore here is my hypothesis:\r\n- either the cache is corrupted somehow with an empty TAR archive\r\n- OR I suspect that the issue comes from python 3.8\r\n", "I just tried again on python 3.8 and I was able to reproduce the issue. Let me work on a fix", "Ok I found the issue. It's not related to python 3.8 in itself though. This issue happens because your local installation of `datasets` is outdated compared to the changes to datasets in #3110\r\n\r\nTo fix this you just have to pull the latest changes from `master` :)\r\n\r\nLet me know if that helps !\r\n\r\n--------------\r\n\r\nHere are more details about my investigation:\r\n\r\nIt's possible to reproduce this issue if you use `datasets<=1.15.1` or before b6469baa22c174b3906c631802a7016fedea6780 and if you load the dataset after revision b6469baa22c174b3906c631802a7016fedea6780. This is because `dl_manager.iter_archive` had issues at that time (and it was not used anywhere anyway).\r\n\r\nIn particular it was returning the absolute path to extracted files instead of the relative path of the file inside the archive. This was an issue because `dl_manager.iter_archive` isn't supposed to extract the TAR archive. Instead, it iterates over all the files inside the archive, without creating a directory with the extracted content.\r\n\r\nTherefore if you want to use the datasets on `master`, make sure that you have an up-to-date local installation of `datasets` as well, or you may face incompatibilities like this.", "Thanks!\r\nBut what about code that is already using older version of datasets? \r\nThe reason I encountered this issue was that suddenly one of my repos with version 1.12.1 started getting 0 examples.\r\nI handled it by adding `revision` to `load_dataset`, but I guess it would still be an issue for other users who doesn't know this.", "Hi, in 1.12.1 it uses the dataset scripts from that time, not the one on master.\r\n\r\nIt only uses the datasets from master if you installed `datasets` from source, or if the dataset isn't available in your local version (in this case it shows a warning and it loads from master).\r\n", "OK, I understand the issue a bit better now.\r\nI see I wasn't on 1.12.1, but on 1.12.1.dev0 and since it is a dev version it uses master.\r\nSo users that use an old dev version must specify revision or else they'll encounter this problem.\r\n\r\nBTW, when I opened the issue I installed the latest master version with\r\n```\r\npip install git+git://github.com/huggingface/datasets@master#egg=datasets\r\n```\r\nand also used `download_mode=\"force_redownload\"`, and it still returned 0 examples.\r\nNow I deleted all of the cache and ran the code again, and it worked.\r\nI'm not sure what exactly happened here, but looks like it was due to a mix of an unofficial version and its cache.\r\n\r\nThanks again!" ]

[]

[ "Google's version of WIT is now available here: https://huggingface.co/datasets/google/wit" ]

[]

[ "Hi ! Thanks for reporting.\r\nI am working on changes in the way data are sliced from arrow. I can probably fix your issue with the changes I'm doing.\r\nIf you have some code to reproduce the issue it would be nice so I can make sure that this case will be supported :)\r\nI'll make a PR in a few days ", "Good if you can take care at your side.\r\nHere is the [colab notebook](https://colab.research.google.com/drive/19c-abm87RTRYgW9G1D8ktfwRW95zDYBZ?usp=sharing)" ]

[ { "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" } ]

[ "This might be interesting to @kayoyin the author of https://github.com/kayoyin/transformer-slt – pinging you just in case :)", "Thanks Amit, this is a great idea! I'm thinking of porting the SLT models from my paper here as well, having this dataset would be perfect for that :)", "Thanks for your contribution, @AmitMY. Are you still interested in adding this dataset?\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\nWe would suggest you create this dataset there. Please, 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" } ]

[ "This has been fixed in this commit: https://github.com/huggingface/datasets/pull/2254/commits/88676c930216cd4cc31741b99827b477d2b46cb6\r\n\r\nIt was introduced in #2246 : using map with `input_columns` doesn't return the other columns anymore" ]

[ { "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" } ]

[ "Are you sure there is a leak? How can I see it? You shared the script but not the output which you believe should indicate a leak.\r\n\r\nI modified your reproduction script to print only once per try as your original was printing too much info and you absolutely must add `gc.collect()` when doing any memory measurements, since python's GC is scheduled so you might be measuring the wrong thing. This gives us:\r\n\r\n```\r\nimport psutil\r\nimport os\r\nimport gc\r\nfrom transformers import BertTokenizer\r\nfrom datasets import load_dataset\r\nfrom torch.utils.data import DataLoader\r\n\r\nBATCH_SIZE = 32\r\nNUM_TRIES = 100\r\n\r\ntokenizer = BertTokenizer.from_pretrained(\"bert-base-uncased\")\r\ndef transform(x):\r\n x.update(tokenizer(x[\"text\"], return_tensors=\"pt\", max_length=64, padding=\"max_length\", truncation=True))\r\n x.pop(\"text\")\r\n x.pop(\"label\")\r\n return x\r\ndataset = load_dataset(\"imdb\", split=\"train\")\r\ndataset.set_transform(transform)\r\ntrain_loader = DataLoader(dataset, batch_size=BATCH_SIZE, shuffle=True, num_workers=4)\r\n\r\nmem_before = psutil.Process(os.getpid()).memory_info().rss / (1024 * 1024)\r\n\r\ncount = 0\r\nwhile count < NUM_TRIES:\r\n for idx, batch in enumerate(train_loader): pass\r\n gc.collect()\r\n mem_after = psutil.Process(os.getpid()).memory_info().rss / (1024 * 1024)\r\n print(count, mem_after - mem_before)\r\n count += 1\r\n```\r\n\r\nNow running it:\r\n\r\n```\r\n$ python dl-leak.py \r\nReusing dataset imdb (/home/stas/.cache/huggingface/datasets/imdb/plain_text/1.0.0/2fdd8b9bcadd6e7055e742a706876ba43f19faee861df134affd7a3f60fc38a1)\r\n0 4.43359375\r\n1 4.4453125\r\n2 4.44921875\r\n3 4.44921875\r\n4 4.4609375\r\n5 4.46484375\r\n6 4.46484375\r\n7 4.46484375\r\n8 4.46484375\r\n9 4.46484375\r\n10 4.46484375\r\n11 4.46484375\r\n12 4.46484375\r\n13 4.46484375\r\n14 4.46484375\r\n15 4.46484375\r\n16 4.46484375\r\n```\r\n\r\nIt's normal that at the beginning there is a small growth in memory usage, but after 5 cycles it gets steady.", "Unless of course you're referring the memory growth during the first try. Is that what you're referring to? And since your ds is small it's hard to see the growth - could it be just because some records are longer and it needs to allocate more memory for those?\r\n\r\nThough while experimenting with this I have observed a peculiar thing, if I concatenate 2 datasets, I don't see any growth at all. But that's probably because the program allocated additional peak RSS memory to concatenate and then is re-using the memory\r\n\r\nI basically tried to see if I make the dataset much longer, I'd expect not to see any memory growth once the 780 records of the imdb ds have been processed once.", "It is hard to say if it is directly reproducible in this setup. Maybe it is specific to the images stored in the CM4 case which cause a memory leak. I am still running your script and seeing if I can reproduce that particular leak in this case.", "I was able to reproduce the leak with:\r\n\r\n```\r\nimport psutil\r\nimport os\r\nimport gc\r\nfrom datasets import load_from_disk\r\nimport time\r\n\r\nDATASET_PATH = \"/hf/m4-master/data/cm4/cm4-10000-v0.1\"\r\n\r\ndataset = load_from_disk(DATASET_PATH)\r\n\r\n# truncate to a tiny dataset\r\ndataset = dataset.select(range(1000))\r\n\r\nprint(f\"dataset: {len(dataset)} records\")\r\n\r\nmem_before = psutil.Process(os.getpid()).memory_info().rss / (1024 * 1024)\r\nfor idx, rec in enumerate(dataset):\r\n if idx % 100 == 0:\r\n gc.collect()\r\n mem_after = psutil.Process(os.getpid()).memory_info().rss / (1024 * 1024)\r\n print(f\"{idx:4d} {mem_after - mem_before:12.4f}MB\")\r\n```\r\nYou need to adjust the DATASET_PATH record.\r\n\r\nwhich you get from\r\n\r\n```\r\ngsutil -m cp \"gs://hf-science-m4/cm4/cm4-10000-v0.1/dataset.arrow\" \"gs://hf-science-m4/cm4/cm4-10000-v0.1/dataset_info.json\" \"gs://hf-science-m4/cm4/cm4-10000-v0.1/state.json\" .\r\n```\r\n(I assume the hf folks have the perms) - it's a smallish dataset (10k)\r\n\r\nthen you run:\r\n```\r\n$ python ds.py\r\ndataset: 1000 records\r\n 0 1.0156MB\r\n 100 126.3906MB\r\n 200 142.8906MB\r\n 300 168.5586MB\r\n 400 218.3867MB\r\n 500 230.7070MB\r\n 600 238.9570MB\r\n 700 263.3789MB\r\n 800 288.1289MB\r\n 900 300.5039MB\r\n```\r\n\r\nyou should be able to see the leak ", "This issue has nothing to do with `PIL`'s decoder. I removed it and the problem is still there.\r\n\r\nI then traced this leak to this single call: `pa_table.to_pydict()` here:\r\n\r\nhttps://github.com/huggingface/datasets/blob/08a7b389cdd6fb49264a72aa8ccfc49a233494b6/src/datasets/formatting/formatting.py#L138-L140\r\n\r\nI can make it leak much faster by modifying that code to repeat `pa_table.to_pydict()` many times in a row. It shouldn't have that impact:\r\n\r\n```\r\nclass PythonArrowExtractor(BaseArrowExtractor[dict, list, dict]):\r\n def extract_row(self, pa_table: pa.Table) -> dict:\r\n x = [pa_table.to_pydict() for x in range(200)]\r\n return _unnest(pa_table.to_pydict())\r\n```\r\n\r\n@lhoestq - do you know what might be happening inside `pa_table.to_pydict()`, as this is in the `pyarrow` domain. Perhaps you know someone to tag from that project?\r\n\r\nProbably next need to remove `datasets` from the equation and make a reproducible case with just `pyarrow` directly.\r\n\r\nThe problem already happens with `pyarrow==6.0.0` or later (minimum for current `datasets`)\r\n\r\nI'm also trying to dig in with `objgraph` to see if there are any circular references which prevent objects from being freed, but no luck there so far. And I'm pretty sure `to_pydict` is not a python code, so the problem is likely to happen somewhere outside of python's GC.", "This appears to be the same issue I think: https://github.com/huggingface/datasets/issues/4528\r\nI dug into the repro code there and it's the same behavior with the same leak, but it's a pure nlp dataset and thus much faster to work with. \r\n", "I went all the way back to `pyarrow==1.0.0` and `datasets==1.12.0` and the problem is still there. How is it even possible that it wasn't noticed all this time. \r\n\r\nCould it be that the leak is in some 3rd party component `pyarrow` relies on? as while downgrading I have only downgraded the above 2 packages.\r\n", "Also found this warning \r\n\r\n> Be careful: if you don't pass the ArrowArray struct to a consumer,\r\n> array memory will leak. This is a low-level function intended for\r\n> expert users.\r\n\r\nsee: https://github.com/apache/arrow/blob/99b57e84277f24e8ec1ddadbb11ef8b4f43c8c89/python/pyarrow/table.pxi#L2515-L2517\r\n\r\nperhaps something triggers this condition?\r\n\r\nI have no idea if it's related - this is just something that came up during my research.", "Does it crash with OOM at some point? If it doesn't, it isn't a leak, just agressive caching or a custom allocator that doesn't like to give memory back (not uncommon). #4528 looks like it hits a steady state.\r\n\r\nI believe the underlying arrow libs use a custom C allocator. Some of those are designed not to give back to OS, but keep heap memory for themselves to re-use (hitting up the OS involves more expensive mutex locks, contention, etc). The greedy behaviour can be undesirable though. There are likely flags to change the allocator behaviour, and one could likely build without any custom allocators (or use a different one).", "> Does it crash with OOM at some point?\r\n\r\nIn the original setup where we noticed this problem, it was indeed ending in an OOM", "> https://github.com/huggingface/datasets/issues/4528 looks like it hits a steady state.\r\n\r\n@rwightman in the plot I shared, the steady state comes from the `time.sleep(100)` I added in the end of the script, to showcase that even the garbage collector couldn't free that allocated memory.\r\n", "Could this be related to this discussion about a potential memory leak in pyarrow: https://issues.apache.org/jira/browse/ARROW-11007 ?\r\n\r\n(Note: I've tried `import pyarrow; pyarrow.jemalloc_set_decay_ms(0)` and the memory leak is still happening on your toy example)", "> @lhoestq - do you know what might be happening inside pa_table.to_pydict(), as this is in the pyarrow domain. Perhaps you know someone to tag from that project?\r\n\r\n`to_pydict` calls `to_pylist` on each column (i.e. on each PyArrow Array). Then it iterates on the array and calls `as_py` on each element. The `as_py` implementation depends on the data type. For strings I think it simply gets the buffer that contains the binary string data that is defined in C++\r\n\r\nThe Arrow team is pretty responsive at [email protected] if it can help\r\n\r\n> Probably next need to remove datasets from the equation and make a reproducible case with just pyarrow directly.\r\n\r\nThat would be ideal indeed. Would be happy to help on this, can you give me access to the bucket so I can try with your data ?", "> That would be ideal indeed. Would be happy to help on this, can you give me access to the bucket so I can try with your data ?\r\n\r\nI added you to the bucket @lhoestq ", "It looks like an issue with memory mapping:\r\n- the amount of memory used in the end corresponds to the size of the dataset\r\n- setting `keep_in_memory=True` in `load_from_disk` loads the dataset in RAM, and **doesn't cause any memory leak**", "Here is a code to reproduce this issue using only PyArrow and a dummy arrow file:\r\n```python\r\nimport psutil\r\nimport os\r\nimport gc\r\nimport pyarrow as pa\r\nimport time\r\n\r\nARROW_PATH = \"tmp.arrow\"\r\n\r\nif not os.path.exists(ARROW_PATH):\r\n arr = pa.array([b\"a\" * (200 * 1024)] * 1000) # ~200MB\r\n table = pa.table({\"a\": arr})\r\n\r\n with open(ARROW_PATH, \"wb\") as f:\r\n writer = pa.RecordBatchStreamWriter(f, schema=table.schema)\r\n writer.write_table(table)\r\n writer.close()\r\n\r\n\r\ndef memory_mapped_arrow_table_from_file(filename: str) -> pa.Table:\r\n memory_mapped_stream = pa.memory_map(filename)\r\n opened_stream = pa.ipc.open_stream(memory_mapped_stream)\r\n pa_table = opened_stream.read_all()\r\n return pa_table\r\n\r\n\r\ntable = memory_mapped_arrow_table_from_file(ARROW_PATH)\r\narr = table[0]\r\n\r\nmem_before = psutil.Process(os.getpid()).memory_info().rss / (1024 * 1024)\r\nfor idx, x in enumerate(arr):\r\n if idx % 100 == 0:\r\n gc.collect()\r\n time.sleep(0.1)\r\n mem_after = psutil.Process(os.getpid()).memory_info().rss / (1024 * 1024)\r\n print(f\"{idx:4d} {mem_after - mem_before:12.4f}MB\")\r\n```\r\nprints\r\n```\r\n 0 0.2500MB\r\n 100 19.8008MB\r\n 200 39.3320MB\r\n 300 58.8633MB\r\n 400 78.3945MB\r\n 500 97.9258MB\r\n 600 117.4570MB\r\n 700 136.9883MB\r\n 800 156.5195MB\r\n 900 176.0508MB\r\n```\r\nNote that this example simply iterates over the `pyarrow.lib.BinaryScalar` objects in the array. Running `.as_py()` is not needed to experience the memory issue.", "@lhoestq that does indeed increase in memory, but if you iterate over array again after the first time, or re-open and remap the same file (repeat `table = memory_mapped_arrow_table_from_file(ARROW_PATH)`) before re-iterating, it doesn't move pas 195MB.... it would appear another step is needed to continue consuming memory past that.. hmmm\r\n\r\nAre the pa_tables held on to anywhere after they are iterated in the real code?\r\n\r\nin my hack, if you do a bunch cut & paste and then change the arr name for each iter \r\n\r\n```\r\ntable = memory_mapped_arrow_table_from_file(ARROW_PATH)\r\narr = table[0]\r\n\r\nfor idx, x in enumerate(arr):\r\n if idx % 100 == 0:\r\n gc.collect()\r\n time.sleep(0.1)\r\n mem_after = psutil.Process(os.getpid()).memory_info().rss / (1024 * 1024)\r\n print(f\"{idx:4d} {mem_after - mem_before:12.4f}MB\")\r\n\r\ntable = memory_mapped_arrow_table_from_file(ARROW_PATH)\r\narr1 = table[0]\r\n\r\nfor idx, x in enumerate(arr1):\r\n if idx % 100 == 0:\r\n gc.collect()\r\n time.sleep(0.1)\r\n mem_after = psutil.Process(os.getpid()).memory_info().rss / (1024 * 1024)\r\n print(f\"{idx:4d} {mem_after - mem_before:12.4f}MB\")\r\n\r\ntable = memory_mapped_arrow_table_from_file(ARROW_PATH)\r\narr2 = table[0]\r\n\r\nfor idx, x in enumerate(arr2):\r\n if idx % 100 == 0:\r\n gc.collect()\r\n time.sleep(0.1)\r\n mem_after = psutil.Process(os.getpid()).memory_info().rss / (1024 * 1024)\r\n print(f\"{idx:4d} {mem_after - mem_before:12.4f}MB\")\r\n```\r\n\r\nit leaks, if all arr are the same name (so prev one gets cleaned up) it does not and goes back to 0, anything that could be holding onto a reference of an intermediary equivalent like arr in the real use case?\r\n\r\n\r\n\r\n", "Yes, we have already established here https://github.com/huggingface/datasets/issues/4883#issuecomment-1232063891 that when one iterates over the whole dataset multiple times, it consumes a bit more memory in the next few repetitions and then remains steady. \r\n\r\nWhich means that when a new iterator is created over the same dataset, all the memory from the previous iterator is re-used.\r\n\r\nSo the leak happens primarily when the iterator is \"drained\" the first time. which tells me that either a circular reference is created somewhere which only gets released when the iterator is destroyed, or there is some global variable that keeps piling up the memory and doesn't release it in time.\r\n\r\nAlso I noticed some `__del__` methods which won't destroy objects automatically and there is usually a warning against using it https://stackoverflow.com/a/1481512/9201239\r\n\r\nThere are also some `weakref`s in the code which too may lead to leaks or weird problems at times.\r\n", "@stas00 my point was, I'm not convinced @lhoestq last example illustrates the leak, but rather the differences between memory mapping and in memory usage patterns. If you destroy arr, memory map impl goes back to 0 each iteration. The amount of memory that 'looks' like it is leaked in first pass differes quite a bit between memory mapped vs in memory, but the underlying issue likely a circular reference, or reference(s) which were not cleaned up that would impact either case, but likely much more visible with mmap.", "Thank you for clarifying, Ross. \r\n\r\nI think we agree that it's almost certain that the `datasets` iterator traps some inner variable that prevents object freeing, since if we create the iterator multiple times (and drain it) after a few runs no new memory is allocated. We could try to dig in more with `objgraph` - my main concern is if the problem happens somewhere outside of python, (i.e. in pyarrow cpp implementation) in which case it'd be much more difficult to trace. \r\n\r\nI wish there was a way on linux to tell the program to free no longer used memory at will.", "FWIW, I revisted some code I had in the works to use HF datasets w/ timm train & val scripts. There is no leak there across multipe epochs. It uses the defaults. \r\n\r\nIt's worth noting that with imagenet `keep_in_memory=True` isn't even an option because the train arrow file is ~140GB and my local memory is less. The virtual address space reflects mmap (> 150GB) and doesn't increase over epochs that I noticed. I have some perf issues to bring up wrt to the current setup, but that's a separate and lower prio discussion to have elsewhere...", "# Notes \r\n\r\nAfter reading many issues and trying many things here is the summary of my learning\r\n\r\nI'm now using @lhoestq repro case as it's pyarrow-isolated: https://github.com/huggingface/datasets/issues/4883#issuecomment-1242034985\r\n\r\n\r\n## 1. pyarrow memory backends\r\n\r\nit has 3 backends, I tried them all with the same results\r\n\r\n```\r\npa.set_memory_pool(pa.jemalloc_memory_pool())\r\npa.set_memory_pool(pa.mimalloc_memory_pool())\r\npa.set_memory_pool(pa.system_memory_pool())\r\n```\r\n\r\n## 2. quick release\r\n\r\nThe `jemalloc` backend supports quick release\r\n\r\n```\r\npa.jemalloc_set_decay_ms(0)\r\n```\r\n\r\nit doesn't make any difference in this case\r\n\r\n## 3. actual memory allocations\r\n\r\nthis is a useful tracer for PA memory allocators\r\n```\r\npa.log_memory_allocations(enable=True)\r\n```\r\n\r\nit nicely reports memory allocations and releases when the arrow file is created the first time.\r\n\r\nbut when we then try to do `enumerate(arr)` this logger reports 0 allocations.\r\n\r\nThis summary also reports no allocations when the script run the second time (post file creation):\r\n```\r\nmem_pool = pa.default_memory_pool()\r\nprint(f\"PyArrow mem pool info: {mem_pool.backend_name} backend, {mem_pool.bytes_allocated()} allocated, \"\r\n f\"{mem_pool.max_memory()} max allocated, \")\r\n\r\nprint(f\"PyArrow total allocated bytes: {pa.total_allocated_bytes()}\")\r\n```\r\n\r\nHowever it's easy to see by using `tracemalloc` which only measures python's memory allocations that it's PA that leaks, since `tracemalloc` shows fixed memory\r\n\r\n(this is bolted on top of the original repro script)\r\n\r\n```\r\nimport tracemalloc\r\ntracemalloc.start()\r\n\r\n[...]\r\nfor idx, x in enumerate(arr):\r\n if idx % 10 == 0:\r\n gc.collect()\r\n time.sleep(0.1)\r\n mem_after = psutil.Process(os.getpid()).memory_info().rss / 2**20\r\n mem_use = pa.total_allocated_bytes() - start_use\r\n mem_peak = pool.max_memory() - start_peak_use\r\n\r\n second_size, second_peak = tracemalloc.get_traced_memory()\r\n mem_diff = (second_size - first_size) / 2**20\r\n mem_peak_diff = (second_peak - first_peak) / 2**20\r\n\r\n # pa.jemalloc_memory_pool().release_unused()\r\n # pa.mimalloc_memory_pool().release_unused()\r\n # pa.system_memory_pool().release_unused()\r\n\r\n print(f\"{idx:4d} {mem_after - mem_before:12.4f}MB {mem_diff:12.4f} {mem_peak_diff:12.4f} {memory_mapped_stream.size()/2**20:4.4}MB {mem_use/2**20:4.4}MB {mem_peak/2**20:4.4}MB\")\r\n\r\n```\r\n\r\ngives:\r\n\r\n```\r\n 0 5.4258MB 0.0110 0.0201 195.3MB 0.0MB 0.0MB\r\n 10 25.3672MB 0.0112 0.0202 195.3MB 0.0MB 0.0MB\r\n 20 45.9336MB 0.0112 0.0203 195.3MB 0.0MB 0.0MB\r\n 30 62.4336MB 0.0112 0.0203 195.3MB 0.0MB 0.0MB\r\n 40 83.0586MB 0.0112 0.0203 195.3MB 0.0MB 0.0MB\r\n 50 103.6836MB 0.0112 0.0203 195.3MB 0.0MB 0.0MB\r\n 60 124.3086MB 0.0112 0.0203 195.3MB 0.0MB 0.0MB\r\n 70 140.8086MB 0.0112 0.0203 195.3MB 0.0MB 0.0MB\r\n 80 161.4336MB 0.0112 0.0203 195.3MB 0.0MB 0.0MB\r\n 90 182.0586MB 0.0112 0.0203 195.3MB 0.0MB 0.0MB\r\n```\r\n\r\nthe 3rd and 4th columns are `tracemalloc`'s report.\r\n\r\nthe 5th column is the size of mmaped stream - fixed.\r\n\r\nthe last 2 are the PA's malloc reports - you can see it's totally fixed and 0.\r\n\r\nSo what gives? PA's memory allocator says nothing was allocated and we can see python doesn't allocate any memory either.\r\n\r\nAs someone suggested in one of the PA issues that **IPC/GRPC could be the issue.** Any suggestions on how debug this one?\r\n\r\nThe main issue is that one can't step through with a python debugger as `arr` is an opaque cpp object binded to python.\r\n\r\nPlease see the next comment for a possible answer.\r\n\r\n# ref-count\r\n\r\nI also traced reference counts and they are all fixed using either `sys.getrefcount(x)` or `len(gc.get_referrers(x))`\r\n\r\nso it's not the python object\r\n\r\n# Important related discussions\r\n\r\nhttps://issues.apache.org/jira/browse/ARROW-11007 - looks very similar to our issue\r\nin particular this part of the report:\r\nhttps://issues.apache.org/jira/browse/ARROW-11007?focusedCommentId=17279642&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-17279642\r\n", "# There is no leak, just badly communicated linux RSS memory usage stats\r\n\r\nNext, lets revisit @rwightman's suggestion that there is actually no leak.\r\n\r\nAfter all - we are using mmap which **will try to map** the file to RAM as much as it can and then page out if there is no memory. i.e. MMAP is only fast if you have a lot of CPU RAM.\r\n\r\nSo let's do it:\r\n\r\n# Memory mapping OOM test\r\n\r\nWe first quickly start a cgroups-controlled shell which will instantly kill any program that consumes more than 1GB of memory:\r\n\r\n```\r\n$ systemd-run --user --scope -p MemoryHigh=1G -p MemoryMax=1G -p MemorySwapMax=1G --setenv=\"MEMLIMIT=1GB\" bash\r\n```\r\n\r\nLet's check that it indeed does so. Let's change @lhoestq's script to allocate a 10GB arrow file:\r\n\r\n```\r\n$ python -c 'import pyarrow as pa; pa.array([b\"a\" * (2000 * 1024)] * 5000)'\r\nKilled\r\n```\r\noops, that didn't work, as we tried to allocate 10GB when only 1GB is allowed. This is what we want!\r\n\r\nLet's do a sanity check - can we allocate 0.1GB?\r\n```\r\npython -c 'import pyarrow as pa; pa.array([b\"a\" * (2000 * 1024)] * 50)'\r\n```\r\n\r\nYes. So the limited shell does the right thing. It let's allocate `< 1GB` of RSS RAM.\r\n\r\nNext let's go back to @lhoestq's script but with 10GB arrow file.\r\n\r\nwe change his repro script https://github.com/huggingface/datasets/issues/4883#issuecomment-1242034985 to 50x larger file\r\n```\r\n arr = pa.array([b\"a\" * (2000 * 1024)] * 5000) # ~10000MB\r\n```\r\nwe first have to run into a normal unlimited shell so that we don't get killed (as the script allocates 10GB)\r\n\r\nlet's run the script now in the 1GB-limited shell while running a monitor:\r\n\r\n```\r\n$ htop -F python -s M_RESIDENT -u `whoami`\r\n```\r\n\r\nso we have 2 sources of RSS info just in case.\r\n\r\n```\r\n$ python pyar.py\r\n 0 4.3516MB 0.0103 0.0194 9.766e+03MB 0.0MB 0.0MB\r\n 10 24.3008MB 0.0104 0.0195 9.766e+03MB 0.0MB 0.0MB\r\n[...]\r\n4980 9730.3672MB 0.0108 0.0199 9.766e+03MB 0.0MB 0.0MB\r\n4990 9750.9922MB 0.0108 0.0199 9.766e+03MB 0.0MB 0.0MB\r\nPyArrow mem pool info: jemalloc backend, 0 allocated, 0 max allocated,\r\nPyArrow total allocated bytes: 0\r\n```\r\n\r\nBut wait, it reported 10GB RSS both in `htop` and in our log!\r\n\r\nSo that means it never allocated 10GB otherwise it'd have been killed.\r\n\r\n**Which tells us that there is no leak whatsoever** and this is just a really difficult situation where MMAPPED memory is reported as part of RSS which it probably shouldn't. As now we have no way how to measure real memory usage.\r\n\r\nI also attached the script with all the different things I have tried in it, so it should be easy to turn them on/off if you want to reproduce any of my findings.\r\n\r\n[pyar.txt](https://github.com/huggingface/datasets/files/9539430/pyar.txt)\r\n\r\njust rename it to `pyra.py` as gh doesn't let attaching scripts...\r\n\r\n(I have to remember to exit that special mem-limited shell or else I won't be able to do anything serious there.)\r\n\r\n", "The original leak in the multi-modal code is very likely something else. But of course now it'd be very difficult to trace it using mmap.\r\n\r\nI think to debug we have to set `keep_in_memory=True` in `load_from_disk` to load the small dataset in RAM, so there will be no mmap misleading reporting component and then continue searching for another source of a leak.", "To add to what @stas00 found, I'm gonna leave some links to where I believe the confusion came from in pyarrow's APIs, for future reference:\r\n* In the section where they talk about [efficiently writing and reading arrow data](https://arrow.apache.org/docs/dev/python/ipc.html?#efficiently-writing-and-reading-arrow-data), they give an example of how \r\n\r\n> Arrow can directly reference the data mapped from disk and avoid having to allocate its own memory. \r\n\r\nAnd where their example shows 0 RSS memory allocation, the way we used to measure RSS shows 39.6719MB allocated. Here's the script to reproduce:\r\n```\r\nimport psutil\r\nimport os\r\nimport gc\r\nimport pyarrow as pa\r\nimport time\r\nimport sys\r\n\r\n# gc.set_debug(gc.DEBUG_LEAK)\r\n# gc.set_threshold(0,0,0)\r\n\r\n#pa.set_memory_pool(pa.mimalloc_memory_pool())\r\n#pa.set_memory_pool(pa.system_memory_pool())\r\n\r\nimport tracemalloc\r\n\r\n#pa.jemalloc_set_decay_ms(0)\r\n# pa.log_memory_allocations(enable=True)\r\n\r\nBATCH_SIZE = 10000\r\nNUM_BATCHES = 1000\r\nschema = pa.schema([pa.field('nums', pa.int32())])\r\nwith pa.OSFile('bigfile.arrow', 'wb') as sink:\r\n with pa.ipc.new_file(sink, schema) as writer:\r\n for row in range(NUM_BATCHES):\r\n batch = pa.record_batch([pa.array(range(BATCH_SIZE), type=pa.int32())], schema)\r\n writer.write(batch)\r\n\r\nstart_use = pa.total_allocated_bytes()\r\npool = pa.default_memory_pool()\r\nstart_peak_use = pool.max_memory()\r\ntracemalloc.start()\r\nfirst_size, first_peak = tracemalloc.get_traced_memory()\r\nmem_before = psutil.Process(os.getpid()).memory_info().rss / 2**20\r\n\r\n\r\n# with pa.OSFile('bigfile.arrow', 'rb') as source:\r\n# loaded_array = pa.ipc.open_file(source).read_all()\r\n\r\nwith pa.memory_map('bigfile.arrow', 'rb') as source:\r\n loaded_array = pa.ipc.open_file(source).read_all()\r\n\r\n\r\nprint(\"LEN:\", len(loaded_array))\r\nprint(\"RSS: {}MB\".format(pa.total_allocated_bytes() >> 20))\r\n\r\ngc.collect()\r\ntime.sleep(0.1)\r\nmem_after = psutil.Process(os.getpid()).memory_info().rss / 2**20\r\nmem_use = pa.total_allocated_bytes() - start_use\r\nmem_peak = pool.max_memory() - start_peak_use\r\nsecond_size, second_peak = tracemalloc.get_traced_memory()\r\nmem_diff = (second_size - first_size) / 2**20\r\nmem_peak_diff = (second_peak - first_peak) / 2**20\r\n\r\nidx = 0\r\nprint(f\"{idx:4d} {mem_after - mem_before:12.4f}MB {mem_diff:12.4f} {mem_peak_diff:12.4f} {mem_use/2**20:4.4}MB {mem_peak/2**20:4.4}MB\")\r\n```\r\ngives:\r\n```\r\n\r\nLEN: 10000000\r\nRSS: 0MB\r\n 0 39.6719MB 0.0132 0.0529 0.0MB 0.0MB\r\n```\r\nWhich again just proves that we uncorrectly measure RSS, in the case of MMAPPED memory\r\n\r\n\r\n* [The recommended way to do memory profiling from Arrow's docs](https://arrow.apache.org/docs/dev/cpp/memory.html#memory-profiling)\r\n", "@lhoestq, I have been working on a detailed article that shows that MMAP doesn't leak and it's mostly ready. I will share when it's ready.\r\n\r\nThe issue is that we still need to be able to debug memory leaks by turning MMAP off.\r\n\r\nBut, once I tried to show the user that using `load_dataset(... keep_in_memory=True)` is the way to debug an actual memory leak - guess I what I discovered? A potential actual leak.\r\n\r\nHere is the repro:\r\n\r\n```\r\n$ cat ds-mmap.py\r\nfrom datasets import load_dataset\r\nimport gc\r\nimport os\r\nimport psutil\r\n\r\nproc = psutil.Process(os.getpid())\r\ndef mem_read():\r\n gc.collect()\r\n return proc.memory_info().rss / 2**20\r\n\r\ndataset = load_dataset(\"wmt19\", 'cs-en', keep_in_memory=True, streaming=False)['train']\r\n\r\nprint(f\"{'idx':>6} {'RSS':>10} {'Δ RSS':>15}\")\r\nstep = 20000\r\nfor i in range(0, 10*step, step):\r\n mem_before = mem_read()\r\n _ = dataset[i:i+step]\r\n mem_after = mem_read()\r\n print(f\"{i:6d} {mem_after:12.4f}MB {mem_after - mem_before:12.4f}MB \")\r\n```\r\n\r\n```\r\npython ds-io.py\r\nReusing dataset wmt19 (/home/stas/.cache/huggingface/datasets/wmt19/cs-en/1.0.0/c3db1bf4240362ed1ef4673b354f468d70aac66d4e67d45f536d493a0840f0d3)\r\n100%|███████████████████████████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 5.66it/s]\r\n idx RSS Δ RSS\r\n 0 1398.4609MB 3.5195MB\r\n 20000 1398.5742MB 0.1133MB\r\n 40000 1398.6016MB 0.0273MB\r\n 60000 1398.6016MB 0.0000MB\r\n 80000 1398.6016MB 0.0000MB\r\n100000 1398.6328MB 0.0312MB\r\n120000 1398.6953MB 0.0625MB\r\n140000 1398.6953MB 0.0000MB\r\n160000 1398.7500MB 0.0547MB\r\n180000 1398.7500MB 0.0000MB\r\n```", "as I suggested on slack perhaps it was due to dataset records length variation, so with your help I wrote another repro with synthetic records which are all identical - which should remove my hypothese from the equation and we should expect 0 incremental growth as we iterate over the datasets. But alas this is not the case. There is a tiny but definite leak-like behavior.\r\n\r\nHere is the new repro:\r\n\r\n```\r\n$ cat ds-synthetic-no-mmap.py\r\nfrom datasets import load_from_disk, Dataset\r\nimport gc\r\nimport sys\r\nimport os\r\nimport psutil\r\n\r\nproc = psutil.Process(os.getpid())\r\ndef mem_read():\r\n gc.collect()\r\n return proc.memory_info().rss / 2**20\r\n\r\nDS_PATH = \"synthetic-ds\"\r\nif not os.path.exists(DS_PATH):\r\n records = 1_000_000\r\n print(\"Creating a synthetic dataset\")\r\n row = dict(foo=[dict(a='a'*500, b='b'*1000)])\r\n ds = Dataset.from_dict({k: [v] * records for k, v in row.items()})\r\n ds.save_to_disk(DS_PATH)\r\n print(\"Done. Please restart the program\")\r\n sys.exit()\r\n\r\ndataset = load_from_disk(DS_PATH, keep_in_memory=True)\r\nprint(f\"Dataset len={len(dataset)}\")\r\n\r\nprint(f\"{'idx':>8} {'RSS':>10} {'Δ RSS':>15}\")\r\nmem_start = 0\r\nstep = 25_000\r\nwarmup_iterations = 4\r\nfor idx, i in enumerate(range(0, len(dataset), step)):\r\n if idx == warmup_iterations: # skip the first few iterations while things get set up\r\n mem_start = mem_read()\r\n mem_before = mem_read()\r\n _ = dataset[i:i+step]\r\n mem_after = mem_read()\r\n print(f\"{i:8d} {mem_after:12.4f}MB {mem_after - mem_before:12.4f}MB\")\r\nmem_end = mem_read()\r\n\r\nprint(f\"Total diff: {mem_end - mem_start:12.4f}MB (after {warmup_iterations} warmup iterations)\")\r\n```\r\n\r\nand the run:\r\n\r\n```\r\n$ python ds-synthetic-no-mmap.py\r\nDataset len=1000000\r\n idx RSS Δ RSS\r\n 0 1601.9258MB 47.9688MB\r\n 25000 1641.6289MB 39.7031MB\r\n 50000 1641.8594MB 0.2305MB\r\n 75000 1642.1289MB 0.2695MB\r\n 100000 1642.1289MB 0.0000MB\r\n 125000 1642.3789MB 0.2500MB\r\n 150000 1642.3789MB 0.0000MB\r\n 175000 1642.6289MB 0.2500MB\r\n 200000 1642.6289MB 0.0000MB\r\n 225000 1642.8789MB 0.2500MB\r\n 250000 1642.8828MB 0.0039MB\r\n 275000 1643.1328MB 0.2500MB\r\n 300000 1643.1328MB 0.0000MB\r\n 325000 1643.3828MB 0.2500MB\r\n 350000 1643.3828MB 0.0000MB\r\n 375000 1643.6328MB 0.2500MB\r\n 400000 1643.6328MB 0.0000MB\r\n 425000 1643.8828MB 0.2500MB\r\n 450000 1643.8828MB 0.0000MB\r\n 475000 1644.1328MB 0.2500MB\r\n 500000 1644.1328MB 0.0000MB\r\n 525000 1644.3828MB 0.2500MB\r\n 550000 1644.3828MB 0.0000MB\r\n 575000 1644.6328MB 0.2500MB\r\n 600000 1644.6328MB 0.0000MB\r\n 625000 1644.8828MB 0.2500MB\r\n 650000 1644.8828MB 0.0000MB\r\n 675000 1645.1328MB 0.2500MB\r\n 700000 1645.1328MB 0.0000MB\r\n 725000 1645.3828MB 0.2500MB\r\n 750000 1645.3828MB 0.0000MB\r\n 775000 1645.6328MB 0.2500MB\r\n 800000 1645.6328MB 0.0000MB\r\n 825000 1645.8828MB 0.2500MB\r\n 850000 1645.8828MB 0.0000MB\r\n 875000 1646.1328MB 0.2500MB\r\n 900000 1646.1328MB 0.0000MB\r\n 925000 1646.3828MB 0.2500MB\r\n 950000 1646.3828MB 0.0000MB\r\n 975000 1646.6328MB 0.2500MB\r\nTotal diff: 4.5039MB (after 4 warmup iterations)\r\n```\r\nso I'm still not sure why we get this.\r\n\r\nAs you can see I started skipping the first few iterations where memory isn't stable yet. As the actual diff is much larger if we count all iterations.\r\n\r\nWhat do you think?", "@stas00 my 2 cents from having looked at a LOT of memory leaks over the years, esp in Python, .3% memory increase over that many iterations of something is difficult to say with certainty it is a leak. \r\n\r\nAlso, just looking at RSS makes it hard to analyze leaks. RSS can stay near constant while you are leaking. RSS is paged in mem, if you have a big leak your RSS might not increase much (leaked mem tends not to get used again so often paged out) while your virtual page allocation could be going through the roof...", "yes, that's true, but unless the leak is big, I'm yet to find another measurement tool.\r\n\r\nTo prove your point here is a very simple IO in a loop program that also reads the same line all over again:\r\n\r\n```\r\n$ cat mmap-no-leak-debug.py\r\nimport gc\r\nimport mmap\r\nimport os\r\nimport psutil\r\nimport sys\r\n\r\nproc = psutil.Process(os.getpid())\r\n\r\nPATH = \"./tmp.txt\"\r\n\r\ndef mem_read():\r\n gc.collect()\r\n return proc.memory_info().rss / 2**20\r\n\r\n# create a large data file with a few long lines\r\nif not os.path.exists(PATH):\r\n with open(PATH, \"w\") as fh:\r\n s = 'a'* 2**27 + \"\\n\" # 128MB\r\n # write ~2GB file\r\n for i in range(16):\r\n fh.write(s)\r\n\r\nprint(f\"{'idx':>4} {'RSS':>10} {'Δ RSS':>12} {'Δ accumulated':>10}\")\r\n\r\ntotal_read = 0\r\ncontent = ''\r\nmem_after = mem_before_acc = mem_after_acc = mem_before = proc.memory_info().rss / 2**20\r\nprint(f\"{0:4d} {mem_after:10.2f}MB {mem_after - 0:10.2f}MB {0:10.2f}MB\")\r\n\r\nmmap_mode = True if \"--mmap\" in sys.argv else False\r\n\r\nwith open(PATH, \"r\") as fh:\r\n\r\n if mmap_mode:\r\n mm = mmap.mmap(fh.fileno(), 0, access=mmap.ACCESS_READ)\r\n\r\n idx = 0\r\n while True:\r\n idx += 1\r\n mem_before = mem_read()\r\n line = mm.readline() if mmap_mode else fh.readline()\r\n if not line:\r\n break\r\n\r\n #total_read += len(line)\r\n\r\n if \"--accumulate\" in sys.argv:\r\n mem_before_acc = mem_read()\r\n content += str(line)\r\n mem_after_acc = mem_read()\r\n\r\n mem_after = mem_read()\r\n\r\n print(f\"{idx:4d} {mem_after:10.2f}MB {mem_after - mem_before:10.2f}MB {mem_after_acc - mem_before_acc:10.2f}MB\")\r\n```\r\n\r\nit has some other instrumentations to do mmap and accumulate data, but let's ignore that for now.\r\n\r\nHere it is running in a simple non-mmap IO:\r\n\r\n```\r\n$ python mmap-no-leak-debug.py\r\n idx RSS Δ RSS Δ accumulated\r\n 0 12.43MB 12.43MB 0.00MB\r\n 1 269.72MB 257.29MB 0.00MB\r\n 2 269.73MB 0.02MB 0.00MB\r\n 3 269.73MB 0.00MB 0.00MB\r\n 4 269.74MB 0.01MB 0.00MB\r\n 5 269.74MB 0.00MB 0.00MB\r\n 6 269.75MB 0.01MB 0.00MB\r\n 7 269.75MB 0.00MB 0.00MB\r\n 8 269.76MB 0.01MB 0.00MB\r\n 9 269.76MB 0.00MB 0.00MB\r\n 10 269.77MB 0.01MB 0.00MB\r\n 11 269.77MB 0.00MB 0.00MB\r\n 12 269.77MB 0.00MB 0.00MB\r\n 13 269.77MB 0.00MB 0.00MB\r\n 14 269.77MB 0.00MB 0.00MB\r\n 15 269.77MB 0.00MB 0.00MB\r\n 16 146.02MB -123.75MB 0.00MB\r\n```\r\n\r\nas you can see even this super-simplistic program that just performs `readline()` slightly increases in RSS over iterations.\r\n\r\nIf you have a better tool for measurement other than RSS, I'm all ears.", "@stas00 if you aren't using memory maps, you should be able to clearly see the increase in the virtual mem for the process as well. Even then, it could still be challenging to determine if it's leak vs fragmentation due to problematic allocation patterns (not uncommon with Python). Using a better mem allocator like tcmalloc via LD_PRELOAD hooks could reduce impact of fragmentation across both Python and c libs. Not sure that plays nice with any allocator that arrow might use itself though. " ]

[ { "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": "d93f0b", "default": false, "description": "", "id": 2725241052, "name": "speech", "node_id": "MDU6TGFiZWwyNzI1MjQxMDUy", "url": "https://api.github.com/repos/huggingface/datasets/labels/speech" } ]

[]

[ { "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" } ]