Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Remove apache_beam import in BeamBasedBuilder._save_info #6265

Merged
merged 5 commits into from
Sep 28, 2023
Merged

Remove apache_beam import in BeamBasedBuilder._save_info #6265

merged 5 commits into from
Sep 28, 2023

Conversation

mariosasko
Copy link
Collaborator

@mariosasko mariosasko commented Sep 27, 2023 •
edited
Loading

... to avoid an ImportError raised in BeamBasedBuilder._save_info when apache_beam is not installed (e.g., when downloading the processed version of a dataset from the HF GCS)

Fix #6260

@HuggingFaceDocBuilderDev
Copy link

HuggingFaceDocBuilderDev commented Sep 27, 2023 •
edited
Loading

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

@github-actions
Copy link

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) 0.005896 / 0.011353 (-0.005457) 0.003642 / 0.011008 (-0.007366) 0.081917 / 0.038508 (0.043409) 0.059513 / 0.023109 (0.036404) 0.341422 / 0.275898 (0.065524) 0.359278 / 0.323480 (0.035798) 0.004707 / 0.007986 (-0.003279) 0.002938 / 0.004328 (-0.001390) 0.063095 / 0.004250 (0.058845) 0.051777 / 0.037052 (0.014725) 0.321114 / 0.258489 (0.062625) 0.363823 / 0.293841 (0.069982) 0.027590 / 0.128546 (-0.100957) 0.007846 / 0.075646 (-0.067800) 0.261197 / 0.419271 (-0.158074) 0.045812 / 0.043533 (0.002279) 0.319787 / 0.255139 (0.064648) 0.341839 / 0.283200 (0.058640) 0.021913 / 0.141683 (-0.119770) 1.397525 / 1.452155 (-0.054630) 1.495902 / 1.492716 (0.003186)

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) 0.224815 / 0.018006 (0.206809) 0.425780 / 0.000490 (0.425290) 0.006934 / 0.000200 (0.006734) 0.000225 / 0.000054 (0.000171)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.024342 / 0.037411 (-0.013070) 0.073923 / 0.014526 (0.059398) 0.082108 / 0.176557 (-0.094448) 0.143017 / 0.737135 (-0.594119) 0.083163 / 0.296338 (-0.213175)

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) 0.398244 / 0.215209 (0.183035) 3.957688 / 2.077655 (1.880033) 1.904615 / 1.504120 (0.400495) 1.710353 / 1.541195 (0.169158) 1.798980 / 1.468490 (0.330490) 0.499307 / 4.584777 (-4.085470) 3.026734 / 3.745712 (-0.718978) 2.923940 / 5.269862 (-2.345922) 1.831870 / 4.565676 (-2.733807) 0.058551 / 0.424275 (-0.365724) 0.006403 / 0.007607 (-0.001204) 0.464164 / 0.226044 (0.238119) 4.644556 / 2.268929 (2.375628) 2.341455 / 55.444624 (-53.103169) 2.004385 / 6.876477 (-4.872092) 2.051819 / 2.142072 (-0.090253) 0.585610 / 4.805227 (-4.219617) 0.124735 / 6.500664 (-6.375929) 0.061150 / 0.075469 (-0.014319)

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) 1.224665 / 1.841788 (-0.617122) 17.476227 / 8.074308 (9.401919) 13.867617 / 10.191392 (3.676225) 0.144177 / 0.680424 (-0.536247) 0.017045 / 0.534201 (-0.517156) 0.337468 / 0.579283 (-0.241815) 0.374476 / 0.434364 (-0.059888) 0.393428 / 0.540337 (-0.146910) 0.535335 / 1.386936 (-0.851601)
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) 0.006208 / 0.011353 (-0.005145) 0.003650 / 0.011008 (-0.007359) 0.062843 / 0.038508 (0.024335) 0.062272 / 0.023109 (0.039162) 0.446336 / 0.275898 (0.170438) 0.477476 / 0.323480 (0.153996) 0.004862 / 0.007986 (-0.003124) 0.002822 / 0.004328 (-0.001506) 0.063427 / 0.004250 (0.059177) 0.049023 / 0.037052 (0.011971) 0.453633 / 0.258489 (0.195144) 0.486494 / 0.293841 (0.192653) 0.028634 / 0.128546 (-0.099912) 0.008187 / 0.075646 (-0.067460) 0.068846 / 0.419271 (-0.350425) 0.041104 / 0.043533 (-0.002429) 0.446646 / 0.255139 (0.191507) 0.468860 / 0.283200 (0.185660) 0.020980 / 0.141683 (-0.120703) 1.455565 / 1.452155 (0.003410) 1.511142 / 1.492716 (0.018426)

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) 0.224242 / 0.018006 (0.206236) 0.408483 / 0.000490 (0.407993) 0.003495 / 0.000200 (0.003296) 0.000076 / 0.000054 (0.000022)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.027286 / 0.037411 (-0.010125) 0.081151 / 0.014526 (0.066625) 0.096598 / 0.176557 (-0.079959) 0.146193 / 0.737135 (-0.590942) 0.092213 / 0.296338 (-0.204125)

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) 0.463837 / 0.215209 (0.248628) 4.636820 / 2.077655 (2.559165) 2.576100 / 1.504120 (1.071980) 2.396974 / 1.541195 (0.855779) 2.461526 / 1.468490 (0.993036) 0.502360 / 4.584777 (-4.082417) 3.099973 / 3.745712 (-0.645739) 2.937260 / 5.269862 (-2.332602) 1.871274 / 4.565676 (-2.694402) 0.057913 / 0.424275 (-0.366362) 0.006511 / 0.007607 (-0.001096) 0.536917 / 0.226044 (0.310873) 5.396966 / 2.268929 (3.128038) 3.015646 / 55.444624 (-52.428978) 2.673793 / 6.876477 (-4.202684) 2.712376 / 2.142072 (0.570304) 0.591632 / 4.805227 (-4.213595) 0.124872 / 6.500664 (-6.375792) 0.061820 / 0.075469 (-0.013649)

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) 1.356828 / 1.841788 (-0.484960) 18.076995 / 8.074308 (10.002687) 15.116482 / 10.191392 (4.925090) 0.151375 / 0.680424 (-0.529049) 0.017867 / 0.534201 (-0.516334) 0.335012 / 0.579283 (-0.244271) 0.384137 / 0.434364 (-0.050226) 0.397792 / 0.540337 (-0.142546) 0.551521 / 1.386936 (-0.835415)

@github-actions
Copy link

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) 0.009418 / 0.011353 (-0.001935) 0.005186 / 0.011008 (-0.005822) 0.112270 / 0.038508 (0.073761) 0.114856 / 0.023109 (0.091747) 0.402267 / 0.275898 (0.126369) 0.445213 / 0.323480 (0.121733) 0.005588 / 0.007986 (-0.002398) 0.004315 / 0.004328 (-0.000013) 0.083561 / 0.004250 (0.079311) 0.087319 / 0.037052 (0.050267) 0.400989 / 0.258489 (0.142500) 0.455636 / 0.293841 (0.161795) 0.045168 / 0.128546 (-0.083378) 0.010939 / 0.075646 (-0.064707) 0.400120 / 0.419271 (-0.019151) 0.071599 / 0.043533 (0.028066) 0.418112 / 0.255139 (0.162973) 0.443889 / 0.283200 (0.160690) 0.032433 / 0.141683 (-0.109250) 1.886313 / 1.452155 (0.434159) 2.012909 / 1.492716 (0.520193)

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) 0.306991 / 0.018006 (0.288985) 0.590426 / 0.000490 (0.589937) 0.011811 / 0.000200 (0.011611) 0.000596 / 0.000054 (0.000542)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.042520 / 0.037411 (0.005108) 0.129808 / 0.014526 (0.115283) 0.125481 / 0.176557 (-0.051075) 0.199181 / 0.737135 (-0.537954) 0.130426 / 0.296338 (-0.165913)

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) 0.526455 / 0.215209 (0.311246) 5.213304 / 2.077655 (3.135649) 2.643406 / 1.504120 (1.139286) 2.611214 / 1.541195 (1.070019) 2.586730 / 1.468490 (1.118240) 0.639103 / 4.584777 (-3.945674) 5.197421 / 3.745712 (1.451709) 4.634642 / 5.269862 (-0.635220) 2.741079 / 4.565676 (-1.824598) 0.073064 / 0.424275 (-0.351211) 0.009441 / 0.007607 (0.001834) 0.635984 / 0.226044 (0.409940) 6.283268 / 2.268929 (4.014339) 3.337205 / 55.444624 (-52.107419) 3.192362 / 6.876477 (-3.684114) 2.910367 / 2.142072 (0.768294) 0.767937 / 4.805227 (-4.037290) 0.177467 / 6.500664 (-6.323198) 0.081162 / 0.075469 (0.005693)

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) 1.803717 / 1.841788 (-0.038071) 26.823235 / 8.074308 (18.748927) 19.714471 / 10.191392 (9.523079) 0.204048 / 0.680424 (-0.476376) 0.025992 / 0.534201 (-0.508209) 0.521438 / 0.579283 (-0.057845) 0.596524 / 0.434364 (0.162160) 0.600763 / 0.540337 (0.060425) 0.945971 / 1.386936 (-0.440965)
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) 0.009126 / 0.011353 (-0.002226) 0.005109 / 0.011008 (-0.005899) 0.083046 / 0.038508 (0.044538) 0.115930 / 0.023109 (0.092821) 0.534311 / 0.275898 (0.258413) 0.552846 / 0.323480 (0.229366) 0.007240 / 0.007986 (-0.000746) 0.004617 / 0.004328 (0.000289) 0.083927 / 0.004250 (0.079676) 0.075926 / 0.037052 (0.038873) 0.534750 / 0.258489 (0.276261) 0.575122 / 0.293841 (0.281281) 0.041001 / 0.128546 (-0.087545) 0.010851 / 0.075646 (-0.064795) 0.096574 / 0.419271 (-0.322697) 0.063533 / 0.043533 (0.020001) 0.546850 / 0.255139 (0.291711) 0.547122 / 0.283200 (0.263922) 0.032437 / 0.141683 (-0.109245) 1.926191 / 1.452155 (0.474036) 2.029841 / 1.492716 (0.537125)

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) 0.275582 / 0.018006 (0.257576) 0.574212 / 0.000490 (0.573722) 0.006863 / 0.000200 (0.006663) 0.000236 / 0.000054 (0.000181)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.045340 / 0.037411 (0.007928) 0.129196 / 0.014526 (0.114670) 0.136637 / 0.176557 (-0.039920) 0.200040 / 0.737135 (-0.537096) 0.136328 / 0.296338 (-0.160011)

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) 0.612379 / 0.215209 (0.397170) 5.874664 / 2.077655 (3.797010) 3.070626 / 1.504120 (1.566506) 2.999319 / 1.541195 (1.458124) 3.000571 / 1.468490 (1.532081) 0.732119 / 4.584777 (-3.852658) 5.193226 / 3.745712 (1.447514) 4.714571 / 5.269862 (-0.555291) 2.870438 / 4.565676 (-1.695239) 0.075793 / 0.424275 (-0.348482) 0.009238 / 0.007607 (0.001631) 0.695192 / 0.226044 (0.469148) 6.897996 / 2.268929 (4.629067) 3.923474 / 55.444624 (-51.521150) 3.458326 / 6.876477 (-3.418151) 3.331652 / 2.142072 (1.189579) 0.821132 / 4.805227 (-3.984095) 0.182252 / 6.500664 (-6.318412) 0.084730 / 0.075469 (0.009260)

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) 1.919861 / 1.841788 (0.078073) 27.437228 / 8.074308 (19.362920) 21.109899 / 10.191392 (10.918507) 0.245998 / 0.680424 (-0.434426) 0.025817 / 0.534201 (-0.508384) 0.517757 / 0.579283 (-0.061526) 0.576375 / 0.434364 (0.142011) 0.625283 / 0.540337 (0.084945) 0.956877 / 1.386936 (-0.430059)

lhoestq
lhoestq approved these changes Sep 28, 2023
View reviewed changes
Copy link
Member

@lhoestq lhoestq left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Cool ! It should get the same credentials as the beam filesystem via environment variables or using credentials files on disk so we're fine

@mariosasko mariosasko marked this pull request as ready for review September 28, 2023 18:23
@mariosasko mariosasko merged commit 0cc77d7 into main Sep 28, 2023
9 of 13 checks passed
@mariosasko mariosasko deleted the remove-beam-import branch September 28, 2023 18:23
@github-actions
Copy link

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) 0.008099 / 0.011353 (-0.003254) 0.004815 / 0.011008 (-0.006194) 0.099657 / 0.038508 (0.061149) 0.064737 / 0.023109 (0.041628) 0.461773 / 0.275898 (0.185875) 0.444810 / 0.323480 (0.121330) 0.004247 / 0.007986 (-0.003739) 0.004956 / 0.004328 (0.000628) 0.068664 / 0.004250 (0.064414) 0.052039 / 0.037052 (0.014986) 0.406750 / 0.258489 (0.148261) 0.452832 / 0.293841 (0.158991) 0.044518 / 0.128546 (-0.084028) 0.013220 / 0.075646 (-0.062426) 0.317713 / 0.419271 (-0.101558) 0.061897 / 0.043533 (0.018364) 0.398664 / 0.255139 (0.143525) 0.531494 / 0.283200 (0.248294) 0.064033 / 0.141683 (-0.077650) 1.590385 / 1.452155 (0.138231) 1.769918 / 1.492716 (0.277202)

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) 0.230795 / 0.018006 (0.212789) 0.568797 / 0.000490 (0.568308) 0.013498 / 0.000200 (0.013298) 0.000448 / 0.000054 (0.000393)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.028394 / 0.037411 (-0.009017) 0.081973 / 0.014526 (0.067447) 0.097623 / 0.176557 (-0.078934) 0.158691 / 0.737135 (-0.578445) 0.101548 / 0.296338 (-0.194791)

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) 0.574459 / 0.215209 (0.359249) 5.709871 / 2.077655 (3.632217) 2.521460 / 1.504120 (1.017340) 2.239463 / 1.541195 (0.698268) 2.195067 / 1.468490 (0.726577) 0.792390 / 4.584777 (-3.792387) 4.841665 / 3.745712 (1.095952) 4.201620 / 5.269862 (-1.068241) 2.664081 / 4.565676 (-1.901595) 0.097661 / 0.424275 (-0.326614) 0.008428 / 0.007607 (0.000821) 0.698729 / 0.226044 (0.472684) 6.908867 / 2.268929 (4.639939) 3.247480 / 55.444624 (-52.197145) 2.563921 / 6.876477 (-4.312556) 2.738249 / 2.142072 (0.596177) 0.972066 / 4.805227 (-3.833161) 0.191196 / 6.500664 (-6.309468) 0.064732 / 0.075469 (-0.010737)

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) 1.421910 / 1.841788 (-0.419877) 20.633538 / 8.074308 (12.559230) 18.054562 / 10.191392 (7.863170) 0.194125 / 0.680424 (-0.486299) 0.028097 / 0.534201 (-0.506104) 0.417857 / 0.579283 (-0.161426) 0.518758 / 0.434364 (0.084394) 0.500199 / 0.540337 (-0.040138) 0.754662 / 1.386936 (-0.632274)
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) 0.008452 / 0.011353 (-0.002901) 0.004646 / 0.011008 (-0.006362) 0.077286 / 0.038508 (0.038778) 0.072507 / 0.023109 (0.049398) 0.439580 / 0.275898 (0.163682) 0.506166 / 0.323480 (0.182686) 0.006035 / 0.007986 (-0.001950) 0.003886 / 0.004328 (-0.000442) 0.075091 / 0.004250 (0.070841) 0.063163 / 0.037052 (0.026110) 0.468550 / 0.258489 (0.210061) 0.523273 / 0.293841 (0.229432) 0.048728 / 0.128546 (-0.079818) 0.012991 / 0.075646 (-0.062655) 0.087964 / 0.419271 (-0.331308) 0.058920 / 0.043533 (0.015387) 0.451247 / 0.255139 (0.196108) 0.489827 / 0.283200 (0.206628) 0.031164 / 0.141683 (-0.110519) 1.675504 / 1.452155 (0.223349) 1.806098 / 1.492716 (0.313382)

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) 0.253567 / 0.018006 (0.235561) 0.508971 / 0.000490 (0.508481) 0.010882 / 0.000200 (0.010682) 0.000111 / 0.000054 (0.000057)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.029490 / 0.037411 (-0.007921) 0.090255 / 0.014526 (0.075729) 0.110075 / 0.176557 (-0.066482) 0.159375 / 0.737135 (-0.577760) 0.109313 / 0.296338 (-0.187025)

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) 0.580252 / 0.215209 (0.365043) 5.911741 / 2.077655 (3.834086) 2.659405 / 1.504120 (1.155285) 2.344943 / 1.541195 (0.803749) 2.390748 / 1.468490 (0.922258) 0.827823 / 4.584777 (-3.756954) 4.973544 / 3.745712 (1.227832) 4.300220 / 5.269862 (-0.969642) 2.826181 / 4.565676 (-1.739495) 0.101013 / 0.424275 (-0.323263) 0.008025 / 0.007607 (0.000418) 0.728414 / 0.226044 (0.502369) 7.508045 / 2.268929 (5.239117) 3.687627 / 55.444624 (-51.756997) 2.902953 / 6.876477 (-3.973524) 3.094624 / 2.142072 (0.952551) 1.054696 / 4.805227 (-3.750531) 0.212297 / 6.500664 (-6.288367) 0.070211 / 0.075469 (-0.005258)

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) 1.567117 / 1.841788 (-0.274670) 21.420746 / 8.074308 (13.346438) 19.857467 / 10.191392 (9.666075) 0.228554 / 0.680424 (-0.451870) 0.032278 / 0.534201 (-0.501923) 0.459966 / 0.579283 (-0.119317) 0.541219 / 0.434364 (0.106855) 0.549599 / 0.540337 (0.009261) 0.731476 / 1.386936 (-0.655460)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@lhoestq lhoestq lhoestq approved these changes

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.

REUSE_DATASET_IF_EXISTS don't work
3 participants
@mariosasko @HuggingFaceDocBuilderDev @lhoestq