Datasets:

---

Francesco-A

/

github-issues_huggingface-datasets

like 0

[ { "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" } ]

[ "Just seen https://github.com/huggingface/datasets/issues/5281" ]

[]

[ "Hi ! Currently we don't have a way to `zip` datasets but we plan to add this soon :)\r\nFor now you'll need to use `map` to add the fields from one dataset to the other. See the comment here for more info : https://github.com/huggingface/datasets/issues/853#issuecomment-727872188", "Good to hear.\r\nCurrently I did not use map , just fetch src and tgt from the 2 dataset and merge them.\r\nIt will be a release if you can deal with it at the backend.\r\nThanks.", "Hi! You can rename the columns and concatenate the datasets along `axis=1` to get the desired result as follows:\r\n```python\r\nds1 = ds1.rename_column(\"text\", \"src\")\r\nds2 = ds2.rename_column(\"text\", \"tgt\")\r\nds = datasets.concatenate_datasets([\"ds1\", \"ds2\"], axis=1)\r\n```" ]

[]

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

[]

[]

[]

[ "<details>\n<summary>Show benchmarks</summary>\n\nPyArrow==6.0.0\n\n<details>\n<summary>Show updated benchmarks!</summary>\n\n### Benchmark: benchmark_array_xd.json\n\n| metric | read_batch_formatted_as_numpy after write_array2d | read_batch_formatted_as_numpy after write_flattened_sequence | read_batch_formatted_as_numpy after write_nested_sequence | read_batch_unformated after write_array2d | read_batch_unformated after write_flattened_sequence | read_batch_unformated after write_nested_sequence | read_col_formatted_as_numpy after write_array2d | read_col_formatted_as_numpy after write_flattened_sequence | read_col_formatted_as_numpy after write_nested_sequence | read_col_unformated after write_array2d | read_col_unformated after write_flattened_sequence | read_col_unformated after write_nested_sequence | read_formatted_as_numpy after write_array2d | read_formatted_as_numpy after write_flattened_sequence | read_formatted_as_numpy after write_nested_sequence | read_unformated after write_array2d | read_unformated after write_flattened_sequence | read_unformated after write_nested_sequence | write_array2d | write_flattened_sequence | write_nested_sequence |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.008740 / 0.011353 (-0.002613) | 0.004501 / 0.011008 (-0.006507) | 0.100045 / 0.038508 (0.061537) | 0.029999 / 0.023109 (0.006890) | 0.303556 / 0.275898 (0.027658) | 0.335342 / 0.323480 (0.011863) | 0.006996 / 0.007986 (-0.000989) | 0.004183 / 0.004328 (-0.000145) | 0.076434 / 0.004250 (0.072183) | 0.033899 / 0.037052 (-0.003153) | 0.301312 / 0.258489 (0.042823) | 0.343136 / 0.293841 (0.049295) | 0.034062 / 0.128546 (-0.094484) | 0.011465 / 0.075646 (-0.064181) | 0.323134 / 0.419271 (-0.096137) | 0.040820 / 0.043533 (-0.002713) | 0.301708 / 0.255139 (0.046569) | 0.329528 / 0.283200 (0.046328) | 0.088393 / 0.141683 (-0.053290) | 1.460996 / 1.452155 (0.008842) | 1.531145 / 1.492716 (0.038429) |\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) | 0.191918 / 0.018006 (0.173912) | 0.414099 / 0.000490 (0.413610) | 0.000411 / 0.000200 (0.000211) | 0.000060 / 0.000054 (0.000005) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.022707 / 0.037411 (-0.014704) | 0.096991 / 0.014526 (0.082465) | 0.106070 / 0.176557 (-0.070487) | 0.151275 / 0.737135 (-0.585860) | 0.108909 / 0.296338 (-0.187430) |\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) | 0.422499 / 0.215209 (0.207289) | 4.205551 / 2.077655 (2.127896) | 1.918960 / 1.504120 (0.414841) | 1.715421 / 1.541195 (0.174227) | 1.768969 / 1.468490 (0.300479) | 0.692243 / 4.584777 (-3.892534) | 3.382452 / 3.745712 (-0.363260) | 1.943695 / 5.269862 (-3.326166) | 1.250482 / 4.565676 (-3.315195) | 0.082084 / 0.424275 (-0.342191) | 0.012446 / 0.007607 (0.004839) | 0.525584 / 0.226044 (0.299539) | 5.275530 / 2.268929 (3.006602) | 2.386207 / 55.444624 (-53.058418) | 2.043920 / 6.876477 (-4.832557) | 2.030932 / 2.142072 (-0.111140) | 0.810233 / 4.805227 (-3.994994) | 0.148139 / 6.500664 (-6.352525) | 0.064617 / 0.075469 (-0.010852) |\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) | 1.227352 / 1.841788 (-0.614436) | 13.527623 / 8.074308 (5.453315) | 14.018551 / 10.191392 (3.827159) | 0.140333 / 0.680424 (-0.540091) | 0.028349 / 0.534201 (-0.505852) | 0.394904 / 0.579283 (-0.184379) | 0.406532 / 0.434364 (-0.027831) | 0.471714 / 0.540337 (-0.068624) | 0.568517 / 1.386936 (-0.818419) |\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) | 0.006623 / 0.011353 (-0.004730) | 0.004464 / 0.011008 (-0.006544) | 0.076342 / 0.038508 (0.037834) | 0.027451 / 0.023109 (0.004341) | 0.343851 / 0.275898 (0.067953) | 0.385723 / 0.323480 (0.062243) | 0.005624 / 0.007986 (-0.002362) | 0.004685 / 0.004328 (0.000356) | 0.075669 / 0.004250 (0.071419) | 0.037297 / 0.037052 (0.000244) | 0.343363 / 0.258489 (0.084874) | 0.396115 / 0.293841 (0.102274) | 0.031577 / 0.128546 (-0.096970) | 0.011557 / 0.075646 (-0.064090) | 0.085626 / 0.419271 (-0.333645) | 0.041699 / 0.043533 (-0.001834) | 0.340826 / 0.255139 (0.085687) | 0.377167 / 0.283200 (0.093967) | 0.088632 / 0.141683 (-0.053051) | 1.464500 / 1.452155 (0.012345) | 1.556686 / 1.492716 (0.063969) |\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) | 0.231136 / 0.018006 (0.213130) | 0.402687 / 0.000490 (0.402197) | 0.000590 / 0.000200 (0.000390) | 0.000059 / 0.000054 (0.000004) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.024926 / 0.037411 (-0.012485) | 0.101062 / 0.014526 (0.086536) | 0.106481 / 0.176557 (-0.070075) | 0.159167 / 0.737135 (-0.577968) | 0.110948 / 0.296338 (-0.185390) |\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) | 0.441813 / 0.215209 (0.226603) | 4.416332 / 2.077655 (2.338677) | 2.080621 / 1.504120 (0.576501) | 1.877832 / 1.541195 (0.336637) | 1.944778 / 1.468490 (0.476288) | 0.704634 / 4.584777 (-3.880143) | 3.433955 / 3.745712 (-0.311758) | 1.863493 / 5.269862 (-3.406368) | 1.168869 / 4.565676 (-3.396807) | 0.084095 / 0.424275 (-0.340180) | 0.012440 / 0.007607 (0.004833) | 0.545122 / 0.226044 (0.319077) | 5.472214 / 2.268929 (3.203285) | 2.514580 / 55.444624 (-52.930044) | 2.164570 / 6.876477 (-4.711907) | 2.193467 / 2.142072 (0.051395) | 0.809056 / 4.805227 (-3.996171) | 0.152343 / 6.500664 (-6.348321) | 0.067610 / 0.075469 (-0.007859) |\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) | 1.280968 / 1.841788 (-0.560820) | 13.887674 / 8.074308 (5.813366) | 13.160405 / 10.191392 (2.969013) | 0.128601 / 0.680424 (-0.551823) | 0.016420 / 0.534201 (-0.517780) | 0.382810 / 0.579283 (-0.196473) | 0.394386 / 0.434364 (-0.039978) | 0.470254 / 0.540337 (-0.070083) | 0.566907 / 1.386936 (-0.820029) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#8cc6950322337ea8873939541c53858b10c0f3b9 \"CML watermark\")\n", "_The documentation is not available anymore as the PR was closed or merged._", "<details>\n<summary>Show benchmarks</summary>\n\nPyArrow==6.0.0\n\n<details>\n<summary>Show updated benchmarks!</summary>\n\n### Benchmark: benchmark_array_xd.json\n\n| metric | read_batch_formatted_as_numpy after write_array2d | read_batch_formatted_as_numpy after write_flattened_sequence | read_batch_formatted_as_numpy after write_nested_sequence | read_batch_unformated after write_array2d | read_batch_unformated after write_flattened_sequence | read_batch_unformated after write_nested_sequence | read_col_formatted_as_numpy after write_array2d | read_col_formatted_as_numpy after write_flattened_sequence | read_col_formatted_as_numpy after write_nested_sequence | read_col_unformated after write_array2d | read_col_unformated after write_flattened_sequence | read_col_unformated after write_nested_sequence | read_formatted_as_numpy after write_array2d | read_formatted_as_numpy after write_flattened_sequence | read_formatted_as_numpy after write_nested_sequence | read_unformated after write_array2d | read_unformated after write_flattened_sequence | read_unformated after write_nested_sequence | write_array2d | write_flattened_sequence | write_nested_sequence |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.008673 / 0.011353 (-0.002679) | 0.004475 / 0.011008 (-0.006533) | 0.102060 / 0.038508 (0.063552) | 0.029438 / 0.023109 (0.006329) | 0.351785 / 0.275898 (0.075887) | 0.388199 / 0.323480 (0.064719) | 0.007011 / 0.007986 (-0.000974) | 0.003317 / 0.004328 (-0.001012) | 0.080931 / 0.004250 (0.076681) | 0.033449 / 0.037052 (-0.003603) | 0.360329 / 0.258489 (0.101840) | 0.400069 / 0.293841 (0.106228) | 0.033628 / 0.128546 (-0.094918) | 0.011462 / 0.075646 (-0.064184) | 0.323781 / 0.419271 (-0.095490) | 0.040686 / 0.043533 (-0.002847) | 0.332715 / 0.255139 (0.077576) | 0.370339 / 0.283200 (0.087139) | 0.084633 / 0.141683 (-0.057050) | 1.459452 / 1.452155 (0.007297) | 1.547719 / 1.492716 (0.055003) |\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) | 0.187051 / 0.018006 (0.169045) | 0.402625 / 0.000490 (0.402135) | 0.002218 / 0.000200 (0.002018) | 0.000070 / 0.000054 (0.000016) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.025240 / 0.037411 (-0.012171) | 0.102201 / 0.014526 (0.087675) | 0.108629 / 0.176557 (-0.067927) | 0.156686 / 0.737135 (-0.580449) | 0.111383 / 0.296338 (-0.184955) |\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) | 0.418099 / 0.215209 (0.202890) | 4.163345 / 2.077655 (2.085690) | 1.868419 / 1.504120 (0.364300) | 1.662066 / 1.541195 (0.120871) | 1.705912 / 1.468490 (0.237422) | 0.696391 / 4.584777 (-3.888386) | 3.338307 / 3.745712 (-0.407405) | 1.923255 / 5.269862 (-3.346607) | 1.249220 / 4.565676 (-3.316457) | 0.082037 / 0.424275 (-0.342238) | 0.012232 / 0.007607 (0.004624) | 0.523913 / 0.226044 (0.297869) | 5.290036 / 2.268929 (3.021107) | 2.319729 / 55.444624 (-53.124896) | 1.987345 / 6.876477 (-4.889132) | 2.044516 / 2.142072 (-0.097556) | 0.812098 / 4.805227 (-3.993129) | 0.147327 / 6.500664 (-6.353337) | 0.063838 / 0.075469 (-0.011631) |\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) | 1.219652 / 1.841788 (-0.622136) | 13.271513 / 8.074308 (5.197205) | 13.799982 / 10.191392 (3.608590) | 0.150055 / 0.680424 (-0.530369) | 0.028804 / 0.534201 (-0.505397) | 0.395452 / 0.579283 (-0.183831) | 0.398758 / 0.434364 (-0.035606) | 0.468575 / 0.540337 (-0.071763) | 0.553324 / 1.386936 (-0.833612) |\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) | 0.006498 / 0.011353 (-0.004855) | 0.004439 / 0.011008 (-0.006569) | 0.076525 / 0.038508 (0.038017) | 0.027184 / 0.023109 (0.004074) | 0.364705 / 0.275898 (0.088807) | 0.409481 / 0.323480 (0.086001) | 0.004831 / 0.007986 (-0.003154) | 0.004524 / 0.004328 (0.000196) | 0.075403 / 0.004250 (0.071153) | 0.039013 / 0.037052 (0.001960) | 0.364042 / 0.258489 (0.105553) | 0.413090 / 0.293841 (0.119249) | 0.032052 / 0.128546 (-0.096495) | 0.011514 / 0.075646 (-0.064132) | 0.085219 / 0.419271 (-0.334053) | 0.041448 / 0.043533 (-0.002085) | 0.350371 / 0.255139 (0.095232) | 0.386670 / 0.283200 (0.103470) | 0.089824 / 0.141683 (-0.051859) | 1.487392 / 1.452155 (0.035238) | 1.537201 / 1.492716 (0.044485) |\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) | 0.231555 / 0.018006 (0.213549) | 0.407505 / 0.000490 (0.407016) | 0.000382 / 0.000200 (0.000182) | 0.000060 / 0.000054 (0.000006) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.026665 / 0.037411 (-0.010747) | 0.105852 / 0.014526 (0.091326) | 0.108228 / 0.176557 (-0.068328) | 0.164164 / 0.737135 (-0.572972) | 0.114284 / 0.296338 (-0.182054) |\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) | 0.448957 / 0.215209 (0.233748) | 4.500058 / 2.077655 (2.422403) | 2.331660 / 1.504120 (0.827541) | 2.119904 / 1.541195 (0.578710) | 2.101489 / 1.468490 (0.632999) | 0.696580 / 4.584777 (-3.888197) | 3.364206 / 3.745712 (-0.381506) | 2.550157 / 5.269862 (-2.719704) | 1.496455 / 4.565676 (-3.069222) | 0.083289 / 0.424275 (-0.340986) | 0.012283 / 0.007607 (0.004676) | 0.555581 / 0.226044 (0.329537) | 5.556284 / 2.268929 (3.287355) | 2.595261 / 55.444624 (-52.849363) | 2.234793 / 6.876477 (-4.641683) | 2.280150 / 2.142072 (0.138078) | 0.817885 / 4.805227 (-3.987343) | 0.151481 / 6.500664 (-6.349183) | 0.066764 / 0.075469 (-0.008705) |\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) | 1.318875 / 1.841788 (-0.522913) | 14.220380 / 8.074308 (6.146072) | 13.922773 / 10.191392 (3.731381) | 0.154608 / 0.680424 (-0.525816) | 0.016343 / 0.534201 (-0.517858) | 0.380758 / 0.579283 (-0.198525) | 0.392595 / 0.434364 (-0.041769) | 0.468844 / 0.540337 (-0.071493) | 0.561047 / 1.386936 (-0.825889) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#d57fdcf2c8110b4b599289695fa065d1fc4936d4 \"CML watermark\")\n" ]

[]

[]

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

[ "I was not able to reproduce this behavior (I tried without using pytorch lightning though, since I don't know what code you ran in pytorch lightning to get this).\r\n\r\nIf you can provide a MWE that would be perfect ! :)", "Hi, I think the issue happened because I was loading datasets under an `if` ... `else` statement and the condition would change the dataset I would need to load but instead the cached one was always returned. However, I believe that is expected behaviour, if so I'll close the issue.\r\n\r\nOtherwise I will try to provide a MWE" ]

[]

[ "Thanks for reviewing and merging!" ]

[]

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

[]

[ "Hi ! I'm sorry to hear that.\r\nThough we are not redistributing the dataset, we just provide a python script that downloads and process the dataset from its original source hosted at https://www.cl.cam.ac.uk\r\n\r\nTherefore I'm not sure what's the issue with licensing. What do you mean exactly ?", "I think that the main issue is that the licesenses of the data are not made clear in the huggingface hub – other people wrongly assumed that the data was license-free, which resulted in commercial use, which is against the licenses.\r\nIs it possible to add the licenses from the original download to huggingface? that would help clear any confusion (licenses can be found here: https://www.cl.cam.ac.uk/research/nl/bea2019st/data/wi+locness_v2.1.bea19.tar.gz)", "Thanks for the clarification @SimonHFL \r\nYou're completely right, we need to show the licenses.\r\nI just added them here: https://huggingface.co/datasets/wi_locness#licensing-information", "Hi guys, I'm one of the authors of this dataset. \r\n\r\nTo clarify, we're happy for you to keep the data in the repo on 2 conditions:\r\n1. You don't host the data yourself.\r\n2. You make it clear that anyone who downloads the data via HuggingFace should read and abide by the license. \r\n\r\nI think you've now met these conditions, so we're all good, but I just wanted to make it clear in case there are any issues in the future. Thanks again to @aseifert for bringing this to our attention! :)", "Thanks for your message @chrisjbryant :)\r\nI'm closing this PR then.\r\n\r\nAnd thanks for reporting @aseifert" ]

[ { "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" } ]

[ "Looks like there are multiple issues regarding this (#2386, #2322) and it's a WIP #2329. Currently these datasets are being loaded in-memory which is causing this issue. Quoting @mariosasko here for a quick fix:\r\n\r\n> set `keep_in_memory` to `False` when loading a dataset (`sst = load_dataset(\"sst\", keep_in_memory=False)`) to prevent it from loading in-memory. Currently, in-memory datasets fail to find cached files due to this check (always False for them)\r\n\r\n", "Hi ! Since `datasets` 1.6.0 we no longer keep small datasets (<250MB) on disk and load them in RAM instead by default. This makes data processing and iterating on data faster. However datasets in RAM currently have no way to reload previous results from the cache (since nothing is written on disk). We are working on making the caching work for datasets in RAM.\r\n\r\nUntil then, I'd recommend passing `keep_in_memory=False` to the calls to `load_dataset` like here:\r\n\r\nhttps://github.com/huggingface/transformers/blob/223943872e8c9c3fc11db3c6e93da07f5177423f/examples/pytorch/language-modeling/run_clm.py#L233\r\n\r\nThis way you say explicitly that you want your dataset to stay on the disk, and it will be able to recover previously computed results from the cache.", "gotcha! thanks Quentin", "OK, It doesn't look like we can use the proposed workaround - see https://github.com/huggingface/transformers/issues/11801\r\n\r\nCould you please add an env var for us to be able to turn off this unwanted in our situation behavior? It is really problematic for dev work, when one needs to restart the training very often and needs a quick startup time. Manual editing of standard scripts is not a practical option when one uses examples.\r\n\r\nThis could also be a problem for tests, which will be slower because of lack of cache, albeit usually we use tiny datasets there. I think we want caching for tests.\r\n\r\nThank you.", "Hi @stas00, \r\n\r\nYou are right: an env variable is needed to turn off this behavior. I am adding it.\r\n\r\nFor the moment there is a config parameter to turn off this behavior: `datasets.config.MAX_IN_MEMORY_DATASET_SIZE_IN_BYTES = None`\r\n\r\nYou can find this info in the docs:\r\n- in the docstring of the parameter `keep_in_memory` of the function [`load_datasets`](https://huggingface.co/docs/datasets/package_reference/loading_methods.html#datasets.load_dataset):\r\n- in a Note in the docs about [Loading a Dataset](https://huggingface.co/docs/datasets/loading_datasets.html#from-the-huggingface-hub)\r\n\r\n> The default in 🤗Datasets is to memory-map the dataset on drive if its size is larger than datasets.config.MAX_IN_MEMORY_DATASET_SIZE_IN_BYTES (default 250 MiB); otherwise, the dataset is copied in-memory. This behavior can be disabled by setting datasets.config.MAX_IN_MEMORY_DATASET_SIZE_IN_BYTES = None, and in this case the dataset is not loaded in memory.", "Yes, but this still requires one to edit the standard example scripts, so if I'm doing that already I just as well can add `keep_in_memory=False`.\r\n\r\nMay be the low hanging fruit is to add `MAX_IN_MEMORY_DATASET_SIZE_IN_BYTES` env var to match the config, and if the user sets it to 0, then it'll be the same as `keep_in_memory=False` or `datasets.config.MAX_IN_MEMORY_DATASET_SIZE_IN_BYTES=0`?", "@stas00, however, for the moment, setting the value to `0` is equivalent to the opposite, i.e. `keep_in_memory=True`. This means the max size until which I load in memory is 0 bytes.\r\n\r\nTell me if this is logical/convenient, or I should change it.", "In my PR, to turn off current default bahavior, you should set env variable to one of: `{\"\", \"OFF\", \"NO\", \"FALSE\"}`.\r\n\r\nFor example:\r\n```\r\nMAX_IN_MEMORY_DATASET_SIZE_IN_BYTES=\r\n```", "IMHO, this behaviour is not very intuitive, as 0 is a normal quantity of bytes. So `MAX_IN_MEMORY_DATASET_SIZE_IN_BYTES=0` to me reads as don't cache ever.\r\n\r\nAlso \"SIZE_IN_BYTES\" that can take one of `{\"\", \"OFF\", \"NO\", \"FALSE\"}` is also quite odd.\r\n\r\nI think supporting a very simple `MAX_IN_MEMORY_DATASET_SIZE_IN_BYTES` that can accept any numerical value to match the name of the variable, requires minimal logic and is very straightforward. \r\n\r\nSo if you could adjust this logic - then `MAX_IN_MEMORY_DATASET_SIZE_IN_BYTES=0` is all that's needed to not do in-memory datasets.\r\n\r\nDoes it make sense?", "I understand your point @stas00, as I am not very convinced with current implementation.\r\n\r\nMy concern is: which numerical value should then pass a user who wants `keep_in_memory=True` by default, independently of dataset size? Currently it is `0` for this case.", "That's a good question, and again the normal bytes can be used for that:\r\n```\r\nMAX_IN_MEMORY_DATASET_SIZE_IN_BYTES=1e12 # (~2**40)\r\n```\r\nSince it's unlikely that anybody will have more than 1TB RAM.\r\n\r\nIt's also silly that it uses BYTES and not MBYTES - that level of refinement doesn't seem to be of a practical use in this context.\r\n\r\nNot sure when it was added and if there are back-compat issues here, but perhaps it could be renamed `MAX_IN_MEMORY_DATASET_SIZE` and support 1M, 1G, 1T, etc. \r\n\r\nBut scientific notation is quite intuitive too, as each 000 zeros is the next M, G, T multiplier. Minus the discrepancy of 1024 vs 1000, which adds up. And it is easy to write down `1e12`, as compared to `1099511627776` (2**40). (`1.1e12` is more exact).\r\n", "Great! Thanks, @stas00.\r\n\r\nI am implementing your suggestion to turn off default value when set to `0`.\r\n\r\nFor the other suggestion (allowing different metric prefixes), I will discuss with @lhoestq to agree on its implementation.", "Awesome! Thank you, @albertvillanova!!!\r\n\r\n" ]

[]

[]

[]

[]

[]

[ "Description: \"A clean data resource containing all explicit time references in a dataset of 52,183 novels whose full text is available via Project Gutenberg and the Hathi Trust Digital Library 2.\" > This is just the Gutenberg part.\r\n\r\nAlso, the paragraph at the top of the file would make a good Dataset Summary in the README :) " ]

[]

[]

[]

[]

[]

[]

[]

[ "_The documentation is not available anymore as the PR was closed or merged._", "@albertvillanova, Can you please review this PR for Issue #4540 ", "@lhoestq Thank you for merging the PR . Is there a slack channel for contributing to the datasets library. I would love to work on the library and make meaningful contributions.", "Hi ! Sure feel free to join our discord ^^ \r\nhttps://discuss.huggingface.co/t/join-the-hugging-face-discord/11263 so that we can discuss together mor eeasily. Otherwise everything happens on github ;)" ]

[]

[ "I close this PR, since it based on 1 week old repo. And I will create a new one" ]

[]

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

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

[ "Thanks for reporting these errors :)\r\n\r\nI can actually see two issues here.\r\n\r\nFirst, datasets like `natural_questions` require apache_beam to be processed. Right now the import is not at the right place so we have this error message. However, even the imports are fixed, the nlp viewer doesn't actually have the resources to process NQ right now so we'll have to wait until we have a version that we've already processed on our google storage (that's what we've done for wikipedia for example).\r\n\r\nSecond, datasets like `newsroom` require manual downloads as we're not allowed to redistribute the data ourselves (if I'm not wrong). An error message should be displayed saying that we're not allowed to show the dataset.\r\n\r\nI can fix the first issue with the imports but for the second one I think we'll have to see with @srush to show a message for datasets that require manual downloads (it can be checked whether a dataset requires manual downloads if `dataset_builder_instance.manual_download_instructions is not None`).\r\n\r\n", "I added apache-beam to the viewer. We can think about how to add newsroom. ", "We don't plan to host the source files of newsroom ourselves for now.\r\nYou can still get the dataset if you follow the download instructions given by `dataset = load_dataset('newsroom')` though.\r\nThe viewer also shows the instructions now.\r\n\r\nClosing this one. If you have other questions, feel free to re-open :)" ]

[]

[ "Hi !\r\nIt works on my side with both the LongFormerTokenizer and the LongFormerTokenizerFast.\r\n\r\nWhich version of transformers/datasets are you using ?", "transformers and datasets are both the latest", "Then I guess you need to give us more informations on your setup (OS, python, GPU, etc) or a Google Colab reproducing the error for us to be able to debug this error.", "And your version of `dill` if possible :)", "I have the same issue with `transformers/BertJapaneseTokenizer`.\r\n\r\n\r\n\r\n```python\r\n# train_ds = Dataset(features: {\r\n# 'title': Value(dtype='string', id=None), \r\n# 'score': Value(dtype='float64', id=None)\r\n# }, num_rows: 99999)\r\n\r\nt = BertJapaneseTokenizer.from_pretrained('bert-base-japanese-whole-word-masking')\r\nencoded = train_ds.map(lambda examples: {'tokens': t.encode(examples['title'])}, batched=True)\r\n```\r\n\r\n<details><summary>Error Message</summary>\r\n\r\n```\r\n---------------------------------------------------------------------------\r\nTypeError Traceback (most recent call last)\r\n<ipython-input-35-2b7d66b291c1> in <module>\r\n 2 \r\n 3 encoded = train_ds.map(lambda examples:\r\n----> 4 {'tokens': t.encode(examples['title'])}, batched=True)\r\n\r\n/usr/local/lib/python3.6/site-packages/datasets/arrow_dataset.py in map(self, function, with_indices, input_columns, batched, batch_size, drop_last_batch, remove_columns, keep_in_memory, load_from_cache_file, cache_file_name, writer_batch_size, features, disable_nullable, fn_kwargs, num_proc, suffix_template, new_fingerprint)\r\n 1242 fn_kwargs=fn_kwargs,\r\n 1243 new_fingerprint=new_fingerprint,\r\n-> 1244 update_data=update_data,\r\n 1245 )\r\n 1246 else:\r\n\r\n/usr/local/lib/python3.6/site-packages/datasets/arrow_dataset.py in wrapper(*args, **kwargs)\r\n 151 \"output_all_columns\": self._output_all_columns,\r\n 152 }\r\n--> 153 out: Union[\"Dataset\", \"DatasetDict\"] = func(self, *args, **kwargs)\r\n 154 if new_format[\"columns\"] is not None:\r\n 155 new_format[\"columns\"] = list(set(new_format[\"columns\"]) & set(out.column_names))\r\n\r\n/usr/local/lib/python3.6/site-packages/datasets/fingerprint.py in wrapper(*args, **kwargs)\r\n 156 kwargs_for_fingerprint[\"fingerprint_name\"] = fingerprint_name\r\n 157 kwargs[fingerprint_name] = update_fingerprint(\r\n--> 158 self._fingerprint, transform, kwargs_for_fingerprint\r\n 159 )\r\n 160 \r\n\r\n/usr/local/lib/python3.6/site-packages/datasets/fingerprint.py in update_fingerprint(fingerprint, transform, transform_args)\r\n 103 for key in sorted(transform_args):\r\n 104 hasher.update(key)\r\n--> 105 hasher.update(transform_args[key])\r\n 106 return hasher.hexdigest()\r\n 107 \r\n\r\n/usr/local/lib/python3.6/site-packages/datasets/fingerprint.py in update(self, value)\r\n 55 def update(self, value):\r\n 56 self.m.update(f\"=={type(value)}==\".encode(\"utf8\"))\r\n---> 57 self.m.update(self.hash(value).encode(\"utf-8\"))\r\n 58 \r\n 59 def hexdigest(self):\r\n\r\n/usr/local/lib/python3.6/site-packages/datasets/fingerprint.py in hash(cls, value)\r\n 51 return cls.dispatch[type(value)](cls, value)\r\n 52 else:\r\n---> 53 return cls.hash_default(value)\r\n 54 \r\n 55 def update(self, value):\r\n\r\n/usr/local/lib/python3.6/site-packages/datasets/fingerprint.py in hash_default(cls, value)\r\n 44 @classmethod\r\n 45 def hash_default(cls, value):\r\n---> 46 return cls.hash_bytes(dumps(value))\r\n 47 \r\n 48 @classmethod\r\n\r\n/usr/local/lib/python3.6/site-packages/datasets/utils/py_utils.py in dumps(obj)\r\n 365 file = StringIO()\r\n 366 with _no_cache_fields(obj):\r\n--> 367 dump(obj, file)\r\n 368 return file.getvalue()\r\n 369 \r\n\r\n/usr/local/lib/python3.6/site-packages/datasets/utils/py_utils.py in dump(obj, file)\r\n 337 def dump(obj, file):\r\n 338 \"\"\"pickle an object to a file\"\"\"\r\n--> 339 Pickler(file, recurse=True).dump(obj)\r\n 340 return\r\n 341 \r\n\r\n/usr/local/lib/python3.6/site-packages/dill/_dill.py in dump(self, obj)\r\n 444 raise PicklingError(msg)\r\n 445 else:\r\n--> 446 StockPickler.dump(self, obj)\r\n 447 stack.clear() # clear record of 'recursion-sensitive' pickled objects\r\n 448 return\r\n\r\n/usr/local/lib/python3.6/pickle.py in dump(self, obj)\r\n 407 if self.proto >= 4:\r\n 408 self.framer.start_framing()\r\n--> 409 self.save(obj)\r\n 410 self.write(STOP)\r\n 411 self.framer.end_framing()\r\n\r\n/usr/local/lib/python3.6/pickle.py in save(self, obj, save_persistent_id)\r\n 474 f = self.dispatch.get(t)\r\n 475 if f is not None:\r\n--> 476 f(self, obj) # Call unbound method with explicit self\r\n 477 return\r\n 478 \r\n\r\n/usr/local/lib/python3.6/site-packages/dill/_dill.py in save_function(pickler, obj)\r\n 1436 globs, obj.__name__,\r\n 1437 obj.__defaults__, obj.__closure__,\r\n-> 1438 obj.__dict__, fkwdefaults), obj=obj)\r\n 1439 else:\r\n 1440 _super = ('super' in getattr(obj.func_code,'co_names',())) and (_byref is not None) and getattr(pickler, '_recurse', False)\r\n\r\n/usr/local/lib/python3.6/pickle.py in save_reduce(self, func, args, state, listitems, dictitems, obj)\r\n 608 else:\r\n 609 save(func)\r\n--> 610 save(args)\r\n 611 write(REDUCE)\r\n 612 \r\n\r\n/usr/local/lib/python3.6/pickle.py in save(self, obj, save_persistent_id)\r\n 474 f = self.dispatch.get(t)\r\n 475 if f is not None:\r\n--> 476 f(self, obj) # Call unbound method with explicit self\r\n 477 return\r\n 478 \r\n\r\n/usr/local/lib/python3.6/pickle.py in save_tuple(self, obj)\r\n 749 write(MARK)\r\n 750 for element in obj:\r\n--> 751 save(element)\r\n 752 \r\n 753 if id(obj) in memo:\r\n\r\n/usr/local/lib/python3.6/pickle.py in save(self, obj, save_persistent_id)\r\n 474 f = self.dispatch.get(t)\r\n 475 if f is not None:\r\n--> 476 f(self, obj) # Call unbound method with explicit self\r\n 477 return\r\n 478 \r\n\r\n/usr/local/lib/python3.6/site-packages/dill/_dill.py in save_module_dict(pickler, obj)\r\n 931 # we only care about session the first pass thru\r\n 932 pickler._session = False\r\n--> 933 StockPickler.save_dict(pickler, obj)\r\n 934 log.info(\"# D2\")\r\n 935 return\r\n\r\n/usr/local/lib/python3.6/pickle.py in save_dict(self, obj)\r\n 819 \r\n 820 self.memoize(obj)\r\n--> 821 self._batch_setitems(obj.items())\r\n 822 \r\n 823 dispatch[dict] = save_dict\r\n\r\n/usr/local/lib/python3.6/pickle.py in _batch_setitems(self, items)\r\n 850 k, v = tmp[0]\r\n 851 save(k)\r\n--> 852 save(v)\r\n 853 write(SETITEM)\r\n 854 # else tmp is empty, and we're done\r\n\r\n/usr/local/lib/python3.6/pickle.py in save(self, obj, save_persistent_id)\r\n 519 \r\n 520 # Save the reduce() output and finally memoize the object\r\n--> 521 self.save_reduce(obj=obj, *rv)\r\n 522 \r\n 523 def persistent_id(self, obj):\r\n\r\n/usr/local/lib/python3.6/pickle.py in save_reduce(self, func, args, state, listitems, dictitems, obj)\r\n 632 \r\n 633 if state is not None:\r\n--> 634 save(state)\r\n 635 write(BUILD)\r\n 636 \r\n\r\n/usr/local/lib/python3.6/pickle.py in save(self, obj, save_persistent_id)\r\n 474 f = self.dispatch.get(t)\r\n 475 if f is not None:\r\n--> 476 f(self, obj) # Call unbound method with explicit self\r\n 477 return\r\n 478 \r\n\r\n/usr/local/lib/python3.6/site-packages/dill/_dill.py in save_module_dict(pickler, obj)\r\n 931 # we only care about session the first pass thru\r\n 932 pickler._session = False\r\n--> 933 StockPickler.save_dict(pickler, obj)\r\n 934 log.info(\"# D2\")\r\n 935 return\r\n\r\n/usr/local/lib/python3.6/pickle.py in save_dict(self, obj)\r\n 819 \r\n 820 self.memoize(obj)\r\n--> 821 self._batch_setitems(obj.items())\r\n 822 \r\n 823 dispatch[dict] = save_dict\r\n\r\n/usr/local/lib/python3.6/pickle.py in _batch_setitems(self, items)\r\n 845 for k, v in tmp:\r\n 846 save(k)\r\n--> 847 save(v)\r\n 848 write(SETITEMS)\r\n 849 elif n:\r\n\r\n/usr/local/lib/python3.6/pickle.py in save(self, obj, save_persistent_id)\r\n 519 \r\n 520 # Save the reduce() output and finally memoize the object\r\n--> 521 self.save_reduce(obj=obj, *rv)\r\n 522 \r\n 523 def persistent_id(self, obj):\r\n\r\n/usr/local/lib/python3.6/pickle.py in save_reduce(self, func, args, state, listitems, dictitems, obj)\r\n 632 \r\n 633 if state is not None:\r\n--> 634 save(state)\r\n 635 write(BUILD)\r\n 636 \r\n\r\n/usr/local/lib/python3.6/pickle.py in save(self, obj, save_persistent_id)\r\n 474 f = self.dispatch.get(t)\r\n 475 if f is not None:\r\n--> 476 f(self, obj) # Call unbound method with explicit self\r\n 477 return\r\n 478 \r\n\r\n/usr/local/lib/python3.6/site-packages/dill/_dill.py in save_module_dict(pickler, obj)\r\n 931 # we only care about session the first pass thru\r\n 932 pickler._session = False\r\n--> 933 StockPickler.save_dict(pickler, obj)\r\n 934 log.info(\"# D2\")\r\n 935 return\r\n\r\n/usr/local/lib/python3.6/pickle.py in save_dict(self, obj)\r\n 819 \r\n 820 self.memoize(obj)\r\n--> 821 self._batch_setitems(obj.items())\r\n 822 \r\n 823 dispatch[dict] = save_dict\r\n\r\n/usr/local/lib/python3.6/pickle.py in _batch_setitems(self, items)\r\n 845 for k, v in tmp:\r\n 846 save(k)\r\n--> 847 save(v)\r\n 848 write(SETITEMS)\r\n 849 elif n:\r\n\r\n/usr/local/lib/python3.6/pickle.py in save(self, obj, save_persistent_id)\r\n 494 reduce = getattr(obj, \"__reduce_ex__\", None)\r\n 495 if reduce is not None:\r\n--> 496 rv = reduce(self.proto)\r\n 497 else:\r\n 498 reduce = getattr(obj, \"__reduce__\", None)\r\n\r\nTypeError: can't pickle Tagger objects\r\n```\r\n\r\n</details>\r\n\r\ntrainsformers: 2.10.0\r\ndatasets: 1.0.2\r\ndill: 0.3.2\r\npython: 3.6.8\r\n\r\nOS: ubuntu 16.04 (Docker Image) on [Deep Learning VM](https://console.cloud.google.com/marketplace/details/click-to-deploy-images/deeplearning) (GCP)\r\nGPU: Tesla P100 (CUDA 10)\r\n", "> I have the same issue with `transformers/BertJapaneseTokenizer`.\r\n\r\nIt looks like it this tokenizer is not supported unfortunately.\r\nThis is because `t.word_tokenizer.mecab` is a `fugashi.fugashi.GenericTagger` which is not compatible with pickle nor dill.\r\n\r\nWe need objects passes to `map` to be picklable for our caching system to work properly.\r\nHere it crashes because the caching system is not able to pickle the GenericTagger.\r\n\r\n\\> Maybe you can create an issue on [fugashi](https://github.com/polm/fugashi/issues) 's repo and ask to make `fugashi.fugashi.GenericTagger` compatible with pickle ?\r\n\r\nWhat you can do in the meantime is use a picklable wrapper of the tokenizer:\r\n\r\n\r\n```python\r\nfrom transformers import BertJapaneseTokenizer, MecabTokenizer\r\n\r\nclass PicklableTokenizer(BertJapaneseTokenizer):\r\n\r\n def __getstate__(self):\r\n state = dict(self.__dict__)\r\n state[\"do_lower_case\"] = self.word_tokenizer.do_lower_case\r\n state[\"never_split\"] = self.word_tokenizer.never_split \r\n del state[\"word_tokenizer\"]\r\n return state\r\n\r\n def __setstate__(self, state):\r\n do_lower_case = state.pop(\"do_lower_case\")\r\n never_split = state.pop(\"never_split\")\r\n self.__dict__ = state\r\n self.word_tokenizer = MecabTokenizer(\r\n do_lower_case=do_lower_case, never_split=never_split)\r\n )\r\n\r\nt = PicklableTokenizer.from_pretrained(\"cl-tohoku/bert-base-japanese-whole-word-masking\")\r\nencoded = train_ds.map(lambda examples: {'tokens': t.encode(examples['title'])}, batched=True) # it works\r\n```", "We can also update the `BertJapaneseTokenizer` in `transformers` as you just shown @lhoestq to make it compatible with pickle. It will be faster than asking on fugashi 's repo and good for the other users of `transformers` as well.\r\n\r\nI'm currently working on `transformers` I'll include it in the https://github.com/huggingface/transformers/pull/7141 PR and the next release of `transformers`.", "Thank you for the rapid and polite response!\r\n\r\n@lhoestq Thanks for the suggestion! I've passed the pickle phase, but another `ArrowInvalid` problem occored. I created another issue #687 .\r\n\r\n@thomwolf Wow, really fast work. I'm looking forward to the next release 🤗" ]

[ { "color": "cfd3d7", "default": true, "description": "This issue or pull request already exists", "id": 1935892865, "name": "duplicate", "node_id": "MDU6TGFiZWwxOTM1ODkyODY1", "url": "https://api.github.com/repos/huggingface/datasets/labels/duplicate" } ]

[ "Hi @cbarond there's already an open issue at https://github.com/dair-ai/emotion_dataset/issues/5, as the data seems to be missing now, so check that issue instead 👍🏻 ", "Thanks @cbarond for reporting and @alvarobartt for pointing to the issue we opened in the author's repo.\r\n\r\nIndeed, this issue was first raised in the \"emotion\" dataset Community tab: https://huggingface.co/datasets/emotion/discussions/3\r\n\r\nI'm closing this issue and leave the issue above for the subsequent updates.\r\n\r\nDuplicate of: https://huggingface.co/datasets/emotion/discussions/3", "try using \"SetFit/emotion\" instead", "> try using \"SetFit/emotion\" instead\r\n\r\nI' replaced \"emotion\" with \"SetFit/Emotion\", but the code is getting stuck at\r\n\r\n`emotions = load_dataset(\"SetFit/emotion\")`\r\n\r\nI pause execution using the debugger, and it takes me to filelock.py:226\r\n\r\n`with self._thread_lock:`\r\n\r\nDo you know a way to get past this issue?", "thanks @honeyimholm - worked for me", "> try using \"SetFit/emotion\" instead\r\n\r\nIt really helps a lot, thank you!", "The dataset loading script has been fixed: https://huggingface.co/datasets/emotion/discussions/4" ]

[]

[ "_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) | 0.006134 / 0.011353 (-0.005219) | 0.003816 / 0.011008 (-0.007193) | 0.098226 / 0.038508 (0.059718) | 0.036830 / 0.023109 (0.013721) | 0.314551 / 0.275898 (0.038653) | 0.372251 / 0.323480 (0.048771) | 0.004762 / 0.007986 (-0.003224) | 0.003041 / 0.004328 (-0.001287) | 0.077651 / 0.004250 (0.073401) | 0.052445 / 0.037052 (0.015393) | 0.324632 / 0.258489 (0.066143) | 0.365724 / 0.293841 (0.071883) | 0.028069 / 0.128546 (-0.100477) | 0.008444 / 0.075646 (-0.067203) | 0.312767 / 0.419271 (-0.106505) | 0.047773 / 0.043533 (0.004240) | 0.305317 / 0.255139 (0.050178) | 0.332007 / 0.283200 (0.048807) | 0.018985 / 0.141683 (-0.122698) | 1.538022 / 1.452155 (0.085868) | 1.575898 / 1.492716 (0.083182) |\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) | 0.204780 / 0.018006 (0.186774) | 0.428125 / 0.000490 (0.427635) | 0.003454 / 0.000200 (0.003254) | 0.000078 / 0.000054 (0.000024) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.025064 / 0.037411 (-0.012348) | 0.099419 / 0.014526 (0.084893) | 0.111068 / 0.176557 (-0.065489) | 0.169775 / 0.737135 (-0.567361) | 0.112067 / 0.296338 (-0.184271) |\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) | 0.429642 / 0.215209 (0.214433) | 4.275556 / 2.077655 (2.197901) | 1.914658 / 1.504120 (0.410539) | 1.706556 / 1.541195 (0.165361) | 1.754228 / 1.468490 (0.285738) | 0.563669 / 4.584777 (-4.021108) | 3.391501 / 3.745712 (-0.354211) | 1.791517 / 5.269862 (-3.478345) | 1.030704 / 4.565676 (-3.534973) | 0.070882 / 0.424275 (-0.353393) | 0.011351 / 0.007607 (0.003744) | 0.529438 / 0.226044 (0.303394) | 5.294316 / 2.268929 (3.025387) | 2.344653 / 55.444624 (-53.099972) | 1.997468 / 6.876477 (-4.879009) | 2.108932 / 2.142072 (-0.033140) | 0.676794 / 4.805227 (-4.128433) | 0.135058 / 6.500664 (-6.365607) | 0.065857 / 0.075469 (-0.009612) |\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) | 1.231864 / 1.841788 (-0.609924) | 13.986694 / 8.074308 (5.912386) | 13.306600 / 10.191392 (3.115208) | 0.145520 / 0.680424 (-0.534904) | 0.016717 / 0.534201 (-0.517484) | 0.366303 / 0.579283 (-0.212980) | 0.391637 / 0.434364 (-0.042727) | 0.425445 / 0.540337 (-0.114892) | 0.507719 / 1.386936 (-0.879217) |\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) | 0.006236 / 0.011353 (-0.005116) | 0.003766 / 0.011008 (-0.007242) | 0.076794 / 0.038508 (0.038286) | 0.037210 / 0.023109 (0.014101) | 0.378387 / 0.275898 (0.102489) | 0.425456 / 0.323480 (0.101977) | 0.004694 / 0.007986 (-0.003291) | 0.002921 / 0.004328 (-0.001407) | 0.076985 / 0.004250 (0.072735) | 0.052188 / 0.037052 (0.015136) | 0.394385 / 0.258489 (0.135896) | 0.432527 / 0.293841 (0.138686) | 0.029091 / 0.128546 (-0.099455) | 0.008364 / 0.075646 (-0.067282) | 0.082583 / 0.419271 (-0.336689) | 0.042928 / 0.043533 (-0.000605) | 0.375321 / 0.255139 (0.120182) | 0.391719 / 0.283200 (0.108519) | 0.019388 / 0.141683 (-0.122295) | 1.550644 / 1.452155 (0.098489) | 1.604882 / 1.492716 (0.112166) |\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) | 0.236859 / 0.018006 (0.218853) | 0.418528 / 0.000490 (0.418039) | 0.000388 / 0.000200 (0.000188) | 0.000059 / 0.000054 (0.000004) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.025548 / 0.037411 (-0.011863) | 0.100644 / 0.014526 (0.086118) | 0.109102 / 0.176557 (-0.067455) | 0.161694 / 0.737135 (-0.575441) | 0.112088 / 0.296338 (-0.184250) |\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) | 0.484128 / 0.215209 (0.268919) | 4.849952 / 2.077655 (2.772297) | 2.512769 / 1.504120 (1.008649) | 2.303295 / 1.541195 (0.762100) | 2.356699 / 1.468490 (0.888209) | 0.564181 / 4.584777 (-4.020596) | 3.421393 / 3.745712 (-0.324319) | 2.570875 / 5.269862 (-2.698987) | 1.474307 / 4.565676 (-3.091370) | 0.068035 / 0.424275 (-0.356240) | 0.011300 / 0.007607 (0.003693) | 0.587867 / 0.226044 (0.361823) | 5.862447 / 2.268929 (3.593519) | 3.004017 / 55.444624 (-52.440607) | 2.664989 / 6.876477 (-4.211488) | 2.740020 / 2.142072 (0.597948) | 0.680840 / 4.805227 (-4.124387) | 0.137001 / 6.500664 (-6.363663) | 0.068098 / 0.075469 (-0.007371) |\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) | 1.297362 / 1.841788 (-0.544426) | 14.207891 / 8.074308 (6.133583) | 14.087562 / 10.191392 (3.896170) | 0.149514 / 0.680424 (-0.530910) | 0.016566 / 0.534201 (-0.517635) | 0.367602 / 0.579283 (-0.211681) | 0.400692 / 0.434364 (-0.033671) | 0.432907 / 0.540337 (-0.107431) | 0.525924 / 1.386936 (-0.861012) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#1ec069feaaf6c28d4e4df76d344693b591a74c3f \"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) | 0.006223 / 0.011353 (-0.005130) | 0.003672 / 0.011008 (-0.007336) | 0.097451 / 0.038508 (0.058943) | 0.036243 / 0.023109 (0.013133) | 0.375650 / 0.275898 (0.099752) | 0.431652 / 0.323480 (0.108172) | 0.004758 / 0.007986 (-0.003227) | 0.002941 / 0.004328 (-0.001387) | 0.077383 / 0.004250 (0.073132) | 0.055342 / 0.037052 (0.018289) | 0.390335 / 0.258489 (0.131846) | 0.427867 / 0.293841 (0.134026) | 0.027619 / 0.128546 (-0.100927) | 0.008244 / 0.075646 (-0.067402) | 0.313499 / 0.419271 (-0.105773) | 0.054987 / 0.043533 (0.011454) | 0.394044 / 0.255139 (0.138905) | 0.398784 / 0.283200 (0.115584) | 0.026499 / 0.141683 (-0.115184) | 1.496907 / 1.452155 (0.044753) | 1.554465 / 1.492716 (0.061749) |\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) | 0.241197 / 0.018006 (0.223190) | 0.427856 / 0.000490 (0.427366) | 0.006264 / 0.000200 (0.006065) | 0.000218 / 0.000054 (0.000164) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.025550 / 0.037411 (-0.011862) | 0.104426 / 0.014526 (0.089901) | 0.110310 / 0.176557 (-0.066246) | 0.173813 / 0.737135 (-0.563322) | 0.112129 / 0.296338 (-0.184209) |\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) | 0.458806 / 0.215209 (0.243597) | 4.576351 / 2.077655 (2.498697) | 2.265670 / 1.504120 (0.761550) | 2.073230 / 1.541195 (0.532035) | 2.135283 / 1.468490 (0.666793) | 0.562506 / 4.584777 (-4.022271) | 3.375101 / 3.745712 (-0.370611) | 1.734393 / 5.269862 (-3.535469) | 1.026622 / 4.565676 (-3.539054) | 0.068144 / 0.424275 (-0.356131) | 0.011092 / 0.007607 (0.003485) | 0.562779 / 0.226044 (0.336734) | 5.608256 / 2.268929 (3.339328) | 2.706468 / 55.444624 (-52.738157) | 2.381607 / 6.876477 (-4.494869) | 2.451027 / 2.142072 (0.308954) | 0.671590 / 4.805227 (-4.133637) | 0.135749 / 6.500664 (-6.364915) | 0.065389 / 0.075469 (-0.010080) |\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) | 1.244806 / 1.841788 (-0.596981) | 14.042150 / 8.074308 (5.967841) | 14.246612 / 10.191392 (4.055220) | 0.134309 / 0.680424 (-0.546114) | 0.017082 / 0.534201 (-0.517119) | 0.366043 / 0.579283 (-0.213240) | 0.400748 / 0.434364 (-0.033616) | 0.425695 / 0.540337 (-0.114643) | 0.509355 / 1.386936 (-0.877581) |\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) | 0.006134 / 0.011353 (-0.005219) | 0.003980 / 0.011008 (-0.007028) | 0.078353 / 0.038508 (0.039845) | 0.038011 / 0.023109 (0.014902) | 0.375784 / 0.275898 (0.099886) | 0.433619 / 0.323480 (0.110139) | 0.004897 / 0.007986 (-0.003088) | 0.002981 / 0.004328 (-0.001347) | 0.077362 / 0.004250 (0.073112) | 0.056108 / 0.037052 (0.019056) | 0.395984 / 0.258489 (0.137495) | 0.427397 / 0.293841 (0.133556) | 0.029325 / 0.128546 (-0.099221) | 0.008498 / 0.075646 (-0.067148) | 0.082478 / 0.419271 (-0.336794) | 0.044085 / 0.043533 (0.000552) | 0.389923 / 0.255139 (0.134784) | 0.391180 / 0.283200 (0.107980) | 0.022452 / 0.141683 (-0.119231) | 1.507758 / 1.452155 (0.055603) | 1.530459 / 1.492716 (0.037743) |\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) | 0.230928 / 0.018006 (0.212922) | 0.408484 / 0.000490 (0.407995) | 0.000806 / 0.000200 (0.000606) | 0.000067 / 0.000054 (0.000012) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.025183 / 0.037411 (-0.012228) | 0.102292 / 0.014526 (0.087766) | 0.108142 / 0.176557 (-0.068415) | 0.161172 / 0.737135 (-0.575963) | 0.114476 / 0.296338 (-0.181862) |\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) | 0.482978 / 0.215209 (0.267769) | 4.816103 / 2.077655 (2.738448) | 2.505567 / 1.504120 (1.001447) | 2.302598 / 1.541195 (0.761404) | 2.371238 / 1.468490 (0.902748) | 0.567467 / 4.584777 (-4.017310) | 3.363407 / 3.745712 (-0.382306) | 1.746213 / 5.269862 (-3.523649) | 1.035468 / 4.565676 (-3.530208) | 0.068431 / 0.424275 (-0.355844) | 0.011069 / 0.007607 (0.003462) | 0.598241 / 0.226044 (0.372196) | 5.953927 / 2.268929 (3.684999) | 3.007493 / 55.444624 (-52.437132) | 2.629399 / 6.876477 (-4.247078) | 2.737201 / 2.142072 (0.595129) | 0.682456 / 4.805227 (-4.122771) | 0.137613 / 6.500664 (-6.363051) | 0.067941 / 0.075469 (-0.007528) |\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) | 1.306015 / 1.841788 (-0.535772) | 14.359240 / 8.074308 (6.284932) | 14.187601 / 10.191392 (3.996209) | 0.138612 / 0.680424 (-0.541812) | 0.016708 / 0.534201 (-0.517493) | 0.366365 / 0.579283 (-0.212918) | 0.396982 / 0.434364 (-0.037382) | 0.426939 / 0.540337 (-0.113398) | 0.520064 / 1.386936 (-0.866872) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#21d0fd041a5eca02d3ee787396216ac613c662ac \"CML watermark\")\n", "They use `token` and emit a deprecation warning if `use_auth_token` is passed instead (see https://github.com/huggingface/transformers/blob/78a2b19fc84ed55c65f4bf20a901edb7ceb73c5f/src/transformers/modeling_utils.py#L1933). \r\n\r\nI think we can update the `examples` scripts after merging this PR.", "> I think we can update the examples scripts after merging this PR.\r\n\r\nWe should do a release before updated in the examples scripts no ? That's why it's an option to not have a deprecation warning until transformers and co are updated with the `token` arg", "> We should do a release before updated in the examples scripts no ? That's why it's an option to not have a deprecation warning until transformers and co are updated with the token arg\r\n\r\nThis would avoid the warning only for the latest `datasets` release. TBH, I don't think this is worth the hassle, considering how simple it is to remove it.", "<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) | 0.007644 / 0.011353 (-0.003709) | 0.004667 / 0.011008 (-0.006341) | 0.117347 / 0.038508 (0.078839) | 0.050620 / 0.023109 (0.027510) | 0.415402 / 0.275898 (0.139504) | 0.485898 / 0.323480 (0.162418) | 0.005848 / 0.007986 (-0.002138) | 0.003736 / 0.004328 (-0.000592) | 0.089798 / 0.004250 (0.085547) | 0.069344 / 0.037052 (0.032292) | 0.441684 / 0.258489 (0.183195) | 0.468972 / 0.293841 (0.175131) | 0.036637 / 0.128546 (-0.091909) | 0.010219 / 0.075646 (-0.065427) | 0.394293 / 0.419271 (-0.024978) | 0.061462 / 0.043533 (0.017929) | 0.409448 / 0.255139 (0.154309) | 0.431557 / 0.283200 (0.148358) | 0.027795 / 0.141683 (-0.113888) | 1.837844 / 1.452155 (0.385690) | 1.862683 / 1.492716 (0.369967) |\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) | 0.230500 / 0.018006 (0.212494) | 0.483139 / 0.000490 (0.482649) | 0.006517 / 0.000200 (0.006317) | 0.000143 / 0.000054 (0.000088) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.033152 / 0.037411 (-0.004259) | 0.133673 / 0.014526 (0.119147) | 0.143853 / 0.176557 (-0.032704) | 0.215254 / 0.737135 (-0.521882) | 0.150676 / 0.296338 (-0.145662) |\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) | 0.503796 / 0.215209 (0.288587) | 5.049981 / 2.077655 (2.972326) | 2.399427 / 1.504120 (0.895307) | 2.167635 / 1.541195 (0.626441) | 2.257448 / 1.468490 (0.788958) | 0.641298 / 4.584777 (-3.943479) | 4.828676 / 3.745712 (1.082964) | 4.346069 / 5.269862 (-0.923793) | 2.103890 / 4.565676 (-2.461786) | 0.079115 / 0.424275 (-0.345160) | 0.013377 / 0.007607 (0.005770) | 0.621207 / 0.226044 (0.395162) | 6.190939 / 2.268929 (3.922011) | 2.920129 / 55.444624 (-52.524495) | 2.549225 / 6.876477 (-4.327252) | 2.719221 / 2.142072 (0.577149) | 0.790949 / 4.805227 (-4.014278) | 0.172032 / 6.500664 (-6.328632) | 0.077779 / 0.075469 (0.002310) |\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) | 1.432572 / 1.841788 (-0.409216) | 21.000031 / 8.074308 (12.925723) | 17.555093 / 10.191392 (7.363701) | 0.166646 / 0.680424 (-0.513778) | 0.020451 / 0.534201 (-0.513750) | 0.488767 / 0.579283 (-0.090516) | 0.737036 / 0.434364 (0.302672) | 0.621694 / 0.540337 (0.081356) | 0.732074 / 1.386936 (-0.654862) |\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) | 0.008198 / 0.011353 (-0.003155) | 0.004987 / 0.011008 (-0.006021) | 0.090714 / 0.038508 (0.052206) | 0.053379 / 0.023109 (0.030270) | 0.425199 / 0.275898 (0.149301) | 0.514036 / 0.323480 (0.190556) | 0.006043 / 0.007986 (-0.001943) | 0.003888 / 0.004328 (-0.000441) | 0.088294 / 0.004250 (0.084043) | 0.073024 / 0.037052 (0.035971) | 0.435983 / 0.258489 (0.177494) | 0.514293 / 0.293841 (0.220452) | 0.039451 / 0.128546 (-0.089095) | 0.010439 / 0.075646 (-0.065207) | 0.096885 / 0.419271 (-0.322387) | 0.060165 / 0.043533 (0.016632) | 0.421053 / 0.255139 (0.165914) | 0.455545 / 0.283200 (0.172345) | 0.027234 / 0.141683 (-0.114449) | 1.768975 / 1.452155 (0.316820) | 1.842853 / 1.492716 (0.350137) |\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) | 0.278940 / 0.018006 (0.260933) | 0.480709 / 0.000490 (0.480219) | 0.000436 / 0.000200 (0.000236) | 0.000070 / 0.000054 (0.000016) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.034900 / 0.037411 (-0.002511) | 0.144893 / 0.014526 (0.130368) | 0.149567 / 0.176557 (-0.026989) | 0.213200 / 0.737135 (-0.523935) | 0.156735 / 0.296338 (-0.139604) |\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) | 0.535897 / 0.215209 (0.320687) | 5.336998 / 2.077655 (3.259343) | 2.685854 / 1.504120 (1.181734) | 2.470177 / 1.541195 (0.928983) | 2.547495 / 1.468490 (1.079004) | 0.642830 / 4.584777 (-3.941947) | 4.595866 / 3.745712 (0.850154) | 2.186696 / 5.269862 (-3.083165) | 1.317969 / 4.565676 (-3.247708) | 0.079268 / 0.424275 (-0.345007) | 0.013792 / 0.007607 (0.006185) | 0.662236 / 0.226044 (0.436192) | 6.604775 / 2.268929 (4.335847) | 3.355888 / 55.444624 (-52.088736) | 2.968911 / 6.876477 (-3.907565) | 3.121862 / 2.142072 (0.979790) | 0.794752 / 4.805227 (-4.010475) | 0.170800 / 6.500664 (-6.329864) | 0.078393 / 0.075469 (0.002924) |\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) | 1.601605 / 1.841788 (-0.240183) | 20.743553 / 8.074308 (12.669245) | 17.543968 / 10.191392 (7.352576) | 0.221884 / 0.680424 (-0.458540) | 0.020779 / 0.534201 (-0.513422) | 0.479677 / 0.579283 (-0.099606) | 0.516207 / 0.434364 (0.081843) | 0.564046 / 0.540337 (0.023709) | 0.711336 / 1.386936 (-0.675600) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#819bb4346434912eb405ce3f3e9f21dc25a2fe85 \"CML watermark\")\n", "Yes, sounds great! Thanks", "yup" ]

[]

[]

[]

[ "I replaced all the \"we\" and applied your suggestion", "Merging this for now, we can continue improving this card in other PRs :)" ]

[ { "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" } ]

[ "I have to admit I'm not a fan of this idea, as this would result in a non-consistent behavior between tabular and non-tabular datasets, which is confusing if done without the context you provided. Instead, we could consider returning a `Dataset` object rather than `DatasetDict` if there is only one split in the generated dataset. But then again, I think this lib is a bit too old to make such changes. @lhoestq @albertvillanova WDYT?\r\n\r\n", "We can brainstorm here to see how we could make it happen ? And then depending on the options we see if it's a change we can do.\r\n\r\nI'm starting with a first reasoning\r\n\r\nCurrently not passing `split=` in `load_dataset` means \"return a dict with each split\".\r\n\r\nNow what would happen if a dataset has no split ? Ideally it should return one Dataset. And passing `split=` would have no sense. So depending on the dataset content, not passing `split=` should return a dict or a Dataset. In particular, those two cases should work:\r\n```python\r\n# case 1: dataset without split\r\nds = load_dataset(\"dataset_without_split\")\r\nds[0], ds[\"column_name\"], list(ds) # we want this\r\n\r\n# case 2: dataset with splits\r\nds = load_dataset(\"dataset_with_splits\")\r\nds[\"train\"] # this works and can't be changed\r\nds = load_dataset(\"dataset_with_splits\", split=\"train\")\r\nds[0], ds[\"column_name\"], list(ds) # this works and can't be changed\r\n```\r\n\r\nI can see several ideas:\r\n1. allowing `load_dataset` to return a different object based on the dataset content - either a Dataset or a DatasetDict\r\n - we can update `get_dataset_split_names` to return None or a list if users want to know in advance what object will be returned. They can also use `isinstance` _a posteriori_\r\n - but in this case we expect users to be careful when loading datasets and always to extra steps to check if they got a Dataset or DatasetDict\r\n2. merge Dataset and DatasetDict objects\r\n - they already share many functions: map, filter, push_to_hub etc.\r\n - we can define `ds[0]` to be the first item of the first split, and consider that the uses accesses rows from the full table of all the splits concatenated\r\n - however there is a collision when doing `ds[\"column_name\"]` or `ds[\"train\"]` that we need to address: the first returns a list, while the other returns a Dataset.\r\n\r\nWhat are your opinions on those two ideas ? Do you have other ideas in mind ?", "I like the first idea more (concatenating splits doesn't seem useful, no?). This is a significant breaking change, so I think we should do a poll (or something similar) to gather more info on the actual \"expected behavior\" and wait for Datasets 3.0 if we decide to implement it.\r\n\r\nPS: @thomwolf also suggested the same thing a while ago (https://github.com/huggingface/datasets/issues/743#issuecomment-746074641).", "I think it's an interesting improvement to the user experience for a case that comes often (no split) so I would definitively support it.\r\n\r\nI would be more in favor of option 2 rather than returning various types of objects from load_dataset and handling carefully the possible collisions indeed", "Related: if a dataset only has one split, we don't show the splits select control in the dataset viewer on the Hub, eg. compare https://huggingface.co/datasets/hf-internal-testing/fixtures_image_utils/viewer/image/test with https://huggingface.co/datasets/glue/viewer/mnli/test.\r\n\r\nSee https://github.com/huggingface/moon-landing/pull/3858 for more details (internal)", "I feel like the second idea is a bit more overkill. \r\n@severo I would say it's a bit irrelevant to the problem we have but is a separate problem @polinaeterna is solving at the moment. 😅 (also discussed on slack)", "OK, sorry for polluting the thread. The relation I saw with the dataset viewer is that from a UX point of view, we hide the concepts of split and configuration whenever possible -> this issue feels like doing the same in the datasets library.", "I would agree that returning different types based on the content of the dataset might be confusing.\r\n\r\nWe can do something similar to what `fetch_*` or `load_*` from `sklearn.datasets` do, which is to have an arg which changes the type of the returned type. For instance, `load_iris` would return a dict, but `load_iris(..., return_X_y=True)` would return a tuple.\r\n\r\nHere we can have a similar arg such as `return_X` which would then only return a single `DataSet` or an array.", "> I feel like the second idea is a bit more overkill.\r\n\r\nOverkill in what sense ?\r\n\r\n> Here we can have a similar arg such as return_X which would then only return a single DataSet or an array.\r\n\r\nRight now one can already pass `split=\"all\"` to get one `Dataset` object with all the data in it (unsplit). We could also have something like `return_all=True` so make the API clearer.\r\n\r\n> I would be more in favor of option 2 rather than returning various types of objects from load_dataset and handling carefully the possible collisions indeed\r\n\r\nI think it would be ok to handle the collision by allowing both `ds[\"train\"]` and `ds[\"column_name\"]` (and maybe adding something like `ds.splits` for those who want to iterate over the splits or add new ones)", "Would it make sense to remove the notion of \"split\" in `load_dataset`? I feel a lof of it comes from the want to have some sort of group of more or less similar dataset. \"train\"/\"test\"/\"validation\" are the traditional ones, but there are some datasets that have much more splits.\r\n\r\nWould it make sense to force `load_dataset` to only load a single `Dataset` object, and fail if it doesn't point to one. And have another method that's like `load_dataset_group_info` that can return a very arbitrary info class (Dict, List whatever), but you need to pass individual infos to `load_dataset` to run anything? Typically I don't think `DatasetDict.map` is really that helpful, but that's my personal opinion. This would help make things more readable (typically knowing if an object is a `Dataset` or a `DatasetDict`)", "> Would it make sense to remove the notion of \"split\" in load_dataset?\r\n\r\nI think we need to keep it - though in practice people can name the splits whatever they want anyway.\r\n\r\n> Would it make sense to force load_dataset to only load a single Dataset object, and fail if it doesn't point to one.\r\n\r\nWe need to keep backward compatibility ideally - in particular the load_dataset + ds[\"train\"] one", "> I think we need to keep it - though in practice people can name the splits whatever they want anyway.\r\n\r\nIt was my understanding that the whole issue was that `load_dataset` returned multiple types of objects.\r\n\r\n> We need to keep backward compatibility ideally - in particular the load_dataset + ds[\"train\"] one\r\n\r\nYeah sorry I meant ideally. One can always start developing `load_dataset_v2` can deprecate the first one and remove it in the longer term.", "> It was my understanding that the whole issue was that load_dataset returned multiple types of objects.\r\n\r\nYes indeed, but we still want to keep a way to load the train/val/test/whatever splits alone ;)", "@thomasw21's solution is good but it will break backwards compatibility. 😅", "Started to experiment with merging Dataset and DatasetDict. My plan is to define the splits of a Dataset in Dataset.info.splits (already exists, but never used). A Dataset would then be the concatenation of its splits if they exist.\r\n\r\nNot sure yet this is the way to go. My plan is to play with it and see and share it with you, so we can see if it makes sense from a UX point of view.", "So just to make sure that I understand the current direction, people will have to be extra careful when handling splits right?\r\nImagine \"potato\" a dataset containing train/validation split:\r\n```\r\nload_dataset(\"potato\") # returns the concatenation of all the splits\r\n```\r\nPreviously the design would force you to choose a split (it would raise otherwise), or manually concat them if you really wanted to play with concatenated splits. Now it would potentially run without raising for a bit of time until you figure out that you've been training on both train and validation split.\r\n\r\nWould it make sense to use a dataset specific default instead of using the concatenation, typically \"potato\" dataset's default would be train?\r\n```\r\nload_dataset(\"potato\") # returns \"train\" split\r\nload_dataset(\"potato\", split=\"train\") # returns \"train\" split\r\nload_dataset(\"potato\", split=\"validation\") # returns \"validation\" split\r\nconcatenate_datasets([load_dataset(\"potato\", split=\"train\"), load_dataset(\"potato\", split=\"validation\")]) # returns concatenation\r\n```", "> load_dataset(\"potato\") # returns \"train\" split\r\n\r\nTo avoid a breaking change we need to be able to do `load_dataset(\"potato\")[\"validation\"]` as well.\r\n\r\nIn that case I'd wonder where the validation split comes from, since the rows of the dataset wouldn't contain the validation split according to your example. That's why I'm more in favor of concatenating.\r\n\r\nA dataset is one table, that optionally has some split info about subsets (e.g. for training an evaluation)\r\n\r\nThis also allows anyone to re-split the dataset the way they want if they're not happy with the default:\r\n\r\n```python\r\nds = load_dataset(\"potato\").train_test_split(test_size=0.2)\r\ntrain_ds = ds[\"train\"]\r\ntest_ds = ds[\"test\"]\r\n```", "Just thinking about this, we could just have `to_dataframe()` as `load_dataset(\"blah\").to_dataframe()` to get the whole dataset, and not change anything else.", "I have a first implementation of option 2 (merging Dataset and DatasetDict) in this PR: https://github.com/huggingface/datasets/pull/5301/\r\n\r\nFeel free to play with it if you're interested, and let me know what you think. In this PR, a dataset is one table that optionally has some split info about subsets.", "@adrinjalali we already have [to_pandas](https://huggingface.co/docs/datasets/package_reference/main_classes#datasets.Dataset.to_pandas) AFAIK that essentially does the same thing (for a dataset, not for a dataset dict), I was wondering if it makes sense to have this as I don't know portion of people who load non-tabular datasets into dataframes. @lhoestq I saw your PR and it will break a lot of things imo, WDYT of this option? ", "> we already have [to_pandas](https://huggingface.co/docs/datasets/package_reference/main_classes#datasets.Dataset.to_pandas) AFAIK that essentially does the same thing (for a dataset, not for a dataset dict)\r\n\r\nyes correct :)\r\n\r\n> I saw your PR and it will break a lot of things imo\r\n\r\nDo you have concrete examples you can share ?\r\n\r\n> WDYT of this option?\r\n\r\nThe to_dataframe option ? I think it not enough, since you'd still get a `DatasetDict({\"train\": Dataset()})` if you load a dataset with no splits (e.g. one CSV), and this doesn't really make sense.\r\n\r\nNote that in the PR I opened you can do\r\n```python\r\nds = load_dataset(\"dataset_with_just_one_csv\") # Dataset type\r\ndf = load_dataset(\"dataset_with_just_one_csv\").to_pandas() # DataFrame type\r\n```", "@lhoestq no I think @adrinjalali and I meant when user calls `to_dataframe` if there's only train split in `DatasetDict` we could directly load that into dataframe. This might cause a confusion given there's to_pandas but I think it's more intuitive and least breaking change. (given people -who use `datasets` for tabular workflows- will eventually call `to_pandas` anyway) ", "So in that case it would be fine to still end up with a dataset dict with a \"train\" split ?", "yeah what I mean is this:\r\n\r\n```py\r\ndataset = load_dataset(\"blah\")\r\n\r\n# deal with a split of the dataset\r\ntrain = dataset[\"train\"]\r\ntrain_df = dataset[\"train\"].to_dataframe()\r\n\r\n# deal with the whole dataset\r\ndataset_df = dataset.to_dataframe()\r\n```\r\n\r\nSo we do two things to improve tabular experience:\r\n- allow datasets to have a single split\r\n- add `to_dataframe` to the root dict level so that users can simply call `df = load_dataset(\"blah\").to_dataframe()` and have it in their `pandas.DataFrame` object.", "Ok ! Note that we already have `Dataset.to_pandas()` so for consistency I'd call it `DatasetDict.to_pandas()` as well, does it sound good to you ? This is something we can add pretty easily", "yeah that sounds perfect @lhoestq !", "> So just to make sure that I understand the current direction, people will have to be extra careful when handling splits right?\r\n\r\nWe can raise an error if someone does `load_dataset(...)[0]` if the dataset is made of several splits, and return the first example if there's one or zero splits (i.e. when it's not ambiguous). Had this idea from the dicussions in #5312 WDYT @thomasw21 ?", "> We can raise an error if someone does load_dataset(...)[0] if the dataset is made of several splits,\r\n\r\nBut then how is that different to have the distinction between DatasetDict and Dataset then? Is it just that \"default behaviour when there are no splits or single split, it returns directly the split when there's no ambiguity\".\r\n\r\nAlso I was wondering how the concatenation could have heavy impacts when running mapping functions/filtering in batch? Typically can batch be somehow mixed?", "> But then how is that different to have the distinction between DatasetDict and Dataset then?\r\n\r\nBecause it doesn't make sense to be able to do `example = ds[0]` or `examples = list(ds)` on a class named `DatasetDict` of type `Dict[str, Dataset]`.\r\n\r\n> Also I was wondering how the concatenation could have heavy impacts when running mapping functions/filtering in batch? Typically can batch be somehow mixed?\r\n\r\nNo, we run each function on each split separated", "> Because it doesn't make sense to be able to do example = ds[0] or examples = list(ds) on a class named DatasetDict of type Dict[str, Dataset].\r\n\r\nHum but you're still going to raise an exception in both those cases with your current change no? (actually list(ds) would return the name of the splits no?)\r\n\r\n> No, we run each function on each split separated\r\n\r\nNice!" ]

[]

[ "Hi Julien ! This sounds related to https://github.com/huggingface/datasets/issues/5695 - TL;DR: you need to have shards smaller than 2GB to avoid this issue\r\n\r\nThe number of rows per shard is computed using an estimated size of the full dataset, which can sometimes lead to shards bigger than `max_shard_size`. The estimation is currently done using the first samples of the dataset (which can surely be improved). We should probably open an issue to fix this once and for all.\r\n\r\nAnyway for your specific dataset I'd suggest you to pass `num_shards` instead of `max_shard_size` for now, and make sure to have enough shards to end up with shards smaller than 2GB", "Hi Quentin! Thanks a lot! Using `num_shards` instead of `max_shard_size` works as expected.\r\n\r\nIndeed the way you describe how the size is computed cannot really work with the dataset I'm building as all the image doesn't have the same resolution and then size. Opening an issue on this might be a good idea." ]

[]

[ "Hi\r\nnot sure how change the lable after creation, but this is an issue not dataset request. thanks ", "Hi ! Currently you have to use `map` . You can see an example of how to do it in this comment: https://github.com/huggingface/datasets/issues/853#issuecomment-727872188\r\n\r\nIn the future we'll add support for a more native way of adding a new column ;)" ]

[]

[ "Thank you @lhoestq. This logging configuration makes more sense to me.\r\n\r\nOnce propagation is allowed, the end-user can customize logging behavior and add custom handlers to the proper top logger in the hierarchy.\r\n\r\nAnd I also agree with following the best practices and removing any custom handlers:\r\n- it is the end user who has to implement any custom handlers\r\n- indeed, the previous logging problem with TensorFlow was due to the fact that absl did not follow best practices and had implemented a custom handler\r\n\r\nOur errors/warnings will be displayed anyway, even if we do not implement any custom handler. Since Python 3.2, logging has a built-in \"default\" handler (logging.lastResort) with the expected default behavior (sending error/warning messages to sys.stderr), which is used only if the end user has not configured any custom handler." ]

[]

[]

[]

[ "It's reachable for the viewer and me, so I suppose it was down at that moment?" ]

[]

[ "Thanks for approving it!" ]

[ { "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" } ]

[ "You can add columns that you wish to set into `torch` format using `dataset.set_format(\"torch\", ['id', 'abc'])` so that input batch of the transform only contains those columns", "Sorry, I miss `output_all_columns` args and thought after `dataset.set_format(\"torch\", columns=columns)` I can only get specific columns I assigned." ]

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

[ "Hi! Very cool dataset! I answered your questions on the forum. Also, feel free to comment `#self-assign` on this issue to self-assign it." ]

[]

[ "thanks ! merging this one" ]

[]

[ "README.md is not created yet. I'll add it soon.", "Thank you for your detailed code review! It's so helpful.\r\nI'll reflect them to the code in 24 hours.\r\n\r\nYou may have told me in Slack (I cannot find the conversation log though I've looked through threads), but I'm sorry it seems I'm still misunderstanding how to get YAML from the tagger.\r\nI'm now asking on Slack if I am looking at the tagger the wrong way.", "One more thing I'd like to ask.\r\nShould I make changes by myself, or can I use the \"Commit suggestion\" feature?\r\nI'm new to this feature and I don't know how the rules work in this repository, so I'd like to ask just in case.", "Thank you very much for merging!" ]

[ { "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 @lhoestq \r\nMaybe this is clearer to explain like this, currently map function, map one example to \"one\" modified one, lets assume we want to map one example to \"multiple\" examples, in which we do not know in advance how many examples they would be per each entry. I greatly appreciate telling me how I can handle this operation, thanks a lot", "Hi,\r\nthis is also my question on how to perform similar operation as \"unbatch\" in tensorflow in great huggingface dataset library. \r\nthanks.", "Hi,\r\n\r\n`Dataset.map` in the batched mode allows you to map a single row to multiple rows. So to perform \"unbatch\", you can do the following:\r\n```python\r\nimport collections\r\n\r\ndef unbatch(batch):\r\n new_batch = collections.defaultdict(list)\r\n keys = batch.keys()\r\n for values in zip(*batch.values()):\r\n ex = {k: v for k, v in zip(keys, values)}\r\n inputs = f\"record query: {ex['query']} entities: {', '.join(ex['entities'])} passage: {ex['passage']}\"\r\n new_batch[\"inputs\"].extend([inputs] * len(ex[\"answers\"]))\r\n new_batch[\"targets\"].extend(ex[\"answers\"])\r\n return new_batch\r\n\r\ndset = dset.map(unbatch, batched=True, remove_columns=dset.column_names)\r\n```", "Dear @mariosasko \r\nFirst, thank you very much for coming back to me on this, I appreciate it a lot. I tried this solution, I am getting errors, do you mind\r\ngiving me one test example to be able to run your code, to understand better the format of the inputs to your function?\r\nin this function https://github.com/google-research/text-to-text-transfer-transformer/blob/3c58859b8fe72c2dbca6a43bc775aa510ba7e706/t5/data/preprocessors.py#L952 they copy each example to the number of \"answers\", do you mean one should not do the copying part and use directly your function? \r\n\r\n\r\nthank you very much for your help and time.", "Hi @mariosasko \r\nI think finally I got this, I think you mean to do things in one step, here is the full example for completeness:\r\n\r\n```\r\ndef unbatch(batch):\r\n new_batch = collections.defaultdict(list)\r\n keys = batch.keys()\r\n for values in zip(*batch.values()):\r\n ex = {k: v for k, v in zip(keys, values)}\r\n # updates the passage.\r\n passage = ex['passage']\r\n passage = re.sub(r'(\\.|\\?|\\!|\\\"|\\')\\n@highlight\\n', r'\\1 ', passage)\r\n passage = re.sub(r'\\n@highlight\\n', '. ', passage)\r\n inputs = f\"record query: {ex['query']} entities: {', '.join(ex['entities'])} passage: {passage}\"\r\n # duplicates the samples based on number of answers.\r\n num_answers = len(ex[\"answers\"])\r\n num_duplicates = np.maximum(1, num_answers)\r\n new_batch[\"inputs\"].extend([inputs] * num_duplicates) #len(ex[\"answers\"]))\r\n new_batch[\"targets\"].extend(ex[\"answers\"] if num_answers > 0 else [\"<unk>\"])\r\n return new_batch\r\n\r\ndata = datasets.load_dataset('super_glue', 'record', split=\"train\", script_version=\"master\")\r\ndata = data.map(unbatch, batched=True, remove_columns=data.column_names)\r\n```\r\n\r\nThanks a lot again, this was a super great way to do it." ]

[ { "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" } ]

[ "Can I use instructions present in below link for time series dataset as well? \r\nhttps://github.com/huggingface/datasets/blob/master/ADD_NEW_DATASET.md ", "cc'ing @kashif and @NielsRogge for visibility!", "@INF800 happy to add this dataset! I will try to set a PR by the end of the day... if you can kindly point me to the dataset? Also, note we have a bunch of time series datasets checked in e.g. `electricity_load_diagrams` or `monash_tsf`, and ideally this dataset could also be in a similar format. ", "Thankyou. This is how raw data looks like before cleaning for an individual stocks:\r\n\r\n1. https://github.com/INF800/marktech/tree/raw-data/f/data/raw\r\n2. https://github.com/INF800/marktech/tree/raw-data/t/data/raw\r\n3. https://github.com/INF800/marktech/tree/raw-data/rdfn/data/raw\r\n4. https://github.com/INF800/marktech/tree/raw-data/irbt/data/raw\r\n5. https://github.com/INF800/marktech/tree/raw-data/hll/data/raw\r\n6. https://github.com/INF800/marktech/tree/raw-data/infy/data/raw\r\n7. https://github.com/INF800/marktech/tree/raw-data/reli/data/raw\r\n8. https://github.com/INF800/marktech/tree/raw-data/hdbk/data/raw\r\n\r\n> Scraping is automated using GitHub Actions. So, everyday we will see a new file added in the above links.\r\n\r\nI can rewrite the cleaning scripts to make sure it fits HF dataset standards. (P.S I am very much new to HF dataset)\r\n\r\nThe data set above can be converted into univariate regression / multivariate regression / sequence to sequence generation dataset etc. So, do we have some kind of transformation modules that will read the dataset as some type of dataset (`GenericTimeData`) and convert it to other possible dataset relating to a specific ML task. **By having this kind of transformation module, I only have to add data once** and use transformation module whenever necessary\r\n\r\nAdditionally, having some kind of versioning for the dataset will be really helpful because it will keep on updating - especially time series datasets ", "thanks @INF800 I'll have a look. I believe it should be possible to incorporate this into the time-series format.", "Referencing https://github.com/qingsongedu/time-series-transformers-review", "@INF800 yes I am aware of the review repository and paper which is more or less a collection of abstracts etc. I am working on a unified library of implementations of these papers together with datasets to be then able to compare/contrast and build upon the research etc. but I am not ready to share them publicly just yet.\r\n\r\nIn any case regarding your dataset at the moment its seems from looking at the csv files, its mixture of textual and numerical data, sometimes in the same column etc. As you know, for time series models we would need just numeric data so I would need your help in disambiguating the dataset you have collected and also perhaps starting with just numerical data to start with... \r\n\r\nDo you think you can make a version with just numerical data?", "> @INF800 yes I am aware of the review repository and paper which is more or less a collection of abstracts etc. I am working on a unified library of implementations of these papers together with datasets to be then able to compare/contrast and build upon the research etc. but I am not ready to share them publicly just yet.\r\n> \r\n> In any case regarding your dataset at the moment its seems from looking at the csv files, its mixture of textual and numerical data, sometimes in the same column etc. As you know, for time series models we would need just numeric data so I would need your help in disambiguating the dataset you have collected and also perhaps starting with just numerical data to start with...\r\n> \r\n> Do you think you can make a version with just numerical data?\r\n\r\nWill share the numeric data and conversion script within end of this week. \r\n\r\nI am on a business trip currently - it is in my desktop." ]

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

[ "C4 is too large to be shown in the viewer" ]

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

[]

[]

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

[]

[]

[]

[ "got that the syntax is like this\r\n\r\ndataset = load_dataset(\"facebook/flores\", \"ace_Arab\")" ]

[ { "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 ! The audio data don't always exist as files on disk - the blobs are often stored in the Arrow files. For now I'd suggest disabling decoding with `.cast_column(\"audio\", Audio(decode=False))` and apply your own decoding that handles corrupted files (maybe to filter them out ?)\r\n\r\ncc @sanchit-gandhi since it's related to our discussion about allowing users to make decoding return `None` and show a warning when there are corrupted files", "Thanks @lhoestq, I wasn't aware of the decode flag. It makes more sense as you say to show a warning when there are corrupted files together with some metadata of the file that allows to filter them from the dataset.\r\n\r\nMy workaround was to catch the LibsndfileError and generate a dummy audio with an unsual sample rate to filter it later. However returning `None` seems better. \r\n\r\n`try:\r\n array, sampling_rate = sf.read(file)\r\nexcept sf.LibsndfileError:\r\n print(\"bad file\")\r\n array = np.array([0.0])\r\n sampling_rate = 99.000` \r\n\r\n" ]

[]

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

[]

[ "@patrickvonplaten \r\nI removed the mp3 files, dummy_data is much smaller now!" ]

[]

[ "Hi @EntilZha - sorry for waiting so long until taking action here. We created a new command and a new recipe of how to add dummy_data. Can you maybe rebase to `master` as explained in 7. of https://github.com/huggingface/nlp/blob/master/CONTRIBUTING.md#how-to-contribute-to-nlp and check that your dummy data is correct following the instructions here: https://github.com/huggingface/nlp/blob/master/CONTRIBUTING.md#how-to-add-a-dataset ? \r\n\r\nIf the tests described in 5. of https://github.com/huggingface/nlp/blob/master/CONTRIBUTING.md#how-to-add-a-dataset pass we can merge the PR :-) ", "I updated to the most recent master and followed the steps, but still having the similar error where it can't find the correct file since the path to the directory is given, rather than the individual files within them. This still something wrong about how I'm inputting the data or how the tests are reading it?", "It's the dummy_data structure. You actually have to call the dummy data file name `dummy_data` (not .json anything). So there should not be a `dummy_data` folder but for each config only a `dummy_data` which contains your json dummy data. Can you maybe try once more - if it doesn't work I do it for you :-). ", "Would that work if there are multiple files? In my case, I'm including something similar to squad 1.0/2.0 where we have the main dataset + an additional challenge set in different files. Would I have the zip decompress to two files in that case?", "This dataset was actually a special case. It helped us improve the dummy data instructions :-), see #195 .Close this PR and merge #194." ]

[]

[ "Even if a dataset is cached, we still make requests to check whether the cache is up-to-date. [This](https://huggingface.co/docs/datasets/v2.13.1/en/loading#offline) section in the docs explains how to avoid them and directly load the cached version.", "Thanks - that works! However it doesn't resolve the original issue (but I am not sure if it is a WSL problem)", "We use `requests` to make HTTP requests (and `aiohttp` in the streaming mode), so I don't think we can provide much help regarding the socket issue (it probably has something to do with WSL). " ]

[]

[]

[]

[]

[]

[]

[]

[ "this seems to work for my specific problem:\r\n\r\n`self.train_ds, self.test_ds, self.val_ds = map(_prepare_ds, ('train', 'test[:25%]+test[50%:75%]', 'test[75%:]'))`", "Currently you can indeed split a dataset using `ds_test = nlp.load_dataset('imdb, split='test[:5000]')` (works also with percentages).\r\n\r\nHowever right now we don't have a way to shuffle a dataset but we are thinking about it in the discussion in #166. Feel free to share your thoughts about it.\r\n\r\nOne trick that you can do until we have a better solution is to shuffle and split the indices of your dataset:\r\n```python\r\nimport nlp\r\nfrom sklearn.model_selection import train_test_split\r\n\r\nimdb = nlp.load_dataset('imbd', split='test')\r\ntest_indices, val_indices = train_test_split(range(len(imdb)))\r\n```\r\n\r\nand then to iterate each split:\r\n```python\r\nfor i in test_indices:\r\n example = imdb[i]\r\n ...\r\n```\r\n", "I added a small guide [here](https://github.com/huggingface/nlp/tree/master/docs/splits.md) that explains how to split a dataset. It is very similar to the tensorflow datasets guide, as we kept the same logic.", "Thanks a lot, the new explanation is very helpful!\r\n\r\nAbout using train_test_split from sklearn: I stumbled across the [same error message as this user ](https://github.com/huggingface/nlp/issues/147 )and thought it can't be used at the moment in this context. Will check it out again.\r\n\r\nOne of the problems is how to shuffle very large datasets, which don't fit into the memory. Well, one strategy could be shuffling data in sections. But in a case where the data is sorted by the labels you have to swap larger sections first. \r\n", "We added a way to shuffle datasets (shuffle the indices and then reorder to make a new dataset).\r\nYou can do `shuffled_dset = dataset.shuffle(seed=my_seed)`. It shuffles the whole dataset.\r\nThere is also `dataset.train_test_split()` which if very handy (with the same signature as sklearn).\r\n\r\nClosing this issue as we added the docs for splits and tools to split datasets. Thanks again for your feedback !", "https://huggingface.co/docs/datasets/v1.0.1/package_reference/builder_classes.html#datasets.Split still links to https://github.com/huggingface/datasets/tree/main/docs/splits.md which is a 404\r\n", "The updated documentation doesn't link to this anymore: https://huggingface.co/docs/datasets/v2.10.0/en/package_reference/builder_classes#datasets.Split" ]

[]

[ "Hii please follow me", "The CI error `FAILED tests/test_file_utils.py::TempSeedTest::test_tensorflow` is not related with this dataset and is fixed on master. You can ignore it", "merging since the Ci is fixed on master" ]

[]

[]

[]

[]

[]

[]

[]

[ "@lhoestq fixed :) " ]

[]

[ "The CI error is unrelated to the change." ]

[]

[ "Hi @manuaero \r\n\r\nThank you for your interest in the COYO dataset.\r\n\r\nOur dataset provides the img-url and alt-text in the form of a parquet, so to utilize the coyo dataset you will need to download it directly.\r\n\r\nWe provide a [guide](https://github.com/kakaobrain/coyo-dataset/blob/main/download/README.md) to download, so check it out.\r\n\r\nThank you." ]

[]

[]

[]

[ "Hi @hazalturkmen, thanks for reporting.\r\n\r\nNote that your dataset repository does not contain any loading script; it only contains a data file named `tr_article_2`.\r\n\r\nWhen there is no loading script but only data files, the `datasets` library tries to infer how to load the data by looking at the data file extensions. However, your data file does not have any extension.\r\n\r\nNote that current supported data file extensions are: 'csv', 'tsv', 'json', 'jsonl', 'parquet', 'txt', 'blp', 'bmp', 'dib', 'bufr', 'cur', 'pcx', 'dcx', 'dds', 'ps', 'eps', 'fit', 'fits', 'fli', 'flc', 'ftc', 'ftu', 'gbr', 'gif', 'grib', 'h5', 'hdf', 'png', 'apng', 'jp2', 'j2k', 'jpc', 'jpf', 'jpx', 'j2c', 'icns', 'ico', 'im', 'iim', 'tif', 'tiff', 'jfif', 'jpe', 'jpg', 'jpeg', 'mpg', 'mpeg', 'msp', 'pcd', 'pxr', 'pbm', 'pgm', 'ppm', 'pnm', 'psd', 'bw', 'rgb', 'rgba', 'sgi', 'ras', 'tga', 'icb', 'vda', 'vst', 'webp', 'wmf', 'emf', 'xbm', 'xpm', 'zip'.\r\n\r\nYou have more info on our docs: [How to share a dataset](https://huggingface.co/docs/datasets/share).", "thanks for reply :)" ]

[]

[ "<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) | 0.006186 / 0.011353 (-0.005167) | 0.003744 / 0.011008 (-0.007264) | 0.097295 / 0.038508 (0.058787) | 0.037106 / 0.023109 (0.013997) | 0.424154 / 0.275898 (0.148256) | 0.474536 / 0.323480 (0.151057) | 0.003454 / 0.007986 (-0.004532) | 0.003865 / 0.004328 (-0.000463) | 0.077348 / 0.004250 (0.073097) | 0.051728 / 0.037052 (0.014675) | 0.437120 / 0.258489 (0.178631) | 0.478379 / 0.293841 (0.184538) | 0.028939 / 0.128546 (-0.099608) | 0.008376 / 0.075646 (-0.067270) | 0.312002 / 0.419271 (-0.107270) | 0.053723 / 0.043533 (0.010190) | 0.424815 / 0.255139 (0.169676) | 0.446203 / 0.283200 (0.163004) | 0.026553 / 0.141683 (-0.115130) | 1.479983 / 1.452155 (0.027828) | 1.530613 / 1.492716 (0.037896) |\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) | 0.196627 / 0.018006 (0.178620) | 0.422361 / 0.000490 (0.421871) | 0.003442 / 0.000200 (0.003242) | 0.000077 / 0.000054 (0.000023) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.022913 / 0.037411 (-0.014499) | 0.096011 / 0.014526 (0.081485) | 0.104091 / 0.176557 (-0.072466) | 0.163273 / 0.737135 (-0.573862) | 0.109142 / 0.296338 (-0.187197) |\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) | 0.431032 / 0.215209 (0.215823) | 4.314391 / 2.077655 (2.236737) | 2.003812 / 1.504120 (0.499692) | 1.799538 / 1.541195 (0.258344) | 1.830026 / 1.468490 (0.361536) | 0.560131 / 4.584777 (-4.024646) | 3.368997 / 3.745712 (-0.376715) | 1.703032 / 5.269862 (-3.566830) | 1.026949 / 4.565676 (-3.538727) | 0.067507 / 0.424275 (-0.356768) | 0.010910 / 0.007607 (0.003303) | 0.532606 / 0.226044 (0.306562) | 5.345179 / 2.268929 (3.076250) | 2.368077 / 55.444624 (-53.076548) | 2.028913 / 6.876477 (-4.847564) | 2.147621 / 2.142072 (0.005549) | 0.675696 / 4.805227 (-4.129531) | 0.134902 / 6.500664 (-6.365762) | 0.065004 / 0.075469 (-0.010465) |\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) | 1.233412 / 1.841788 (-0.608376) | 13.767465 / 8.074308 (5.693157) | 13.933653 / 10.191392 (3.742261) | 0.129010 / 0.680424 (-0.551414) | 0.016708 / 0.534201 (-0.517493) | 0.362341 / 0.579283 (-0.216942) | 0.390902 / 0.434364 (-0.043462) | 0.429156 / 0.540337 (-0.111182) | 0.521166 / 1.386936 (-0.865770) |\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) | 0.006169 / 0.011353 (-0.005184) | 0.003839 / 0.011008 (-0.007169) | 0.078784 / 0.038508 (0.040276) | 0.040218 / 0.023109 (0.017109) | 0.360439 / 0.275898 (0.084541) | 0.423957 / 0.323480 (0.100477) | 0.003456 / 0.007986 (-0.004529) | 0.002900 / 0.004328 (-0.001428) | 0.078820 / 0.004250 (0.074569) | 0.047240 / 0.037052 (0.010187) | 0.372081 / 0.258489 (0.113592) | 0.424263 / 0.293841 (0.130422) | 0.027977 / 0.128546 (-0.100569) | 0.008400 / 0.075646 (-0.067246) | 0.084399 / 0.419271 (-0.334872) | 0.043303 / 0.043533 (-0.000230) | 0.361583 / 0.255139 (0.106444) | 0.394987 / 0.283200 (0.111787) | 0.020006 / 0.141683 (-0.121677) | 1.520208 / 1.452155 (0.068053) | 1.587335 / 1.492716 (0.094619) |\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) | 0.223847 / 0.018006 (0.205840) | 0.402194 / 0.000490 (0.401704) | 0.000384 / 0.000200 (0.000184) | 0.000057 / 0.000054 (0.000003) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.024902 / 0.037411 (-0.012509) | 0.099076 / 0.014526 (0.084550) | 0.108041 / 0.176557 (-0.068516) | 0.159385 / 0.737135 (-0.577750) | 0.111442 / 0.296338 (-0.184896) |\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) | 0.446232 / 0.215209 (0.231023) | 4.464927 / 2.077655 (2.387272) | 2.155234 / 1.504120 (0.651114) | 1.953645 / 1.541195 (0.412450) | 1.965991 / 1.468490 (0.497501) | 0.553473 / 4.584777 (-4.031304) | 3.321397 / 3.745712 (-0.424315) | 1.693761 / 5.269862 (-3.576101) | 1.006299 / 4.565676 (-3.559378) | 0.067013 / 0.424275 (-0.357262) | 0.011116 / 0.007607 (0.003509) | 0.555014 / 0.226044 (0.328970) | 5.535694 / 2.268929 (3.266765) | 2.598339 / 55.444624 (-52.846285) | 2.249298 / 6.876477 (-4.627179) | 2.243419 / 2.142072 (0.101347) | 0.667603 / 4.805227 (-4.137624) | 0.133322 / 6.500664 (-6.367343) | 0.065473 / 0.075469 (-0.009996) |\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) | 1.293051 / 1.841788 (-0.548737) | 14.103731 / 8.074308 (6.029423) | 14.215204 / 10.191392 (4.023812) | 0.143990 / 0.680424 (-0.536434) | 0.016805 / 0.534201 (-0.517396) | 0.363264 / 0.579283 (-0.216019) | 0.392769 / 0.434364 (-0.041594) | 0.425291 / 0.540337 (-0.115046) | 0.515479 / 1.386936 (-0.871457) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#e03a58f3f5d7e6f07279fb833e62d859a0babaad \"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) | 0.006346 / 0.011353 (-0.005006) | 0.004130 / 0.011008 (-0.006878) | 0.096898 / 0.038508 (0.058390) | 0.042564 / 0.023109 (0.019455) | 0.343748 / 0.275898 (0.067850) | 0.412515 / 0.323480 (0.089035) | 0.006153 / 0.007986 (-0.001833) | 0.003345 / 0.004328 (-0.000984) | 0.075314 / 0.004250 (0.071064) | 0.061478 / 0.037052 (0.024426) | 0.362948 / 0.258489 (0.104459) | 0.401533 / 0.293841 (0.107692) | 0.032363 / 0.128546 (-0.096184) | 0.008780 / 0.075646 (-0.066867) | 0.328691 / 0.419271 (-0.090580) | 0.054253 / 0.043533 (0.010721) | 0.340783 / 0.255139 (0.085644) | 0.360705 / 0.283200 (0.077505) | 0.023183 / 0.141683 (-0.118500) | 1.484078 / 1.452155 (0.031924) | 1.528581 / 1.492716 (0.035865) |\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) | 0.208732 / 0.018006 (0.190726) | 0.452572 / 0.000490 (0.452082) | 0.002936 / 0.000200 (0.002737) | 0.000082 / 0.000054 (0.000028) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.024616 / 0.037411 (-0.012795) | 0.107547 / 0.014526 (0.093021) | 0.114492 / 0.176557 (-0.062065) | 0.171770 / 0.737135 (-0.565365) | 0.122538 / 0.296338 (-0.173800) |\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) | 0.406140 / 0.215209 (0.190930) | 4.062391 / 2.077655 (1.984736) | 1.865962 / 1.504120 (0.361842) | 1.682236 / 1.541195 (0.141041) | 1.738119 / 1.468490 (0.269629) | 0.532244 / 4.584777 (-4.052533) | 3.816421 / 3.745712 (0.070709) | 2.981205 / 5.269862 (-2.288656) | 1.519497 / 4.565676 (-3.046179) | 0.065904 / 0.424275 (-0.358371) | 0.011277 / 0.007607 (0.003670) | 0.512789 / 0.226044 (0.286745) | 5.107618 / 2.268929 (2.838690) | 2.419399 / 55.444624 (-53.025226) | 2.079262 / 6.876477 (-4.797214) | 2.150447 / 2.142072 (0.008375) | 0.696737 / 4.805227 (-4.108490) | 0.142497 / 6.500664 (-6.358167) | 0.063521 / 0.075469 (-0.011949) |\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) | 1.180692 / 1.841788 (-0.661095) | 14.343084 / 8.074308 (6.268776) | 13.303719 / 10.191392 (3.112327) | 0.164234 / 0.680424 (-0.516190) | 0.017439 / 0.534201 (-0.516762) | 0.399712 / 0.579283 (-0.179571) | 0.428248 / 0.434364 (-0.006115) | 0.471909 / 0.540337 (-0.068428) | 0.573853 / 1.386936 (-0.813083) |\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) | 0.006210 / 0.011353 (-0.005143) | 0.004104 / 0.011008 (-0.006905) | 0.075140 / 0.038508 (0.036632) | 0.044647 / 0.023109 (0.021538) | 0.370120 / 0.275898 (0.094222) | 0.452936 / 0.323480 (0.129457) | 0.003943 / 0.007986 (-0.004042) | 0.003285 / 0.004328 (-0.001043) | 0.075267 / 0.004250 (0.071017) | 0.055517 / 0.037052 (0.018465) | 0.396385 / 0.258489 (0.137896) | 0.447870 / 0.293841 (0.154029) | 0.031342 / 0.128546 (-0.097204) | 0.008720 / 0.075646 (-0.066926) | 0.082702 / 0.419271 (-0.336570) | 0.051010 / 0.043533 (0.007477) | 0.350546 / 0.255139 (0.095407) | 0.425395 / 0.283200 (0.142195) | 0.024483 / 0.141683 (-0.117200) | 1.467341 / 1.452155 (0.015186) | 1.537187 / 1.492716 (0.044471) |\n\n### Benchmark: benchmark_getitem\\_100B.json\n\n| metric | get_batch_of\\_1024\\_random_rows | get_batch_of\\_1024\\_rows | get_first_row | get_last_row |\n|--------|---|---|---|---|\n| new / old (diff) | 0.218067 / 0.018006 (0.200061) | 0.441603 / 0.000490 (0.441114) | 0.003711 / 0.000200 (0.003512) | 0.000092 / 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) | 0.028669 / 0.037411 (-0.008742) | 0.112941 / 0.014526 (0.098415) | 0.122584 / 0.176557 (-0.053972) | 0.176494 / 0.737135 (-0.560641) | 0.129369 / 0.296338 (-0.166970) |\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) | 0.434543 / 0.215209 (0.219334) | 4.344056 / 2.077655 (2.266401) | 2.079286 / 1.504120 (0.575166) | 1.887264 / 1.541195 (0.346069) | 1.910386 / 1.468490 (0.441896) | 0.538824 / 4.584777 (-4.045953) | 3.844786 / 3.745712 (0.099074) | 2.902091 / 5.269862 (-2.367770) | 1.270852 / 4.565676 (-3.294824) | 0.066324 / 0.424275 (-0.357951) | 0.011346 / 0.007607 (0.003739) | 0.537122 / 0.226044 (0.311078) | 5.367354 / 2.268929 (3.098426) | 2.533672 / 55.444624 (-52.910952) | 2.203260 / 6.876477 (-4.673217) | 2.224310 / 2.142072 (0.082237) | 0.663806 / 4.805227 (-4.141422) | 0.142758 / 6.500664 (-6.357906) | 0.063870 / 0.075469 (-0.011599) |\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) | 1.260487 / 1.841788 (-0.581301) | 14.800106 / 8.074308 (6.725798) | 13.993488 / 10.191392 (3.802096) | 0.165829 / 0.680424 (-0.514595) | 0.017347 / 0.534201 (-0.516854) | 0.401819 / 0.579283 (-0.177464) | 0.424577 / 0.434364 (-0.009787) | 0.475161 / 0.540337 (-0.065176) | 0.574659 / 1.386936 (-0.812277) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#02e1e9ab6df4720f57b2d08c0b800cecac79a7c8 \"CML watermark\")\n" ]

[ { "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" } ]

[ "Can you make it a draft PR until you've added the dataset please ? @ekdnam ", "Done", "Thanks for your contribution, @ekdnam. 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." ]

[]