Fix CI 404 errors

[albertvillanova]

Currently our CI usually raises 404 errors when trying to delete temporary repositories. See, e.g.: https://github.com/huggingface/datasets/actions/runs/6314980985/job/17146507884

FAILED tests/test_upstream_hub.py::TestPushToHub::test_push_dataset_dict_to_hub_multiple_files_with_max_shard_size - huggingface_hub.utils._errors.RepositoryNotFoundError: 404 Client Error. (Request ID: Root=1-6512fb99-4a52c561752ece3d77eb6d57;2b61cae4-613d-4a73-bbb1-2faf9e32b02d)
Repository Not Found for url: https://hub-ci.huggingface.co/api/repos/delete.
Please make sure you specified the correct `repo_id` and `repo_type`.
If you are trying to access a private or gated repo, make sure you are authenticated.

FAILED tests/test_upstream_hub.py::TestPushToHub::test_push_dataset_to_hub_custom_features_audio - huggingface_hub.utils._errors.RepositoryNotFoundError: 404 Client Error. (Request ID: Root=1-6512fbb2-0333dd666d42f0e173c2bb68;dfdc4271-b49b-4008-8c49-f05cf7c1d53d)
Repository Not Found for url: https://hub-ci.huggingface.co/api/repos/delete.
Please make sure you specified the correct `repo_id` and `repo_type`.
If you are trying to access a private or gated repo, make sure you are authenticated.

FAILED tests/test_upstream_hub.py::TestPushToHub::test_push_dataset_dict_to_hub_custom_splits - huggingface_hub.utils._errors.RepositoryNotFoundError: 404 Client Error. (Request ID: Root=1-6512fbca-167690694f39770a5b3a444e;baeaa905-0a57-4585-ac97-9aaae12dd47d)
Repository Not Found for url: https://hub-ci.huggingface.co/api/repos/delete.
Please make sure you specified the correct `repo_id` and `repo_type`.
If you are trying to access a private or gated repo, make sure you are authenticated.

I think this can be caused by collisions in temporary repository IDs because we create them in multiprocessing:

with temporary_repo(f"{CI_HUB_USER}/test-{int(time.time() * 10e3)}") as ds_name:

This can also be caused when there is another issue that does not allow the creation of the repository, thus making it impossible to delete it.

This PR tries to fix this issue by increasing the precision of the number on the repository ID: 10e6 instead of 10e3.
Additionally, this PR catches RepositoryNotFoundError.

[github-actions]

Show benchmarks

PyArrow==8.0.0

Show updated benchmarks!

Benchmark: benchmark_array_xd.json

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
new / old (diff)	0.008220 / 0.011353 (-0.003133)	0.005560 / 0.011008 (-0.005448)	0.100147 / 0.038508 (0.061639)	0.070106 / 0.023109 (0.046996)	0.411906 / 0.275898 (0.136008)	0.432825 / 0.323480 (0.109345)	0.004795 / 0.007986 (-0.003190)	0.004094 / 0.004328 (-0.000235)	0.075719 / 0.004250 (0.071468)	0.067426 / 0.037052 (0.030374)	0.428531 / 0.258489 (0.170042)	0.437114 / 0.293841 (0.143273)	0.045603 / 0.128546 (-0.082943)	0.013333 / 0.075646 (-0.062313)	0.353137 / 0.419271 (-0.066134)	0.067902 / 0.043533 (0.024369)	0.396633 / 0.255139 (0.141494)	0.399185 / 0.283200 (0.115985)	0.036377 / 0.141683 (-0.105306)	1.624249 / 1.452155 (0.172094)	1.792575 / 1.492716 (0.299859)

Benchmark: benchmark_getitem_100B.json

metric	get_batch_of_1024_random_rows	get_batch_of_1024_rows	get_first_row	get_last_row
new / old (diff)	0.315847 / 0.018006 (0.297840)	0.595009 / 0.000490 (0.594519)	0.018876 / 0.000200 (0.018676)	0.000613 / 0.000054 (0.000558)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.029886 / 0.037411 (-0.007526)	0.085765 / 0.014526 (0.071239)	0.108680 / 0.176557 (-0.067877)	0.174588 / 0.737135 (-0.562548)	0.104494 / 0.296338 (-0.191844)

Benchmark: benchmark_iterating.json

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
new / old (diff)	0.594429 / 0.215209 (0.379220)	5.912352 / 2.077655 (3.834698)	2.408501 / 1.504120 (0.904381)	2.050914 / 1.541195 (0.509720)	2.199349 / 1.468490 (0.730859)	0.813797 / 4.584777 (-3.770980)	5.169577 / 3.745712 (1.423864)	4.653951 / 5.269862 (-0.615911)	2.805423 / 4.565676 (-1.760253)	0.092278 / 0.424275 (-0.331997)	0.007394 / 0.007607 (-0.000213)	0.684029 / 0.226044 (0.457985)	6.964260 / 2.268929 (4.695331)	3.108408 / 55.444624 (-52.336217)	2.470907 / 6.876477 (-4.405569)	2.460153 / 2.142072 (0.318081)	0.986445 / 4.805227 (-3.818782)	0.213069 / 6.500664 (-6.287596)	0.074061 / 0.075469 (-0.001408)

Benchmark: benchmark_map_filter.json

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
new / old (diff)	1.590732 / 1.841788 (-0.251056)	23.736918 / 8.074308 (15.662609)	21.223910 / 10.191392 (11.032518)	0.236173 / 0.680424 (-0.444251)	0.030056 / 0.534201 (-0.504145)	0.489461 / 0.579283 (-0.089822)	0.607582 / 0.434364 (0.173218)	0.539889 / 0.540337 (-0.000449)	0.817942 / 1.386936 (-0.568994)

PyArrow==latest

Show updated benchmarks!

Benchmark: benchmark_array_xd.json

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
new / old (diff)	0.008042 / 0.011353 (-0.003311)	0.004836 / 0.011008 (-0.006173)	0.075434 / 0.038508 (0.036926)	0.080818 / 0.023109 (0.057709)	0.474797 / 0.275898 (0.198899)	0.526168 / 0.323480 (0.202689)	0.006463 / 0.007986 (-0.001522)	0.004031 / 0.004328 (-0.000297)	0.074141 / 0.004250 (0.069891)	0.068265 / 0.037052 (0.031212)	0.562550 / 0.258489 (0.304061)	0.544820 / 0.293841 (0.250979)	0.047263 / 0.128546 (-0.081283)	0.014113 / 0.075646 (-0.061534)	0.086061 / 0.419271 (-0.333210)	0.062475 / 0.043533 (0.018942)	0.479912 / 0.255139 (0.224773)	0.494784 / 0.283200 (0.211584)	0.035847 / 0.141683 (-0.105836)	1.726452 / 1.452155 (0.274297)	1.770113 / 1.492716 (0.277396)

Benchmark: benchmark_getitem_100B.json

metric	get_batch_of_1024_random_rows	get_batch_of_1024_rows	get_first_row	get_last_row
new / old (diff)	0.286713 / 0.018006 (0.268707)	0.609704 / 0.000490 (0.609214)	0.009342 / 0.000200 (0.009143)	0.000134 / 0.000054 (0.000080)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.035137 / 0.037411 (-0.002275)	0.099331 / 0.014526 (0.084805)	0.108971 / 0.176557 (-0.067586)	0.170952 / 0.737135 (-0.566183)	0.111736 / 0.296338 (-0.184603)

Benchmark: benchmark_iterating.json

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
new / old (diff)	0.617434 / 0.215209 (0.402225)	6.204351 / 2.077655 (4.126697)	2.854347 / 1.504120 (1.350227)	2.557424 / 1.541195 (1.016229)	2.638173 / 1.468490 (1.169683)	0.854234 / 4.584777 (-3.730543)	5.383288 / 3.745712 (1.637576)	4.698098 / 5.269862 (-0.571763)	2.903860 / 4.565676 (-1.661817)	0.094689 / 0.424275 (-0.329586)	0.007892 / 0.007607 (0.000285)	0.729420 / 0.226044 (0.503376)	7.356691 / 2.268929 (5.087763)	3.708039 / 55.444624 (-51.736585)	2.979734 / 6.876477 (-3.896743)	2.978983 / 2.142072 (0.836911)	1.040554 / 4.805227 (-3.764673)	0.211246 / 6.500664 (-6.289418)	0.079880 / 0.075469 (0.004411)

Benchmark: benchmark_map_filter.json

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
new / old (diff)	1.676057 / 1.841788 (-0.165731)	23.428443 / 8.074308 (15.354135)	21.016293 / 10.191392 (10.824901)	0.260927 / 0.680424 (-0.419497)	0.030689 / 0.534201 (-0.503512)	0.495652 / 0.579283 (-0.083632)	0.622976 / 0.434364 (0.188612)	0.561175 / 0.540337 (0.020837)	0.786733 / 1.386936 (-0.600203)

[HuggingFaceDocBuilderDev]

The documentation is not available anymore as the PR was closed or merged.

[github-actions]

Show benchmarks

PyArrow==8.0.0

Show updated benchmarks!

Benchmark: benchmark_array_xd.json

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
new / old (diff)	0.005942 / 0.011353 (-0.005410)	0.003706 / 0.011008 (-0.007302)	0.081002 / 0.038508 (0.042493)	0.056854 / 0.023109 (0.033745)	0.358668 / 0.275898 (0.082770)	0.369718 / 0.323480 (0.046238)	0.005202 / 0.007986 (-0.002784)	0.002841 / 0.004328 (-0.001487)	0.062976 / 0.004250 (0.058726)	0.051308 / 0.037052 (0.014255)	0.373636 / 0.258489 (0.115147)	0.390480 / 0.293841 (0.096639)	0.027480 / 0.128546 (-0.101067)	0.007960 / 0.075646 (-0.067686)	0.262719 / 0.419271 (-0.156552)	0.046488 / 0.043533 (0.002955)	0.347299 / 0.255139 (0.092160)	0.393448 / 0.283200 (0.110249)	0.019445 / 0.141683 (-0.122238)	1.431314 / 1.452155 (-0.020841)	1.495578 / 1.492716 (0.002862)

Benchmark: benchmark_getitem_100B.json

metric	get_batch_of_1024_random_rows	get_batch_of_1024_rows	get_first_row	get_last_row
new / old (diff)	0.223724 / 0.018006 (0.205718)	0.416929 / 0.000490 (0.416440)	0.005253 / 0.000200 (0.005053)	0.000217 / 0.000054 (0.000163)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.023571 / 0.037411 (-0.013841)	0.073503 / 0.014526 (0.058978)	0.081366 / 0.176557 (-0.095190)	0.142716 / 0.737135 (-0.594420)	0.082612 / 0.296338 (-0.213727)

Benchmark: benchmark_iterating.json

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
new / old (diff)	0.407319 / 0.215209 (0.192109)	4.141404 / 2.077655 (2.063749)	1.910842 / 1.504120 (0.406722)	1.731694 / 1.541195 (0.190499)	1.805228 / 1.468490 (0.336738)	0.497109 / 4.584777 (-4.087668)	3.107624 / 3.745712 (-0.638088)	2.890687 / 5.269862 (-2.379174)	1.795913 / 4.565676 (-2.769763)	0.057099 / 0.424275 (-0.367176)	0.006414 / 0.007607 (-0.001194)	0.482127 / 0.226044 (0.256083)	4.835158 / 2.268929 (2.566229)	2.368909 / 55.444624 (-53.075715)	2.001608 / 6.876477 (-4.874868)	2.004492 / 2.142072 (-0.137580)	0.579910 / 4.805227 (-4.225317)	0.123541 / 6.500664 (-6.377123)	0.059651 / 0.075469 (-0.015818)

Benchmark: benchmark_map_filter.json

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
new / old (diff)	1.242364 / 1.841788 (-0.599424)	16.982676 / 8.074308 (8.908368)	13.718885 / 10.191392 (3.527493)	0.132759 / 0.680424 (-0.547665)	0.017012 / 0.534201 (-0.517189)	0.333447 / 0.579283 (-0.245836)	0.360149 / 0.434364 (-0.074215)	0.385526 / 0.540337 (-0.154811)	0.536915 / 1.386936 (-0.850021)

PyArrow==latest

Show updated benchmarks!

Benchmark: benchmark_array_xd.json

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
new / old (diff)	0.005946 / 0.011353 (-0.005407)	0.003442 / 0.011008 (-0.007566)	0.062595 / 0.038508 (0.024087)	0.058699 / 0.023109 (0.035590)	0.442626 / 0.275898 (0.166728)	0.473773 / 0.323480 (0.150293)	0.004622 / 0.007986 (-0.003364)	0.002812 / 0.004328 (-0.001516)	0.064099 / 0.004250 (0.059849)	0.046784 / 0.037052 (0.009731)	0.466049 / 0.258489 (0.207560)	0.487912 / 0.293841 (0.194071)	0.028372 / 0.128546 (-0.100174)	0.007992 / 0.075646 (-0.067654)	0.068151 / 0.419271 (-0.351120)	0.041010 / 0.043533 (-0.002523)	0.442331 / 0.255139 (0.187192)	0.469686 / 0.283200 (0.186487)	0.019694 / 0.141683 (-0.121989)	1.467928 / 1.452155 (0.015774)	1.525635 / 1.492716 (0.032918)

Benchmark: benchmark_getitem_100B.json

metric	get_batch_of_1024_random_rows	get_batch_of_1024_rows	get_first_row	get_last_row
new / old (diff)	0.204459 / 0.018006 (0.186453)	0.407766 / 0.000490 (0.407276)	0.003898 / 0.000200 (0.003698)	0.000077 / 0.000054 (0.000023)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.025909 / 0.037411 (-0.011503)	0.080341 / 0.014526 (0.065816)	0.088231 / 0.176557 (-0.088325)	0.144056 / 0.737135 (-0.593079)	0.089769 / 0.296338 (-0.206569)

Benchmark: benchmark_iterating.json

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
new / old (diff)	0.462876 / 0.215209 (0.247667)	4.625983 / 2.077655 (2.548329)	2.580079 / 1.504120 (1.075959)	2.402792 / 1.541195 (0.861597)	2.424982 / 1.468490 (0.956491)	0.503654 / 4.584777 (-4.081123)	3.178995 / 3.745712 (-0.566717)	2.956126 / 5.269862 (-2.313735)	1.847837 / 4.565676 (-2.717840)	0.057964 / 0.424275 (-0.366311)	0.006405 / 0.007607 (-0.001202)	0.536036 / 0.226044 (0.309992)	5.374416 / 2.268929 (3.105487)	3.036440 / 55.444624 (-52.408184)	2.682054 / 6.876477 (-4.194422)	2.683462 / 2.142072 (0.541390)	0.592751 / 4.805227 (-4.212477)	0.124313 / 6.500664 (-6.376351)	0.061127 / 0.075469 (-0.014342)

Benchmark: benchmark_map_filter.json

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
new / old (diff)	1.383539 / 1.841788 (-0.458249)	17.766221 / 8.074308 (9.691913)	15.306600 / 10.191392 (5.115208)	0.145035 / 0.680424 (-0.535389)	0.018078 / 0.534201 (-0.516123)	0.330102 / 0.579283 (-0.249181)	0.375380 / 0.434364 (-0.058984)	0.388531 / 0.540337 (-0.151807)	0.548720 / 1.386936 (-0.838216)

[github-actions]

Show benchmarks

PyArrow==8.0.0

Show updated benchmarks!

Benchmark: benchmark_array_xd.json

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
new / old (diff)	0.006757 / 0.011353 (-0.004596)	0.004110 / 0.011008 (-0.006898)	0.084727 / 0.038508 (0.046219)	0.074328 / 0.023109 (0.051219)	0.310467 / 0.275898 (0.034569)	0.343209 / 0.323480 (0.019729)	0.004228 / 0.007986 (-0.003757)	0.003400 / 0.004328 (-0.000929)	0.065546 / 0.004250 (0.061296)	0.063057 / 0.037052 (0.026005)	0.315023 / 0.258489 (0.056534)	0.356395 / 0.293841 (0.062554)	0.031959 / 0.128546 (-0.096588)	0.008577 / 0.075646 (-0.067069)	0.289075 / 0.419271 (-0.130196)	0.055011 / 0.043533 (0.011478)	0.308861 / 0.255139 (0.053722)	0.328691 / 0.283200 (0.045491)	0.027037 / 0.141683 (-0.114646)	1.464314 / 1.452155 (0.012159)	1.549644 / 1.492716 (0.056927)

Benchmark: benchmark_getitem_100B.json

metric	get_batch_of_1024_random_rows	get_batch_of_1024_rows	get_first_row	get_last_row
new / old (diff)	0.238330 / 0.018006 (0.220324)	0.451570 / 0.000490 (0.451080)	0.010873 / 0.000200 (0.010673)	0.000341 / 0.000054 (0.000286)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.029909 / 0.037411 (-0.007503)	0.085222 / 0.014526 (0.070696)	0.100180 / 0.176557 (-0.076377)	0.154842 / 0.737135 (-0.582293)	0.099253 / 0.296338 (-0.197086)

Benchmark: benchmark_iterating.json

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
new / old (diff)	0.401603 / 0.215209 (0.186394)	4.009781 / 2.077655 (1.932126)	2.021807 / 1.504120 (0.517687)	1.861017 / 1.541195 (0.319822)	2.009072 / 1.468490 (0.540582)	0.483798 / 4.584777 (-4.100979)	3.580394 / 3.745712 (-0.165318)	3.464587 / 5.269862 (-1.805275)	2.018400 / 4.565676 (-2.547276)	0.057134 / 0.424275 (-0.367141)	0.007303 / 0.007607 (-0.000304)	0.473627 / 0.226044 (0.247582)	4.722634 / 2.268929 (2.453706)	2.490884 / 55.444624 (-52.953741)	2.121568 / 6.876477 (-4.754909)	2.200699 / 2.142072 (0.058626)	0.576728 / 4.805227 (-4.228499)	0.135633 / 6.500664 (-6.365032)	0.061625 / 0.075469 (-0.013844)

Benchmark: benchmark_map_filter.json

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
new / old (diff)	1.250545 / 1.841788 (-0.591243)	19.167642 / 8.074308 (11.093334)	14.189891 / 10.191392 (3.998499)	0.164552 / 0.680424 (-0.515872)	0.018215 / 0.534201 (-0.515986)	0.389962 / 0.579283 (-0.189321)	0.413972 / 0.434364 (-0.020392)	0.460253 / 0.540337 (-0.080085)	0.647897 / 1.386936 (-0.739039)

PyArrow==latest

Show updated benchmarks!

Benchmark: benchmark_array_xd.json

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
new / old (diff)	0.006714 / 0.011353 (-0.004639)	0.004081 / 0.011008 (-0.006927)	0.065627 / 0.038508 (0.027119)	0.077644 / 0.023109 (0.054535)	0.409950 / 0.275898 (0.134052)	0.442940 / 0.323480 (0.119460)	0.005523 / 0.007986 (-0.002463)	0.003366 / 0.004328 (-0.000962)	0.065425 / 0.004250 (0.061174)	0.056222 / 0.037052 (0.019169)	0.429928 / 0.258489 (0.171439)	0.457136 / 0.293841 (0.163296)	0.032356 / 0.128546 (-0.096190)	0.008676 / 0.075646 (-0.066970)	0.071785 / 0.419271 (-0.347486)	0.048458 / 0.043533 (0.004925)	0.408003 / 0.255139 (0.152864)	0.433529 / 0.283200 (0.150330)	0.023232 / 0.141683 (-0.118450)	1.483640 / 1.452155 (0.031485)	1.552425 / 1.492716 (0.059709)

Benchmark: benchmark_getitem_100B.json

metric	get_batch_of_1024_random_rows	get_batch_of_1024_rows	get_first_row	get_last_row
new / old (diff)	0.282347 / 0.018006 (0.264341)	0.448742 / 0.000490 (0.448253)	0.039590 / 0.000200 (0.039390)	0.000407 / 0.000054 (0.000353)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.032516 / 0.037411 (-0.004896)	0.095269 / 0.014526 (0.080744)	0.106363 / 0.176557 (-0.070193)	0.157945 / 0.737135 (-0.579191)	0.106783 / 0.296338 (-0.189556)

Benchmark: benchmark_iterating.json

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
new / old (diff)	0.436334 / 0.215209 (0.221125)	4.348147 / 2.077655 (2.270492)	2.326830 / 1.504120 (0.822710)	2.162586 / 1.541195 (0.621391)	2.257769 / 1.468490 (0.789279)	0.491677 / 4.584777 (-4.093099)	3.707385 / 3.745712 (-0.038328)	3.567147 / 5.269862 (-1.702715)	2.099451 / 4.565676 (-2.466226)	0.058486 / 0.424275 (-0.365789)	0.007324 / 0.007607 (-0.000283)	0.510962 / 0.226044 (0.284917)	5.106550 / 2.268929 (2.837622)	2.785723 / 55.444624 (-52.658901)	2.452928 / 6.876477 (-4.423548)	2.545034 / 2.142072 (0.402961)	0.611124 / 4.805227 (-4.194103)	0.133503 / 6.500664 (-6.367161)	0.061118 / 0.075469 (-0.014351)

Benchmark: benchmark_map_filter.json

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
new / old (diff)	1.386640 / 1.841788 (-0.455148)	20.485670 / 8.074308 (12.411362)	15.332223 / 10.191392 (5.140831)	0.164070 / 0.680424 (-0.516354)	0.019962 / 0.534201 (-0.514239)	0.394217 / 0.579283 (-0.185066)	0.428442 / 0.434364 (-0.005922)	0.473784 / 0.540337 (-0.066553)	0.665141 / 1.386936 (-0.721795)

[albertvillanova]

The CI errors seem unrelated to this PR but I think they need further investigation in another PR.

FAILED tests/test_upstream_hub.py::TestPushToHub::test_push_dataset_dict_to_hub_multiple_files - KeyError: 'url'

[github-actions]

Show benchmarks

PyArrow==8.0.0

Show updated benchmarks!

Benchmark: benchmark_array_xd.json

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
new / old (diff)	0.008766 / 0.011353 (-0.002587)	0.005289 / 0.011008 (-0.005720)	0.097220 / 0.038508 (0.058712)	0.072246 / 0.023109 (0.049137)	0.369359 / 0.275898 (0.093461)	0.422571 / 0.323480 (0.099091)	0.004941 / 0.007986 (-0.003044)	0.006103 / 0.004328 (0.001774)	0.075828 / 0.004250 (0.071578)	0.065795 / 0.037052 (0.028743)	0.412835 / 0.258489 (0.154346)	0.430062 / 0.293841 (0.136221)	0.045806 / 0.128546 (-0.082741)	0.013760 / 0.075646 (-0.061887)	0.351542 / 0.419271 (-0.067729)	0.064836 / 0.043533 (0.021304)	0.370162 / 0.255139 (0.115023)	0.434949 / 0.283200 (0.151749)	0.039198 / 0.141683 (-0.102485)	1.670940 / 1.452155 (0.218785)	1.809677 / 1.492716 (0.316961)

Benchmark: benchmark_getitem_100B.json

metric	get_batch_of_1024_random_rows	get_batch_of_1024_rows	get_first_row	get_last_row
new / old (diff)	0.295104 / 0.018006 (0.277097)	0.594584 / 0.000490 (0.594095)	0.010923 / 0.000200 (0.010723)	0.000479 / 0.000054 (0.000425)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.029174 / 0.037411 (-0.008237)	0.094637 / 0.014526 (0.080111)	0.102948 / 0.176557 (-0.073608)	0.171048 / 0.737135 (-0.566087)	0.111465 / 0.296338 (-0.184873)

Benchmark: benchmark_iterating.json

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
new / old (diff)	0.582017 / 0.215209 (0.366808)	5.727008 / 2.077655 (3.649354)	2.563211 / 1.504120 (1.059091)	2.308912 / 1.541195 (0.767717)	2.301258 / 1.468490 (0.832768)	0.819594 / 4.584777 (-3.765183)	5.177536 / 3.745712 (1.431824)	4.473602 / 5.269862 (-0.796260)	2.743819 / 4.565676 (-1.821857)	0.090052 / 0.424275 (-0.334223)	0.007903 / 0.007607 (0.000295)	0.679142 / 0.226044 (0.453097)	6.887891 / 2.268929 (4.618962)	3.337926 / 55.444624 (-52.106699)	2.659228 / 6.876477 (-4.217249)	2.641289 / 2.142072 (0.499216)	0.974829 / 4.805227 (-3.830398)	0.205775 / 6.500664 (-6.294890)	0.075268 / 0.075469 (-0.000201)

Benchmark: benchmark_map_filter.json

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
new / old (diff)	1.500562 / 1.841788 (-0.341226)	22.688483 / 8.074308 (14.614175)	19.634878 / 10.191392 (9.443486)	0.227409 / 0.680424 (-0.453015)	0.029794 / 0.534201 (-0.504407)	0.475204 / 0.579283 (-0.104079)	0.579379 / 0.434364 (0.145016)	0.541244 / 0.540337 (0.000907)	0.739187 / 1.386936 (-0.647749)

PyArrow==latest

Show updated benchmarks!

Benchmark: benchmark_array_xd.json

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
new / old (diff)	0.008641 / 0.011353 (-0.002712)	0.006139 / 0.011008 (-0.004870)	0.075048 / 0.038508 (0.036540)	0.074070 / 0.023109 (0.050961)	0.508288 / 0.275898 (0.232390)	0.539770 / 0.323480 (0.216290)	0.006092 / 0.007986 (-0.001894)	0.003748 / 0.004328 (-0.000581)	0.077945 / 0.004250 (0.073695)	0.056989 / 0.037052 (0.019936)	0.526889 / 0.258489 (0.268400)	0.560862 / 0.293841 (0.267021)	0.046507 / 0.128546 (-0.082040)	0.013249 / 0.075646 (-0.062397)	0.088363 / 0.419271 (-0.330908)	0.058776 / 0.043533 (0.015243)	0.495869 / 0.255139 (0.240730)	0.538615 / 0.283200 (0.255415)	0.034055 / 0.141683 (-0.107628)	1.658713 / 1.452155 (0.206558)	1.736599 / 1.492716 (0.243883)

Benchmark: benchmark_getitem_100B.json

metric	get_batch_of_1024_random_rows	get_batch_of_1024_rows	get_first_row	get_last_row
new / old (diff)	0.288355 / 0.018006 (0.270349)	0.571481 / 0.000490 (0.570991)	0.006765 / 0.000200 (0.006565)	0.000101 / 0.000054 (0.000047)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.031836 / 0.037411 (-0.005575)	0.101312 / 0.014526 (0.086786)	0.111433 / 0.176557 (-0.065124)	0.169599 / 0.737135 (-0.567536)	0.114595 / 0.296338 (-0.181743)

Benchmark: benchmark_iterating.json

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
new / old (diff)	0.645258 / 0.215209 (0.430049)	6.446653 / 2.077655 (4.368998)	2.983498 / 1.504120 (1.479379)	2.573820 / 1.541195 (1.032625)	2.624286 / 1.468490 (1.155796)	0.815997 / 4.584777 (-3.768780)	5.140248 / 3.745712 (1.394536)	4.636915 / 5.269862 (-0.632947)	2.866313 / 4.565676 (-1.699364)	0.096643 / 0.424275 (-0.327633)	0.008452 / 0.007607 (0.000845)	0.765837 / 0.226044 (0.539793)	7.622897 / 2.268929 (5.353968)	3.796247 / 55.444624 (-51.648378)	3.019349 / 6.876477 (-3.857128)	3.034187 / 2.142072 (0.892115)	1.001682 / 4.805227 (-3.803546)	0.211841 / 6.500664 (-6.288823)	0.073351 / 0.075469 (-0.002119)

Benchmark: benchmark_map_filter.json

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
new / old (diff)	1.740254 / 1.841788 (-0.101534)	23.465619 / 8.074308 (15.391311)	21.651670 / 10.191392 (11.460278)	0.226129 / 0.680424 (-0.454294)	0.029611 / 0.534201 (-0.504590)	0.441140 / 0.579283 (-0.138143)	0.605591 / 0.434364 (0.171227)	0.552427 / 0.540337 (0.012090)	0.771975 / 1.386936 (-0.614961)

[mariosasko]

Thanks a lot for working on this!

[mariosasko]

The CI errors seem unrelated to this PR but I think they need further investigation in another PR.

FAILED tests/test_upstream_hub.py::TestPushToHub::test_push_dataset_dict_to_hub_multiple_files - KeyError: 'url'

We need to wait for huggingface_hub's next release to fix this (see huggingface/huggingface_hub#1675; 409 error is currently ignored, hence the KeyError)

Also, we should be able to fix test_push_dataset_dict_to_hub_overwrite_files by inserting gc.collect() (to drop the "reference" to an Arrow file) between the load_dataset calls to avoid the PermissionError (also reported in #3139)

(Indeed, this can be addressed in subsequent PRs.)

[github-actions]

Show benchmarks

PyArrow==8.0.0

Show updated benchmarks!

Benchmark: benchmark_array_xd.json

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
new / old (diff)	0.008988 / 0.011353 (-0.002365)	0.005270 / 0.011008 (-0.005738)	0.114577 / 0.038508 (0.076068)	0.091630 / 0.023109 (0.068521)	0.409217 / 0.275898 (0.133319)	0.440903 / 0.323480 (0.117424)	0.005226 / 0.007986 (-0.002760)	0.004289 / 0.004328 (-0.000040)	0.082246 / 0.004250 (0.077995)	0.084926 / 0.037052 (0.047873)	0.407822 / 0.258489 (0.149333)	0.440891 / 0.293841 (0.147051)	0.052225 / 0.128546 (-0.076321)	0.014218 / 0.075646 (-0.061429)	0.436994 / 0.419271 (0.017722)	0.066433 / 0.043533 (0.022901)	0.413909 / 0.255139 (0.158770)	0.425729 / 0.283200 (0.142530)	0.039576 / 0.141683 (-0.102107)	1.905604 / 1.452155 (0.453449)	1.907032 / 1.492716 (0.414315)

Benchmark: benchmark_getitem_100B.json

metric	get_batch_of_1024_random_rows	get_batch_of_1024_rows	get_first_row	get_last_row
new / old (diff)	0.313662 / 0.018006 (0.295655)	0.614541 / 0.000490 (0.614051)	0.015631 / 0.000200 (0.015431)	0.000507 / 0.000054 (0.000453)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.029049 / 0.037411 (-0.008362)	0.094626 / 0.014526 (0.080100)	0.104718 / 0.176557 (-0.071838)	0.187346 / 0.737135 (-0.549790)	0.108001 / 0.296338 (-0.188337)

Benchmark: benchmark_iterating.json

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
new / old (diff)	0.578997 / 0.215209 (0.363788)	5.815546 / 2.077655 (3.737892)	2.411301 / 1.504120 (0.907181)	2.110088 / 1.541195 (0.568893)	2.147839 / 1.468490 (0.679349)	0.861285 / 4.584777 (-3.723492)	5.264245 / 3.745712 (1.518533)	4.695786 / 5.269862 (-0.574076)	2.867522 / 4.565676 (-1.698154)	0.096523 / 0.424275 (-0.327752)	0.008777 / 0.007607 (0.001170)	0.716316 / 0.226044 (0.490272)	7.257574 / 2.268929 (4.988645)	3.141502 / 55.444624 (-52.303123)	2.480604 / 6.876477 (-4.395872)	2.530031 / 2.142072 (0.387958)	1.054274 / 4.805227 (-3.750953)	0.210781 / 6.500664 (-6.289883)	0.073837 / 0.075469 (-0.001632)

Benchmark: benchmark_map_filter.json

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
new / old (diff)	1.607689 / 1.841788 (-0.234099)	23.856780 / 8.074308 (15.782472)	19.507196 / 10.191392 (9.315804)	0.232712 / 0.680424 (-0.447712)	0.027037 / 0.534201 (-0.507164)	0.466613 / 0.579283 (-0.112670)	0.571139 / 0.434364 (0.136775)	0.543109 / 0.540337 (0.002771)	0.785558 / 1.386936 (-0.601378)

PyArrow==latest

Show updated benchmarks!

Benchmark: benchmark_array_xd.json

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
new / old (diff)	0.008104 / 0.011353 (-0.003249)	0.004923 / 0.011008 (-0.006086)	0.075093 / 0.038508 (0.036585)	0.075218 / 0.023109 (0.052109)	0.476615 / 0.275898 (0.200717)	0.506984 / 0.323480 (0.183504)	0.006371 / 0.007986 (-0.001614)	0.004818 / 0.004328 (0.000489)	0.075634 / 0.004250 (0.071383)	0.059513 / 0.037052 (0.022461)	0.523763 / 0.258489 (0.265274)	0.531858 / 0.293841 (0.238017)	0.048168 / 0.128546 (-0.080379)	0.014110 / 0.075646 (-0.061537)	0.086052 / 0.419271 (-0.333219)	0.058369 / 0.043533 (0.014836)	0.475537 / 0.255139 (0.220398)	0.509429 / 0.283200 (0.226229)	0.033924 / 0.141683 (-0.107758)	1.657490 / 1.452155 (0.205336)	1.762544 / 1.492716 (0.269828)

Benchmark: benchmark_getitem_100B.json

metric	get_batch_of_1024_random_rows	get_batch_of_1024_rows	get_first_row	get_last_row
new / old (diff)	0.263863 / 0.018006 (0.245857)	0.584468 / 0.000490 (0.583978)	0.007063 / 0.000200 (0.006863)	0.000181 / 0.000054 (0.000126)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.032229 / 0.037411 (-0.005183)	0.096750 / 0.014526 (0.082224)	0.117798 / 0.176557 (-0.058758)	0.173376 / 0.737135 (-0.563760)	0.117241 / 0.296338 (-0.179098)

Benchmark: benchmark_iterating.json

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
new / old (diff)	0.701935 / 0.215209 (0.486726)	6.544655 / 2.077655 (4.467001)	3.055531 / 1.504120 (1.551411)	2.896339 / 1.541195 (1.355144)	3.013157 / 1.468490 (1.544667)	0.852989 / 4.584777 (-3.731788)	5.399355 / 3.745712 (1.653643)	5.119811 / 5.269862 (-0.150051)	3.167269 / 4.565676 (-1.398407)	0.096962 / 0.424275 (-0.327313)	0.008843 / 0.007607 (0.001236)	0.776170 / 0.226044 (0.550125)	7.735093 / 2.268929 (5.466164)	3.792629 / 55.444624 (-51.651996)	3.249911 / 6.876477 (-3.626565)	3.235590 / 2.142072 (1.093517)	1.046426 / 4.805227 (-3.758801)	0.239854 / 6.500664 (-6.260810)	0.100648 / 0.075469 (0.025179)

Benchmark: benchmark_map_filter.json

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
new / old (diff)	1.774488 / 1.841788 (-0.067300)	25.646958 / 8.074308 (17.572650)	23.181577 / 10.191392 (12.990185)	0.231948 / 0.680424 (-0.448476)	0.030147 / 0.534201 (-0.504054)	0.464161 / 0.579283 (-0.115122)	0.598980 / 0.434364 (0.164616)	0.571156 / 0.540337 (0.030819)	0.833221 / 1.386936 (-0.553715)

[mariosasko]

LGTM!

[github-actions]

Show benchmarks

PyArrow==8.0.0

Show updated benchmarks!

Benchmark: benchmark_array_xd.json

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
new / old (diff)	0.006010 / 0.011353 (-0.005343)	0.003662 / 0.011008 (-0.007346)	0.079971 / 0.038508 (0.041463)	0.066790 / 0.023109 (0.043681)	0.311387 / 0.275898 (0.035489)	0.346781 / 0.323480 (0.023301)	0.003500 / 0.007986 (-0.004485)	0.002831 / 0.004328 (-0.001498)	0.063238 / 0.004250 (0.058988)	0.056163 / 0.037052 (0.019110)	0.317456 / 0.258489 (0.058967)	0.356106 / 0.293841 (0.062265)	0.027358 / 0.128546 (-0.101188)	0.007906 / 0.075646 (-0.067741)	0.261779 / 0.419271 (-0.157492)	0.046385 / 0.043533 (0.002852)	0.312587 / 0.255139 (0.057448)	0.339513 / 0.283200 (0.056314)	0.021474 / 0.141683 (-0.120209)	1.418637 / 1.452155 (-0.033518)	1.510257 / 1.492716 (0.017540)

Benchmark: benchmark_getitem_100B.json

metric	get_batch_of_1024_random_rows	get_batch_of_1024_rows	get_first_row	get_last_row
new / old (diff)	0.211761 / 0.018006 (0.193755)	0.424387 / 0.000490 (0.423898)	0.002579 / 0.000200 (0.002379)	0.000065 / 0.000054 (0.000011)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.024038 / 0.037411 (-0.013374)	0.072524 / 0.014526 (0.057998)	0.083443 / 0.176557 (-0.093113)	0.144835 / 0.737135 (-0.592300)	0.084754 / 0.296338 (-0.211585)

Benchmark: benchmark_iterating.json

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
new / old (diff)	0.392423 / 0.215209 (0.177214)	3.927220 / 2.077655 (1.849565)	1.877853 / 1.504120 (0.373733)	1.699275 / 1.541195 (0.158081)	1.793144 / 1.468490 (0.324654)	0.503809 / 4.584777 (-4.080968)	3.052569 / 3.745712 (-0.693143)	2.907432 / 5.269862 (-2.362429)	1.811220 / 4.565676 (-2.754457)	0.057249 / 0.424275 (-0.367026)	0.006433 / 0.007607 (-0.001174)	0.463257 / 0.226044 (0.237213)	4.631038 / 2.268929 (2.362109)	2.315870 / 55.444624 (-53.128754)	2.000476 / 6.876477 (-4.876001)	2.043581 / 2.142072 (-0.098492)	0.588911 / 4.805227 (-4.216317)	0.125370 / 6.500664 (-6.375295)	0.061721 / 0.075469 (-0.013748)

Benchmark: benchmark_map_filter.json

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
new / old (diff)	1.244486 / 1.841788 (-0.597301)	17.862422 / 8.074308 (9.788114)	13.890205 / 10.191392 (3.698813)	0.145467 / 0.680424 (-0.534957)	0.016856 / 0.534201 (-0.517345)	0.329357 / 0.579283 (-0.249926)	0.367550 / 0.434364 (-0.066814)	0.377541 / 0.540337 (-0.162796)	0.534087 / 1.386936 (-0.852849)

PyArrow==latest

Show updated benchmarks!

Benchmark: benchmark_array_xd.json

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
new / old (diff)	0.006030 / 0.011353 (-0.005323)	0.003650 / 0.011008 (-0.007359)	0.063300 / 0.038508 (0.024792)	0.058877 / 0.023109 (0.035767)	0.454662 / 0.275898 (0.178764)	0.489362 / 0.323480 (0.165882)	0.004856 / 0.007986 (-0.003130)	0.002909 / 0.004328 (-0.001420)	0.063356 / 0.004250 (0.059105)	0.047867 / 0.037052 (0.010814)	0.465461 / 0.258489 (0.206972)	0.506684 / 0.293841 (0.212843)	0.028599 / 0.128546 (-0.099947)	0.008076 / 0.075646 (-0.067570)	0.068695 / 0.419271 (-0.350576)	0.041487 / 0.043533 (-0.002045)	0.448676 / 0.255139 (0.193537)	0.471206 / 0.283200 (0.188007)	0.020401 / 0.141683 (-0.121282)	1.461181 / 1.452155 (0.009026)	1.517079 / 1.492716 (0.024363)

Benchmark: benchmark_getitem_100B.json

metric	get_batch_of_1024_random_rows	get_batch_of_1024_rows	get_first_row	get_last_row
new / old (diff)	0.222827 / 0.018006 (0.204821)	0.425074 / 0.000490 (0.424585)	0.004153 / 0.000200 (0.003953)	0.000081 / 0.000054 (0.000026)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.026980 / 0.037411 (-0.010431)	0.080786 / 0.014526 (0.066260)	0.092040 / 0.176557 (-0.084517)	0.146082 / 0.737135 (-0.591053)	0.092739 / 0.296338 (-0.203600)

Benchmark: benchmark_iterating.json

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
new / old (diff)	0.461663 / 0.215209 (0.246454)	4.604828 / 2.077655 (2.527173)	2.566926 / 1.504120 (1.062806)	2.394419 / 1.541195 (0.853224)	2.458375 / 1.468490 (0.989885)	0.505140 / 4.584777 (-4.079637)	3.155916 / 3.745712 (-0.589796)	3.014474 / 5.269862 (-2.255388)	1.900296 / 4.565676 (-2.665380)	0.058063 / 0.424275 (-0.366212)	0.006409 / 0.007607 (-0.001198)	0.541165 / 0.226044 (0.315120)	5.410700 / 2.268929 (3.141772)	3.010239 / 55.444624 (-52.434386)	2.668103 / 6.876477 (-4.208373)	2.730418 / 2.142072 (0.588346)	0.603471 / 4.805227 (-4.201756)	0.129852 / 6.500664 (-6.370812)	0.061507 / 0.075469 (-0.013962)

Benchmark: benchmark_map_filter.json

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
new / old (diff)	1.355272 / 1.841788 (-0.486516)	18.170088 / 8.074308 (10.095780)	15.583855 / 10.191392 (5.392463)	0.146246 / 0.680424 (-0.534178)	0.018093 / 0.534201 (-0.516108)	0.331695 / 0.579283 (-0.247588)	0.380845 / 0.434364 (-0.053519)	0.388564 / 0.540337 (-0.151774)	0.551465 / 1.386936 (-0.835471)