Skip to content

Add trust_remote_code argument #6429

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
Merged
merged 14 commits into from
Nov 28, 2023
Merged

Add trust_remote_code argument #6429

merged 14 commits into from
Nov 28, 2023

Conversation

@lhoestq
Copy link
Member

@lhoestq lhoestq commented Nov 16, 2023 •
edited
Loading

Draft about adding trust_remote_code to load_dataset.

ds = load_dataset(..., trust_remote_code=True)  # run remote code (current default)

It would default to True (current behavior) and in the next major release it will prompt the user to check the code before running it (we'll communicate on this before doing it of course).

# in the future
ds = load_dataset(...)  # prompt the user to check the code before running it (future default)
ds = load_dataset(..., trust_remote_code=True)  # run remote code
ds = load_dataset(..., trust_remote_code=False)  # disallow remote code

Related to #6400

Will do a separate PR to use the parquet export when possible

@github-actions
Copy link

github-actions bot commented Nov 16, 2023

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) 0.004947 / 0.011353 (-0.006405) 0.002961 / 0.011008 (-0.008047) 0.063474 / 0.038508 (0.024966) 0.030162 / 0.023109 (0.007053) 0.232388 / 0.275898 (-0.043511) 0.257654 / 0.323480 (-0.065826) 0.002969 / 0.007986 (-0.005017) 0.002336 / 0.004328 (-0.001993) 0.049724 / 0.004250 (0.045473) 0.045608 / 0.037052 (0.008555) 0.236079 / 0.258489 (-0.022410) 0.267809 / 0.293841 (-0.026032) 0.023805 / 0.128546 (-0.104741) 0.007177 / 0.075646 (-0.068470) 0.202167 / 0.419271 (-0.217104) 0.056181 / 0.043533 (0.012648) 0.256464 / 0.255139 (0.001325) 0.271908 / 0.283200 (-0.011292) 0.020211 / 0.141683 (-0.121472) 1.114112 / 1.452155 (-0.338042) 1.174879 / 1.492716 (-0.317837)

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) 0.093457 / 0.018006 (0.075451) 0.307643 / 0.000490 (0.307154) 0.000212 / 0.000200 (0.000012) 0.000047 / 0.000054 (-0.000008)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.018635 / 0.037411 (-0.018777) 0.062099 / 0.014526 (0.047573) 0.073619 / 0.176557 (-0.102938) 0.119986 / 0.737135 (-0.617149) 0.075439 / 0.296338 (-0.220899)

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) 0.280142 / 0.215209 (0.064933) 2.733790 / 2.077655 (0.656136) 1.457633 / 1.504120 (-0.046487) 1.336288 / 1.541195 (-0.204907) 1.363191 / 1.468490 (-0.105299) 0.399331 / 4.584777 (-4.185446) 2.343099 / 3.745712 (-1.402614) 2.617059 / 5.269862 (-2.652802) 1.575912 / 4.565676 (-2.989765) 0.045621 / 0.424275 (-0.378655) 0.004825 / 0.007607 (-0.002782) 0.346669 / 0.226044 (0.120625) 3.225982 / 2.268929 (0.957054) 1.787067 / 55.444624 (-53.657557) 1.503883 / 6.876477 (-5.372593) 1.527593 / 2.142072 (-0.614479) 0.466806 / 4.805227 (-4.338421) 0.098537 / 6.500664 (-6.402127) 0.042028 / 0.075469 (-0.033441)

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) 0.945040 / 1.841788 (-0.896748) 11.970022 / 8.074308 (3.895714) 10.261176 / 10.191392 (0.069784) 0.138231 / 0.680424 (-0.542193) 0.013933 / 0.534201 (-0.520268) 0.270640 / 0.579283 (-0.308643) 0.263185 / 0.434364 (-0.171178) 0.306686 / 0.540337 (-0.233651) 0.423164 / 1.386936 (-0.963772)
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) 0.004765 / 0.011353 (-0.006588) 0.003158 / 0.011008 (-0.007850) 0.047813 / 0.038508 (0.009305) 0.053363 / 0.023109 (0.030254) 0.278570 / 0.275898 (0.002671) 0.291500 / 0.323480 (-0.031980) 0.003987 / 0.007986 (-0.003998) 0.002430 / 0.004328 (-0.001898) 0.048059 / 0.004250 (0.043809) 0.038595 / 0.037052 (0.001542) 0.276383 / 0.258489 (0.017894) 0.304234 / 0.293841 (0.010393) 0.024402 / 0.128546 (-0.104144) 0.007303 / 0.075646 (-0.068343) 0.055091 / 0.419271 (-0.364180) 0.032735 / 0.043533 (-0.010797) 0.270905 / 0.255139 (0.015766) 0.287181 / 0.283200 (0.003981) 0.018919 / 0.141683 (-0.122764) 1.153814 / 1.452155 (-0.298341) 1.197009 / 1.492716 (-0.295707)

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) 0.093743 / 0.018006 (0.075737) 0.302877 / 0.000490 (0.302387) 0.000223 / 0.000200 (0.000023) 0.000052 / 0.000054 (-0.000002)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.021279 / 0.037411 (-0.016133) 0.070886 / 0.014526 (0.056360) 0.081628 / 0.176557 (-0.094928) 0.119721 / 0.737135 (-0.617414) 0.083093 / 0.296338 (-0.213245)

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) 0.297788 / 0.215209 (0.082579) 2.915235 / 2.077655 (0.837580) 1.587580 / 1.504120 (0.083460) 1.461699 / 1.541195 (-0.079495) 1.520609 / 1.468490 (0.052119) 0.398363 / 4.584777 (-4.186413) 2.408415 / 3.745712 (-1.337297) 2.552776 / 5.269862 (-2.717086) 1.508219 / 4.565676 (-3.057457) 0.045884 / 0.424275 (-0.378391) 0.004842 / 0.007607 (-0.002765) 0.341376 / 0.226044 (0.115331) 3.420192 / 2.268929 (1.151264) 1.974938 / 55.444624 (-53.469686) 1.678283 / 6.876477 (-5.198194) 1.702439 / 2.142072 (-0.439633) 0.467056 / 4.805227 (-4.338172) 0.098684 / 6.500664 (-6.401980) 0.041052 / 0.075469 (-0.034417)

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) 0.990145 / 1.841788 (-0.851643) 12.143198 / 8.074308 (4.068890) 10.911039 / 10.191392 (0.719647) 0.130384 / 0.680424 (-0.550040) 0.015602 / 0.534201 (-0.518599) 0.270799 / 0.579283 (-0.308484) 0.279060 / 0.434364 (-0.155304) 0.315108 / 0.540337 (-0.225230) 0.413576 / 1.386936 (-0.973360)

@HuggingFaceDocBuilderDev
Copy link

HuggingFaceDocBuilderDev commented Nov 16, 2023 •
edited
Loading

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

mariosasko
mariosasko reviewed Nov 16, 2023
View reviewed changes
Copy link
Collaborator

@mariosasko mariosasko left a comment

Choose a reason for hiding this comment

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

Nice! Some nits:

albertvillanova
albertvillanova reviewed Nov 17, 2023
View reviewed changes
Copy link
Member

@albertvillanova albertvillanova left a comment

Choose a reason for hiding this comment

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

Thanks. Just a suggestion below.

src/datasets/load.py
streaming: bool = False,
num_proc: Optional[int] = None,
storage_options: Optional[Dict] = None,
trust_remote_code: bool = None,
Copy link
Member

@albertvillanova albertvillanova Nov 17, 2023

Choose a reason for hiding this comment

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

I think we should raise a FutureWarning to warn users that the default behavior will change in the future if they do not explicitly pass trust_remote_code.

Copy link
Member Author

@lhoestq lhoestq Nov 17, 2023

Choose a reason for hiding this comment

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

good idea !

@lhoestq lhoestq mentioned this pull request Nov 23, 2023
Merged
1 task
@github-actions
Copy link

github-actions bot commented Nov 24, 2023

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) 0.004911 / 0.011353 (-0.006442) 0.002808 / 0.011008 (-0.008200) 0.061367 / 0.038508 (0.022859) 0.050154 / 0.023109 (0.027045) 0.250403 / 0.275898 (-0.025495) 0.273831 / 0.323480 (-0.049649) 0.002914 / 0.007986 (-0.005071) 0.002493 / 0.004328 (-0.001836) 0.048288 / 0.004250 (0.044037) 0.039219 / 0.037052 (0.002167) 0.260043 / 0.258489 (0.001554) 0.288177 / 0.293841 (-0.005664) 0.023123 / 0.128546 (-0.105423) 0.006981 / 0.075646 (-0.068666) 0.201306 / 0.419271 (-0.217965) 0.035670 / 0.043533 (-0.007863) 0.255237 / 0.255139 (0.000098) 0.283701 / 0.283200 (0.000502) 0.019349 / 0.141683 (-0.122334) 1.100963 / 1.452155 (-0.351192) 1.152725 / 1.492716 (-0.339992)

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) 0.106350 / 0.018006 (0.088344) 0.300577 / 0.000490 (0.300087) 0.000206 / 0.000200 (0.000006) 0.000043 / 0.000054 (-0.000011)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.019028 / 0.037411 (-0.018384) 0.062643 / 0.014526 (0.048118) 0.072771 / 0.176557 (-0.103786) 0.119873 / 0.737135 (-0.617263) 0.074470 / 0.296338 (-0.221869)

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) 0.287032 / 0.215209 (0.071823) 2.826134 / 2.077655 (0.748480) 1.507362 / 1.504120 (0.003242) 1.382929 / 1.541195 (-0.158266) 1.385361 / 1.468490 (-0.083129) 0.412081 / 4.584777 (-4.172696) 2.384289 / 3.745712 (-1.361423) 2.551316 / 5.269862 (-2.718546) 1.562954 / 4.565676 (-3.002722) 0.046669 / 0.424275 (-0.377606) 0.004804 / 0.007607 (-0.002803) 0.337751 / 0.226044 (0.111707) 3.378894 / 2.268929 (1.109965) 1.848817 / 55.444624 (-53.595807) 1.564560 / 6.876477 (-5.311917) 1.579577 / 2.142072 (-0.562496) 0.484531 / 4.805227 (-4.320697) 0.101157 / 6.500664 (-6.399507) 0.042272 / 0.075469 (-0.033197)

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) 0.948289 / 1.841788 (-0.893498) 11.490877 / 8.074308 (3.416569) 10.492787 / 10.191392 (0.301395) 0.128575 / 0.680424 (-0.551849) 0.013716 / 0.534201 (-0.520485) 0.271075 / 0.579283 (-0.308208) 0.269749 / 0.434364 (-0.164615) 0.306378 / 0.540337 (-0.233959) 0.400204 / 1.386936 (-0.986732)
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) 0.004821 / 0.011353 (-0.006532) 0.002773 / 0.011008 (-0.008235) 0.048934 / 0.038508 (0.010426) 0.049490 / 0.023109 (0.026380) 0.271107 / 0.275898 (-0.004791) 0.291472 / 0.323480 (-0.032008) 0.004734 / 0.007986 (-0.003252) 0.002437 / 0.004328 (-0.001892) 0.048840 / 0.004250 (0.044590) 0.039757 / 0.037052 (0.002704) 0.276037 / 0.258489 (0.017548) 0.298220 / 0.293841 (0.004379) 0.024595 / 0.128546 (-0.103952) 0.007320 / 0.075646 (-0.068327) 0.054693 / 0.419271 (-0.364578) 0.032672 / 0.043533 (-0.010861) 0.271555 / 0.255139 (0.016416) 0.287685 / 0.283200 (0.004485) 0.017159 / 0.141683 (-0.124524) 1.118496 / 1.452155 (-0.333659) 1.177389 / 1.492716 (-0.315327)

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) 0.090469 / 0.018006 (0.072463) 0.306014 / 0.000490 (0.305525) 0.000218 / 0.000200 (0.000018) 0.000044 / 0.000054 (-0.000010)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.021452 / 0.037411 (-0.015960) 0.070014 / 0.014526 (0.055488) 0.081917 / 0.176557 (-0.094639) 0.120615 / 0.737135 (-0.616520) 0.081745 / 0.296338 (-0.214593)

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) 0.294049 / 0.215209 (0.078840) 2.886802 / 2.077655 (0.809147) 1.607817 / 1.504120 (0.103697) 1.474172 / 1.541195 (-0.067023) 1.474744 / 1.468490 (0.006254) 0.398178 / 4.584777 (-4.186599) 2.455908 / 3.745712 (-1.289804) 2.463003 / 5.269862 (-2.806858) 1.560402 / 4.565676 (-3.005275) 0.046208 / 0.424275 (-0.378067) 0.004862 / 0.007607 (-0.002745) 0.350862 / 0.226044 (0.124817) 3.463958 / 2.268929 (1.195030) 1.934696 / 55.444624 (-53.509928) 1.660090 / 6.876477 (-5.216387) 1.770920 / 2.142072 (-0.371153) 0.468409 / 4.805227 (-4.336819) 0.096812 / 6.500664 (-6.403852) 0.040580 / 0.075469 (-0.034889)

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) 0.978102 / 1.841788 (-0.863686) 11.943265 / 8.074308 (3.868957) 10.684995 / 10.191392 (0.493603) 0.131554 / 0.680424 (-0.548870) 0.015608 / 0.534201 (-0.518593) 0.271449 / 0.579283 (-0.307834) 0.282485 / 0.434364 (-0.151879) 0.302376 / 0.540337 (-0.237962) 0.524908 / 1.386936 (-0.862028)

@github-actions
Copy link

github-actions bot commented Nov 24, 2023

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) 0.004926 / 0.011353 (-0.006427) 0.003020 / 0.011008 (-0.007988) 0.061899 / 0.038508 (0.023391) 0.063836 / 0.023109 (0.040726) 0.239252 / 0.275898 (-0.036646) 0.268320 / 0.323480 (-0.055160) 0.003939 / 0.007986 (-0.004046) 0.002557 / 0.004328 (-0.001772) 0.048469 / 0.004250 (0.044219) 0.038707 / 0.037052 (0.001655) 0.247563 / 0.258489 (-0.010926) 0.281171 / 0.293841 (-0.012670) 0.023564 / 0.128546 (-0.104983) 0.007699 / 0.075646 (-0.067948) 0.207561 / 0.419271 (-0.211710) 0.036362 / 0.043533 (-0.007171) 0.248324 / 0.255139 (-0.006814) 0.269673 / 0.283200 (-0.013527) 0.018841 / 0.141683 (-0.122842) 1.123407 / 1.452155 (-0.328748) 1.170422 / 1.492716 (-0.322295)

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) 0.096278 / 0.018006 (0.078272) 0.311477 / 0.000490 (0.310988) 0.000217 / 0.000200 (0.000017) 0.000042 / 0.000054 (-0.000012)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.019470 / 0.037411 (-0.017942) 0.071888 / 0.014526 (0.057362) 0.074264 / 0.176557 (-0.102292) 0.124413 / 0.737135 (-0.612723) 0.075602 / 0.296338 (-0.220737)

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) 0.284755 / 0.215209 (0.069546) 2.770789 / 2.077655 (0.693135) 1.478276 / 1.504120 (-0.025843) 1.375287 / 1.541195 (-0.165907) 1.398032 / 1.468490 (-0.070458) 0.420457 / 4.584777 (-4.164320) 2.445929 / 3.745712 (-1.299783) 2.819548 / 5.269862 (-2.450313) 1.628506 / 4.565676 (-2.937171) 0.047687 / 0.424275 (-0.376588) 0.004861 / 0.007607 (-0.002746) 0.340173 / 0.226044 (0.114129) 3.340703 / 2.268929 (1.071774) 1.882803 / 55.444624 (-53.561821) 1.587206 / 6.876477 (-5.289271) 1.645298 / 2.142072 (-0.496774) 0.490957 / 4.805227 (-4.314270) 0.102779 / 6.500664 (-6.397885) 0.048372 / 0.075469 (-0.027098)

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) 0.958311 / 1.841788 (-0.883477) 12.354981 / 8.074308 (4.280673) 10.864826 / 10.191392 (0.673434) 0.149053 / 0.680424 (-0.531371) 0.015078 / 0.534201 (-0.519123) 0.270117 / 0.579283 (-0.309166) 0.274495 / 0.434364 (-0.159869) 0.307584 / 0.540337 (-0.232753) 0.405603 / 1.386936 (-0.981333)
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) 0.004996 / 0.011353 (-0.006357) 0.002995 / 0.011008 (-0.008014) 0.047897 / 0.038508 (0.009389) 0.056413 / 0.023109 (0.033303) 0.277669 / 0.275898 (0.001771) 0.300679 / 0.323480 (-0.022801) 0.004094 / 0.007986 (-0.003892) 0.002519 / 0.004328 (-0.001810) 0.049536 / 0.004250 (0.045285) 0.042341 / 0.037052 (0.005288) 0.281533 / 0.258489 (0.023044) 0.306771 / 0.293841 (0.012930) 0.025379 / 0.128546 (-0.103167) 0.007495 / 0.075646 (-0.068152) 0.054453 / 0.419271 (-0.364818) 0.032616 / 0.043533 (-0.010917) 0.277844 / 0.255139 (0.022705) 0.296265 / 0.283200 (0.013065) 0.019462 / 0.141683 (-0.122221) 1.115841 / 1.452155 (-0.336313) 1.169662 / 1.492716 (-0.323054)

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) 0.095459 / 0.018006 (0.077453) 0.301590 / 0.000490 (0.301100) 0.000230 / 0.000200 (0.000030) 0.000061 / 0.000054 (0.000006)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.022182 / 0.037411 (-0.015229) 0.085367 / 0.014526 (0.070842) 0.084006 / 0.176557 (-0.092550) 0.121260 / 0.737135 (-0.615876) 0.084137 / 0.296338 (-0.212202)

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) 0.310335 / 0.215209 (0.095126) 3.002531 / 2.077655 (0.924876) 1.642282 / 1.504120 (0.138162) 1.573044 / 1.541195 (0.031849) 1.572076 / 1.468490 (0.103586) 0.422037 / 4.584777 (-4.162740) 2.495295 / 3.745712 (-1.250417) 2.523707 / 5.269862 (-2.746155) 1.725824 / 4.565676 (-2.839853) 0.047814 / 0.424275 (-0.376461) 0.004868 / 0.007607 (-0.002739) 0.352833 / 0.226044 (0.126789) 3.477241 / 2.268929 (1.208313) 1.983888 / 55.444624 (-53.460736) 1.696883 / 6.876477 (-5.179594) 1.831665 / 2.142072 (-0.310407) 0.502976 / 4.805227 (-4.302251) 0.101264 / 6.500664 (-6.399400) 0.041779 / 0.075469 (-0.033690)

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) 0.981629 / 1.841788 (-0.860159) 12.550634 / 8.074308 (4.476326) 11.113382 / 10.191392 (0.921990) 0.136565 / 0.680424 (-0.543859) 0.016742 / 0.534201 (-0.517459) 0.274316 / 0.579283 (-0.304967) 0.284687 / 0.434364 (-0.149676) 0.309966 / 0.540337 (-0.230372) 0.557990 / 1.386936 (-0.828946)

@github-actions
Copy link

github-actions bot commented Nov 24, 2023

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) 0.004980 / 0.011353 (-0.006373) 0.002786 / 0.011008 (-0.008222) 0.062460 / 0.038508 (0.023952) 0.051811 / 0.023109 (0.028702) 0.231734 / 0.275898 (-0.044164) 0.254075 / 0.323480 (-0.069405) 0.002884 / 0.007986 (-0.005102) 0.002317 / 0.004328 (-0.002011) 0.049044 / 0.004250 (0.044793) 0.038984 / 0.037052 (0.001931) 0.241193 / 0.258489 (-0.017296) 0.272091 / 0.293841 (-0.021750) 0.023098 / 0.128546 (-0.105448) 0.007190 / 0.075646 (-0.068456) 0.201409 / 0.419271 (-0.217863) 0.036100 / 0.043533 (-0.007433) 0.238185 / 0.255139 (-0.016954) 0.257127 / 0.283200 (-0.026072) 0.019542 / 0.141683 (-0.122141) 1.127925 / 1.452155 (-0.324230) 1.174354 / 1.492716 (-0.318362)

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) 0.099608 / 0.018006 (0.081601) 0.315046 / 0.000490 (0.314556) 0.000282 / 0.000200 (0.000082) 0.000042 / 0.000054 (-0.000013)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.018710 / 0.037411 (-0.018701) 0.062557 / 0.014526 (0.048031) 0.074021 / 0.176557 (-0.102536) 0.119670 / 0.737135 (-0.617465) 0.076491 / 0.296338 (-0.219847)

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) 0.282940 / 0.215209 (0.067731) 2.788542 / 2.077655 (0.710887) 1.496039 / 1.504120 (-0.008080) 1.367542 / 1.541195 (-0.173653) 1.393705 / 1.468490 (-0.074785) 0.405910 / 4.584777 (-4.178867) 2.422544 / 3.745712 (-1.323168) 2.602822 / 5.269862 (-2.667039) 1.586853 / 4.565676 (-2.978823) 0.045440 / 0.424275 (-0.378836) 0.004792 / 0.007607 (-0.002815) 0.342059 / 0.226044 (0.116015) 3.366880 / 2.268929 (1.097952) 1.810566 / 55.444624 (-53.634058) 1.527112 / 6.876477 (-5.349364) 1.548906 / 2.142072 (-0.593166) 0.479491 / 4.805227 (-4.325736) 0.099807 / 6.500664 (-6.400857) 0.041951 / 0.075469 (-0.033518)

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) 0.953723 / 1.841788 (-0.888065) 11.837240 / 8.074308 (3.762932) 10.562979 / 10.191392 (0.371587) 0.145064 / 0.680424 (-0.535360) 0.014285 / 0.534201 (-0.519916) 0.270605 / 0.579283 (-0.308678) 0.264086 / 0.434364 (-0.170278) 0.308000 / 0.540337 (-0.232337) 0.403916 / 1.386936 (-0.983020)
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) 0.004796 / 0.011353 (-0.006557) 0.002997 / 0.011008 (-0.008011) 0.048702 / 0.038508 (0.010193) 0.053377 / 0.023109 (0.030267) 0.271852 / 0.275898 (-0.004046) 0.293366 / 0.323480 (-0.030114) 0.004041 / 0.007986 (-0.003945) 0.002459 / 0.004328 (-0.001869) 0.048197 / 0.004250 (0.043947) 0.040094 / 0.037052 (0.003042) 0.275837 / 0.258489 (0.017348) 0.301174 / 0.293841 (0.007333) 0.024433 / 0.128546 (-0.104113) 0.007203 / 0.075646 (-0.068444) 0.054080 / 0.419271 (-0.365192) 0.033237 / 0.043533 (-0.010295) 0.271177 / 0.255139 (0.016038) 0.293062 / 0.283200 (0.009862) 0.018399 / 0.141683 (-0.123284) 1.149527 / 1.452155 (-0.302628) 1.202717 / 1.492716 (-0.290000)

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) 0.093168 / 0.018006 (0.075162) 0.290536 / 0.000490 (0.290046) 0.000290 / 0.000200 (0.000090) 0.000074 / 0.000054 (0.000020)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.021191 / 0.037411 (-0.016221) 0.069990 / 0.014526 (0.055465) 0.080636 / 0.176557 (-0.095920) 0.120151 / 0.737135 (-0.616984) 0.082944 / 0.296338 (-0.213395)

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) 0.289673 / 0.215209 (0.074463) 2.828419 / 2.077655 (0.750764) 1.590741 / 1.504120 (0.086621) 1.480969 / 1.541195 (-0.060226) 1.512761 / 1.468490 (0.044271) 0.398328 / 4.584777 (-4.186449) 2.441134 / 3.745712 (-1.304578) 2.487606 / 5.269862 (-2.782256) 1.586604 / 4.565676 (-2.979073) 0.045578 / 0.424275 (-0.378697) 0.004842 / 0.007607 (-0.002766) 0.344556 / 0.226044 (0.118512) 3.395982 / 2.268929 (1.127053) 1.963354 / 55.444624 (-53.481271) 1.680496 / 6.876477 (-5.195980) 1.827916 / 2.142072 (-0.314157) 0.476203 / 4.805227 (-4.329024) 0.098016 / 6.500664 (-6.402648) 0.041234 / 0.075469 (-0.034235)

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) 0.977820 / 1.841788 (-0.863968) 12.139614 / 8.074308 (4.065306) 10.643071 / 10.191392 (0.451679) 0.130928 / 0.680424 (-0.549496) 0.015341 / 0.534201 (-0.518860) 0.271304 / 0.579283 (-0.307979) 0.284671 / 0.434364 (-0.149693) 0.306210 / 0.540337 (-0.234128) 0.546498 / 1.386936 (-0.840438)

@github-actions
Copy link

github-actions bot commented Nov 24, 2023

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) 0.004748 / 0.011353 (-0.006605) 0.002942 / 0.011008 (-0.008066) 0.061298 / 0.038508 (0.022790) 0.052873 / 0.023109 (0.029764) 0.250573 / 0.275898 (-0.025325) 0.270636 / 0.323480 (-0.052844) 0.002925 / 0.007986 (-0.005061) 0.003126 / 0.004328 (-0.001203) 0.047340 / 0.004250 (0.043090) 0.038662 / 0.037052 (0.001609) 0.252151 / 0.258489 (-0.006338) 0.284700 / 0.293841 (-0.009141) 0.025145 / 0.128546 (-0.103402) 0.007075 / 0.075646 (-0.068572) 0.200501 / 0.419271 (-0.218771) 0.035623 / 0.043533 (-0.007910) 0.249657 / 0.255139 (-0.005482) 0.272384 / 0.283200 (-0.010815) 0.018331 / 0.141683 (-0.123351) 1.095064 / 1.452155 (-0.357091) 1.145304 / 1.492716 (-0.347412)

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) 0.092548 / 0.018006 (0.074542) 0.299338 / 0.000490 (0.298848) 0.000212 / 0.000200 (0.000012) 0.000046 / 0.000054 (-0.000009)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.018723 / 0.037411 (-0.018688) 0.062226 / 0.014526 (0.047700) 0.072840 / 0.176557 (-0.103717) 0.120073 / 0.737135 (-0.617063) 0.074536 / 0.296338 (-0.221802)

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) 0.284862 / 0.215209 (0.069653) 2.791842 / 2.077655 (0.714188) 1.506481 / 1.504120 (0.002361) 1.368952 / 1.541195 (-0.172243) 1.372555 / 1.468490 (-0.095935) 0.408292 / 4.584777 (-4.176485) 2.381155 / 3.745712 (-1.364558) 2.613617 / 5.269862 (-2.656244) 1.575892 / 4.565676 (-2.989785) 0.047526 / 0.424275 (-0.376749) 0.004792 / 0.007607 (-0.002815) 0.344818 / 0.226044 (0.118773) 3.344965 / 2.268929 (1.076036) 1.883659 / 55.444624 (-53.560965) 1.596039 / 6.876477 (-5.280437) 1.584410 / 2.142072 (-0.557662) 0.486672 / 4.805227 (-4.318555) 0.101464 / 6.500664 (-6.399200) 0.041824 / 0.075469 (-0.033645)

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) 0.930491 / 1.841788 (-0.911296) 11.636526 / 8.074308 (3.562218) 10.371829 / 10.191392 (0.180437) 0.138181 / 0.680424 (-0.542243) 0.014307 / 0.534201 (-0.519894) 0.268322 / 0.579283 (-0.310961) 0.264173 / 0.434364 (-0.170191) 0.303649 / 0.540337 (-0.236688) 0.399958 / 1.386936 (-0.986978)
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) 0.004802 / 0.011353 (-0.006551) 0.002861 / 0.011008 (-0.008147) 0.048843 / 0.038508 (0.010335) 0.053887 / 0.023109 (0.030778) 0.278690 / 0.275898 (0.002792) 0.302729 / 0.323480 (-0.020751) 0.003929 / 0.007986 (-0.004057) 0.002376 / 0.004328 (-0.001953) 0.048146 / 0.004250 (0.043896) 0.039842 / 0.037052 (0.002790) 0.281595 / 0.258489 (0.023106) 0.305813 / 0.293841 (0.011972) 0.024214 / 0.128546 (-0.104333) 0.007201 / 0.075646 (-0.068446) 0.053604 / 0.419271 (-0.365667) 0.032841 / 0.043533 (-0.010691) 0.276168 / 0.255139 (0.021029) 0.293869 / 0.283200 (0.010669) 0.017550 / 0.141683 (-0.124132) 1.121508 / 1.452155 (-0.330647) 1.177694 / 1.492716 (-0.315022)

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) 0.091805 / 0.018006 (0.073799) 0.299026 / 0.000490 (0.298536) 0.000219 / 0.000200 (0.000019) 0.000051 / 0.000054 (-0.000004)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.021094 / 0.037411 (-0.016318) 0.069769 / 0.014526 (0.055243) 0.081191 / 0.176557 (-0.095366) 0.118884 / 0.737135 (-0.618252) 0.081955 / 0.296338 (-0.214383)

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) 0.292159 / 0.215209 (0.076950) 2.874473 / 2.077655 (0.796819) 1.614695 / 1.504120 (0.110575) 1.492123 / 1.541195 (-0.049071) 1.505293 / 1.468490 (0.036803) 0.394498 / 4.584777 (-4.190279) 2.455539 / 3.745712 (-1.290173) 2.458184 / 5.269862 (-2.811677) 1.569108 / 4.565676 (-2.996569) 0.046576 / 0.424275 (-0.377699) 0.005093 / 0.007607 (-0.002514) 0.346142 / 0.226044 (0.120098) 3.398171 / 2.268929 (1.129242) 1.971953 / 55.444624 (-53.472672) 1.695275 / 6.876477 (-5.181201) 1.840511 / 2.142072 (-0.301562) 0.465932 / 4.805227 (-4.339295) 0.098578 / 6.500664 (-6.402086) 0.040456 / 0.075469 (-0.035013)

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) 0.977636 / 1.841788 (-0.864152) 12.083585 / 8.074308 (4.009277) 10.509082 / 10.191392 (0.317690) 0.130717 / 0.680424 (-0.549707) 0.015958 / 0.534201 (-0.518243) 0.273504 / 0.579283 (-0.305780) 0.276498 / 0.434364 (-0.157866) 0.306139 / 0.540337 (-0.234199) 0.522521 / 1.386936 (-0.864415)

@github-actions
Copy link

github-actions bot commented Nov 24, 2023

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) 0.004859 / 0.011353 (-0.006493) 0.002423 / 0.011008 (-0.008585) 0.060969 / 0.038508 (0.022461) 0.048758 / 0.023109 (0.025649) 0.245400 / 0.275898 (-0.030498) 0.263686 / 0.323480 (-0.059794) 0.002852 / 0.007986 (-0.005134) 0.002273 / 0.004328 (-0.002055) 0.047648 / 0.004250 (0.043398) 0.038310 / 0.037052 (0.001258) 0.249849 / 0.258489 (-0.008640) 0.279305 / 0.293841 (-0.014536) 0.022897 / 0.128546 (-0.105649) 0.006882 / 0.075646 (-0.068764) 0.202793 / 0.419271 (-0.216478) 0.034557 / 0.043533 (-0.008976) 0.252147 / 0.255139 (-0.002992) 0.267414 / 0.283200 (-0.015785) 0.019956 / 0.141683 (-0.121727) 1.106181 / 1.452155 (-0.345973) 1.158423 / 1.492716 (-0.334293)

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) 0.086848 / 0.018006 (0.068842) 0.295235 / 0.000490 (0.294745) 0.000211 / 0.000200 (0.000011) 0.000041 / 0.000054 (-0.000014)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.018209 / 0.037411 (-0.019203) 0.061967 / 0.014526 (0.047441) 0.071551 / 0.176557 (-0.105005) 0.117525 / 0.737135 (-0.619611) 0.073401 / 0.296338 (-0.222937)

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) 0.272388 / 0.215209 (0.057179) 2.689797 / 2.077655 (0.612143) 1.440897 / 1.504120 (-0.063223) 1.334689 / 1.541195 (-0.206505) 1.356395 / 1.468490 (-0.112095) 0.387201 / 4.584777 (-4.197576) 2.342908 / 3.745712 (-1.402804) 2.480156 / 5.269862 (-2.789706) 1.512342 / 4.565676 (-3.053335) 0.042324 / 0.424275 (-0.381951) 0.004744 / 0.007607 (-0.002863) 0.323568 / 0.226044 (0.097523) 3.190021 / 2.268929 (0.921093) 1.765046 / 55.444624 (-53.679578) 1.513958 / 6.876477 (-5.362519) 1.504943 / 2.142072 (-0.637129) 0.452302 / 4.805227 (-4.352925) 0.094728 / 6.500664 (-6.405936) 0.038641 / 0.075469 (-0.036828)

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) 0.939721 / 1.841788 (-0.902067) 11.174180 / 8.074308 (3.099872) 10.046717 / 10.191392 (-0.144675) 0.124877 / 0.680424 (-0.555547) 0.013687 / 0.534201 (-0.520514) 0.261002 / 0.579283 (-0.318282) 0.267349 / 0.434364 (-0.167015) 0.306545 / 0.540337 (-0.233792) 0.389322 / 1.386936 (-0.997614)
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) 0.004702 / 0.011353 (-0.006651) 0.002431 / 0.011008 (-0.008577) 0.046138 / 0.038508 (0.007630) 0.048356 / 0.023109 (0.025246) 0.272154 / 0.275898 (-0.003744) 0.292676 / 0.323480 (-0.030804) 0.003870 / 0.007986 (-0.004115) 0.002294 / 0.004328 (-0.002035) 0.048129 / 0.004250 (0.043879) 0.039026 / 0.037052 (0.001974) 0.273900 / 0.258489 (0.015411) 0.295927 / 0.293841 (0.002086) 0.024044 / 0.128546 (-0.104502) 0.006906 / 0.075646 (-0.068740) 0.053268 / 0.419271 (-0.366004) 0.032360 / 0.043533 (-0.011173) 0.273470 / 0.255139 (0.018331) 0.286207 / 0.283200 (0.003007) 0.017580 / 0.141683 (-0.124103) 1.091064 / 1.452155 (-0.361091) 1.159645 / 1.492716 (-0.333071)

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) 0.087149 / 0.018006 (0.069143) 0.293489 / 0.000490 (0.293000) 0.000217 / 0.000200 (0.000017) 0.000052 / 0.000054 (-0.000003)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.021779 / 0.037411 (-0.015632) 0.066453 / 0.014526 (0.051928) 0.078517 / 0.176557 (-0.098039) 0.117317 / 0.737135 (-0.619819) 0.079828 / 0.296338 (-0.216511)

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) 0.287605 / 0.215209 (0.072396) 2.811094 / 2.077655 (0.733439) 1.572474 / 1.504120 (0.068354) 1.450294 / 1.541195 (-0.090900) 1.456052 / 1.468490 (-0.012438) 0.402095 / 4.584777 (-4.182682) 2.444709 / 3.745712 (-1.301003) 2.390837 / 5.269862 (-2.879024) 1.530519 / 4.565676 (-3.035157) 0.043520 / 0.424275 (-0.380755) 0.004788 / 0.007607 (-0.002819) 0.337436 / 0.226044 (0.111391) 3.326111 / 2.268929 (1.057182) 1.889273 / 55.444624 (-53.555352) 1.624423 / 6.876477 (-5.252054) 1.715766 / 2.142072 (-0.426307) 0.484570 / 4.805227 (-4.320657) 0.091691 / 6.500664 (-6.408973) 0.038278 / 0.075469 (-0.037191)

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) 0.961708 / 1.841788 (-0.880079) 11.496471 / 8.074308 (3.422162) 10.211589 / 10.191392 (0.020197) 0.127584 / 0.680424 (-0.552840) 0.015178 / 0.534201 (-0.519023) 0.267290 / 0.579283 (-0.311993) 0.259305 / 0.434364 (-0.175059) 0.303433 / 0.540337 (-0.236905) 0.508016 / 1.386936 (-0.878920)

@github-actions
Copy link

github-actions bot commented Nov 24, 2023

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) 0.004558 / 0.011353 (-0.006795) 0.002563 / 0.011008 (-0.008445) 0.061314 / 0.038508 (0.022806) 0.049312 / 0.023109 (0.026203) 0.240988 / 0.275898 (-0.034910) 0.260548 / 0.323480 (-0.062932) 0.002817 / 0.007986 (-0.005169) 0.002904 / 0.004328 (-0.001425) 0.048515 / 0.004250 (0.044264) 0.037511 / 0.037052 (0.000459) 0.244880 / 0.258489 (-0.013609) 0.276118 / 0.293841 (-0.017723) 0.022636 / 0.128546 (-0.105910) 0.006694 / 0.075646 (-0.068953) 0.201336 / 0.419271 (-0.217936) 0.035228 / 0.043533 (-0.008305) 0.242424 / 0.255139 (-0.012715) 0.260178 / 0.283200 (-0.023022) 0.017836 / 0.141683 (-0.123847) 1.122296 / 1.452155 (-0.329859) 1.189024 / 1.492716 (-0.303692)

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) 0.090051 / 0.018006 (0.072045) 0.298562 / 0.000490 (0.298073) 0.000216 / 0.000200 (0.000016) 0.000051 / 0.000054 (-0.000004)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.018228 / 0.037411 (-0.019184) 0.062379 / 0.014526 (0.047853) 0.073482 / 0.176557 (-0.103075) 0.120341 / 0.737135 (-0.616794) 0.073868 / 0.296338 (-0.222470)

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) 0.280195 / 0.215209 (0.064986) 2.743333 / 2.077655 (0.665678) 1.470078 / 1.504120 (-0.034042) 1.335874 / 1.541195 (-0.205321) 1.342961 / 1.468490 (-0.125529) 0.409203 / 4.584777 (-4.175574) 2.392217 / 3.745712 (-1.353495) 2.544161 / 5.269862 (-2.725701) 1.544016 / 4.565676 (-3.021660) 0.059485 / 0.424275 (-0.364790) 0.004833 / 0.007607 (-0.002775) 0.335114 / 0.226044 (0.109070) 3.289009 / 2.268929 (1.020080) 1.854666 / 55.444624 (-53.589959) 1.566282 / 6.876477 (-5.310195) 1.561287 / 2.142072 (-0.580786) 0.484961 / 4.805227 (-4.320267) 0.099651 / 6.500664 (-6.401013) 0.041408 / 0.075469 (-0.034061)

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) 0.941743 / 1.841788 (-0.900044) 11.165692 / 8.074308 (3.091383) 10.236693 / 10.191392 (0.045301) 0.129694 / 0.680424 (-0.550730) 0.014879 / 0.534201 (-0.519322) 0.275120 / 0.579283 (-0.304163) 0.263822 / 0.434364 (-0.170542) 0.306429 / 0.540337 (-0.233909) 0.397611 / 1.386936 (-0.989325)
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) 0.004714 / 0.011353 (-0.006639) 0.002430 / 0.011008 (-0.008578) 0.047644 / 0.038508 (0.009136) 0.049710 / 0.023109 (0.026601) 0.271950 / 0.275898 (-0.003948) 0.290996 / 0.323480 (-0.032483) 0.003888 / 0.007986 (-0.004097) 0.002367 / 0.004328 (-0.001962) 0.047623 / 0.004250 (0.043372) 0.039574 / 0.037052 (0.002522) 0.274540 / 0.258489 (0.016051) 0.298065 / 0.293841 (0.004224) 0.024677 / 0.128546 (-0.103869) 0.006844 / 0.075646 (-0.068802) 0.053180 / 0.419271 (-0.366091) 0.032391 / 0.043533 (-0.011141) 0.273222 / 0.255139 (0.018083) 0.290336 / 0.283200 (0.007136) 0.017911 / 0.141683 (-0.123772) 1.105879 / 1.452155 (-0.346276) 1.176979 / 1.492716 (-0.315737)

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) 0.089563 / 0.018006 (0.071557) 0.296392 / 0.000490 (0.295903) 0.000214 / 0.000200 (0.000014) 0.000053 / 0.000054 (-0.000001)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.021588 / 0.037411 (-0.015824) 0.069951 / 0.014526 (0.055425) 0.080397 / 0.176557 (-0.096160) 0.118772 / 0.737135 (-0.618363) 0.080356 / 0.296338 (-0.215983)

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) 0.288492 / 0.215209 (0.073283) 2.839553 / 2.077655 (0.761898) 1.597504 / 1.504120 (0.093384) 1.475001 / 1.541195 (-0.066193) 1.481561 / 1.468490 (0.013071) 0.411851 / 4.584777 (-4.172926) 2.397322 / 3.745712 (-1.348390) 2.444078 / 5.269862 (-2.825784) 1.557106 / 4.565676 (-3.008571) 0.047159 / 0.424275 (-0.377116) 0.004842 / 0.007607 (-0.002765) 0.346221 / 0.226044 (0.120177) 3.387900 / 2.268929 (1.118972) 1.962167 / 55.444624 (-53.482457) 1.675017 / 6.876477 (-5.201460) 1.788745 / 2.142072 (-0.353328) 0.488063 / 4.805227 (-4.317164) 0.098878 / 6.500664 (-6.401786) 0.040369 / 0.075469 (-0.035100)

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) 0.977999 / 1.841788 (-0.863789) 11.671558 / 8.074308 (3.597250) 10.327847 / 10.191392 (0.136455) 0.129317 / 0.680424 (-0.551107) 0.015600 / 0.534201 (-0.518601) 0.267967 / 0.579283 (-0.311316) 0.273811 / 0.434364 (-0.160553) 0.301749 / 0.540337 (-0.238588) 0.515493 / 1.386936 (-0.871443)

@lhoestq lhoestq marked this pull request as ready for review November 27, 2023 13:33
@lhoestq
Copy link
Member Author

lhoestq commented Nov 27, 2023

I added tests and docs @mariosasko @albertvillanova let le know what you think !

mariosasko
mariosasko reviewed Nov 27, 2023
View reviewed changes
Copy link
Collaborator

@mariosasko mariosasko left a comment

Choose a reason for hiding this comment

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

Just a couple of nits:

@github-actions
Copy link

github-actions bot commented Nov 27, 2023

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) 0.004867 / 0.011353 (-0.006486) 0.002952 / 0.011008 (-0.008056) 0.062008 / 0.038508 (0.023500) 0.055279 / 0.023109 (0.032170) 0.248160 / 0.275898 (-0.027738) 0.276173 / 0.323480 (-0.047307) 0.003945 / 0.007986 (-0.004041) 0.002371 / 0.004328 (-0.001958) 0.048385 / 0.004250 (0.044134) 0.038997 / 0.037052 (0.001945) 0.257465 / 0.258489 (-0.001024) 0.286920 / 0.293841 (-0.006921) 0.023031 / 0.128546 (-0.105515) 0.007075 / 0.075646 (-0.068571) 0.201897 / 0.419271 (-0.217375) 0.035637 / 0.043533 (-0.007896) 0.252050 / 0.255139 (-0.003089) 0.272580 / 0.283200 (-0.010620) 0.018578 / 0.141683 (-0.123105) 1.129427 / 1.452155 (-0.322727) 1.172182 / 1.492716 (-0.320534)

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) 0.091806 / 0.018006 (0.073800) 0.298632 / 0.000490 (0.298143) 0.000202 / 0.000200 (0.000002) 0.000047 / 0.000054 (-0.000007)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.019123 / 0.037411 (-0.018288) 0.062603 / 0.014526 (0.048077) 0.074352 / 0.176557 (-0.102205) 0.120431 / 0.737135 (-0.616704) 0.074622 / 0.296338 (-0.221717)

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) 0.276019 / 0.215209 (0.060810) 2.701610 / 2.077655 (0.623955) 1.398388 / 1.504120 (-0.105732) 1.270902 / 1.541195 (-0.270292) 1.307992 / 1.468490 (-0.160499) 0.396350 / 4.584777 (-4.188427) 2.351064 / 3.745712 (-1.394648) 2.606229 / 5.269862 (-2.663632) 1.591075 / 4.565676 (-2.974601) 0.046429 / 0.424275 (-0.377846) 0.004832 / 0.007607 (-0.002775) 0.327887 / 0.226044 (0.101843) 3.277847 / 2.268929 (1.008918) 1.767210 / 55.444624 (-53.677414) 1.483997 / 6.876477 (-5.392479) 1.515689 / 2.142072 (-0.626383) 0.471326 / 4.805227 (-4.333902) 0.098821 / 6.500664 (-6.401843) 0.041914 / 0.075469 (-0.033555)

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) 0.956278 / 1.841788 (-0.885510) 11.924373 / 8.074308 (3.850065) 10.493926 / 10.191392 (0.302534) 0.140214 / 0.680424 (-0.540210) 0.013679 / 0.534201 (-0.520522) 0.270304 / 0.579283 (-0.308979) 0.266518 / 0.434364 (-0.167846) 0.310113 / 0.540337 (-0.230224) 0.399811 / 1.386936 (-0.987125)
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) 0.004793 / 0.011353 (-0.006560) 0.002879 / 0.011008 (-0.008130) 0.048632 / 0.038508 (0.010124) 0.051413 / 0.023109 (0.028304) 0.272704 / 0.275898 (-0.003194) 0.291541 / 0.323480 (-0.031939) 0.003913 / 0.007986 (-0.004072) 0.002387 / 0.004328 (-0.001941) 0.049045 / 0.004250 (0.044795) 0.040164 / 0.037052 (0.003112) 0.273052 / 0.258489 (0.014563) 0.300139 / 0.293841 (0.006298) 0.024225 / 0.128546 (-0.104321) 0.007060 / 0.075646 (-0.068587) 0.054360 / 0.419271 (-0.364911) 0.032882 / 0.043533 (-0.010650) 0.270295 / 0.255139 (0.015157) 0.312253 / 0.283200 (0.029054) 0.017413 / 0.141683 (-0.124270) 1.137306 / 1.452155 (-0.314849) 1.203705 / 1.492716 (-0.289011)

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) 0.091083 / 0.018006 (0.073077) 0.301607 / 0.000490 (0.301117) 0.000219 / 0.000200 (0.000019) 0.000052 / 0.000054 (-0.000003)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.021753 / 0.037411 (-0.015658) 0.069693 / 0.014526 (0.055167) 0.080481 / 0.176557 (-0.096075) 0.118581 / 0.737135 (-0.618555) 0.082231 / 0.296338 (-0.214108)

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) 0.300014 / 0.215209 (0.084805) 2.885934 / 2.077655 (0.808279) 1.594120 / 1.504120 (0.090000) 1.472312 / 1.541195 (-0.068883) 1.491663 / 1.468490 (0.023173) 0.412946 / 4.584777 (-4.171831) 2.494168 / 3.745712 (-1.251544) 2.527987 / 5.269862 (-2.741875) 1.589187 / 4.565676 (-2.976490) 0.046594 / 0.424275 (-0.377681) 0.004810 / 0.007607 (-0.002797) 0.345496 / 0.226044 (0.119452) 3.428850 / 2.268929 (1.159921) 1.952696 / 55.444624 (-53.491929) 1.663285 / 6.876477 (-5.213191) 1.822187 / 2.142072 (-0.319885) 0.483798 / 4.805227 (-4.321430) 0.101403 / 6.500664 (-6.399261) 0.041773 / 0.075469 (-0.033696)

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) 0.974247 / 1.841788 (-0.867541) 12.459980 / 8.074308 (4.385672) 10.354792 / 10.191392 (0.163400) 0.129083 / 0.680424 (-0.551341) 0.015225 / 0.534201 (-0.518976) 0.267673 / 0.579283 (-0.311610) 0.281011 / 0.434364 (-0.153352) 0.303054 / 0.540337 (-0.237283) 0.405719 / 1.386936 (-0.981217)

@lhoestq
Copy link
Member Author

lhoestq commented Nov 27, 2023

I switched to using deepmind/code_contests in examples in the docs to avoid having to pass trust_remote_code, and remove the DEFAULT naming stuff :)

@github-actions
Copy link

github-actions bot commented Nov 27, 2023

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) 0.005169 / 0.011353 (-0.006184) 0.003066 / 0.011008 (-0.007942) 0.068884 / 0.038508 (0.030376) 0.060345 / 0.023109 (0.037236) 0.243050 / 0.275898 (-0.032848) 0.265523 / 0.323480 (-0.057957) 0.002918 / 0.007986 (-0.005067) 0.002495 / 0.004328 (-0.001834) 0.051538 / 0.004250 (0.047288) 0.040010 / 0.037052 (0.002957) 0.249603 / 0.258489 (-0.008886) 0.287955 / 0.293841 (-0.005886) 0.024003 / 0.128546 (-0.104543) 0.007111 / 0.075646 (-0.068535) 0.205041 / 0.419271 (-0.214231) 0.036296 / 0.043533 (-0.007237) 0.246135 / 0.255139 (-0.009004) 0.268801 / 0.283200 (-0.014399) 0.018451 / 0.141683 (-0.123232) 1.130387 / 1.452155 (-0.321767) 1.162041 / 1.492716 (-0.330675)

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) 0.096370 / 0.018006 (0.078364) 0.309867 / 0.000490 (0.309377) 0.000229 / 0.000200 (0.000029) 0.000051 / 0.000054 (-0.000004)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.018688 / 0.037411 (-0.018723) 0.062859 / 0.014526 (0.048333) 0.076383 / 0.176557 (-0.100173) 0.120385 / 0.737135 (-0.616750) 0.080192 / 0.296338 (-0.216147)

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) 0.282994 / 0.215209 (0.067785) 2.742341 / 2.077655 (0.664686) 1.432041 / 1.504120 (-0.072079) 1.303282 / 1.541195 (-0.237913) 1.347198 / 1.468490 (-0.121292) 0.399145 / 4.584777 (-4.185632) 2.359766 / 3.745712 (-1.385947) 2.753577 / 5.269862 (-2.516285) 1.639953 / 4.565676 (-2.925724) 0.047111 / 0.424275 (-0.377164) 0.004946 / 0.007607 (-0.002661) 0.338857 / 0.226044 (0.112813) 3.328709 / 2.268929 (1.059781) 1.794729 / 55.444624 (-53.649895) 1.508514 / 6.876477 (-5.367963) 1.550737 / 2.142072 (-0.591335) 0.484227 / 4.805227 (-4.321000) 0.101001 / 6.500664 (-6.399663) 0.042792 / 0.075469 (-0.032677)

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) 0.956471 / 1.841788 (-0.885317) 12.031362 / 8.074308 (3.957054) 10.512914 / 10.191392 (0.321522) 0.141841 / 0.680424 (-0.538583) 0.014343 / 0.534201 (-0.519858) 0.273916 / 0.579283 (-0.305367) 0.266150 / 0.434364 (-0.168214) 0.312020 / 0.540337 (-0.228317) 0.410465 / 1.386936 (-0.976471)
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) 0.004945 / 0.011353 (-0.006408) 0.003288 / 0.011008 (-0.007720) 0.048247 / 0.038508 (0.009739) 0.057892 / 0.023109 (0.034783) 0.269741 / 0.275898 (-0.006157) 0.293728 / 0.323480 (-0.029752) 0.004789 / 0.007986 (-0.003197) 0.002477 / 0.004328 (-0.001852) 0.047825 / 0.004250 (0.043575) 0.040780 / 0.037052 (0.003727) 0.273355 / 0.258489 (0.014865) 0.300057 / 0.293841 (0.006216) 0.024481 / 0.128546 (-0.104066) 0.007285 / 0.075646 (-0.068361) 0.053046 / 0.419271 (-0.366226) 0.032342 / 0.043533 (-0.011190) 0.272293 / 0.255139 (0.017154) 0.290842 / 0.283200 (0.007642) 0.017546 / 0.141683 (-0.124137) 1.155816 / 1.452155 (-0.296339) 1.195839 / 1.492716 (-0.296878)

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) 0.094177 / 0.018006 (0.076170) 0.305122 / 0.000490 (0.304632) 0.000237 / 0.000200 (0.000037) 0.000063 / 0.000054 (0.000009)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.021817 / 0.037411 (-0.015595) 0.070711 / 0.014526 (0.056185) 0.084028 / 0.176557 (-0.092528) 0.120160 / 0.737135 (-0.616975) 0.083085 / 0.296338 (-0.213254)

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) 0.289127 / 0.215209 (0.073918) 2.826365 / 2.077655 (0.748710) 1.582910 / 1.504120 (0.078790) 1.472796 / 1.541195 (-0.068399) 1.497491 / 1.468490 (0.029000) 0.412276 / 4.584777 (-4.172501) 2.430692 / 3.745712 (-1.315020) 2.556444 / 5.269862 (-2.713418) 1.625782 / 4.565676 (-2.939895) 0.047921 / 0.424275 (-0.376354) 0.004809 / 0.007607 (-0.002798) 0.345569 / 0.226044 (0.119524) 3.417785 / 2.268929 (1.148856) 1.959223 / 55.444624 (-53.485401) 1.672765 / 6.876477 (-5.203712) 1.852444 / 2.142072 (-0.289628) 0.489225 / 4.805227 (-4.316002) 0.100624 / 6.500664 (-6.400040) 0.041242 / 0.075469 (-0.034227)

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) 0.971130 / 1.841788 (-0.870658) 12.652204 / 8.074308 (4.577896) 10.661821 / 10.191392 (0.470429) 0.147636 / 0.680424 (-0.532787) 0.015738 / 0.534201 (-0.518463) 0.272763 / 0.579283 (-0.306520) 0.282623 / 0.434364 (-0.151741) 0.341303 / 0.540337 (-0.199035) 0.412149 / 1.386936 (-0.974787)

@github-actions
Copy link

github-actions bot commented Nov 27, 2023

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) 0.004589 / 0.011353 (-0.006764) 0.002730 / 0.011008 (-0.008279) 0.061862 / 0.038508 (0.023353) 0.050945 / 0.023109 (0.027836) 0.240776 / 0.275898 (-0.035122) 0.266000 / 0.323480 (-0.057480) 0.003823 / 0.007986 (-0.004162) 0.002345 / 0.004328 (-0.001983) 0.047821 / 0.004250 (0.043571) 0.037813 / 0.037052 (0.000761) 0.251075 / 0.258489 (-0.007415) 0.279430 / 0.293841 (-0.014411) 0.022957 / 0.128546 (-0.105590) 0.007294 / 0.075646 (-0.068353) 0.206092 / 0.419271 (-0.213180) 0.035308 / 0.043533 (-0.008225) 0.247197 / 0.255139 (-0.007942) 0.264988 / 0.283200 (-0.018212) 0.017588 / 0.141683 (-0.124095) 1.093291 / 1.452155 (-0.358864) 1.165477 / 1.492716 (-0.327240)

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) 0.104057 / 0.018006 (0.086051) 0.303424 / 0.000490 (0.302934) 0.000223 / 0.000200 (0.000023) 0.000051 / 0.000054 (-0.000004)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.019040 / 0.037411 (-0.018371) 0.063161 / 0.014526 (0.048635) 0.085333 / 0.176557 (-0.091224) 0.155973 / 0.737135 (-0.581162) 0.077528 / 0.296338 (-0.218810)

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) 0.276104 / 0.215209 (0.060895) 2.738174 / 2.077655 (0.660519) 1.479484 / 1.504120 (-0.024636) 1.354094 / 1.541195 (-0.187100) 1.385312 / 1.468490 (-0.083178) 0.401398 / 4.584777 (-4.183379) 2.368503 / 3.745712 (-1.377209) 2.586405 / 5.269862 (-2.683457) 1.573978 / 4.565676 (-2.991699) 0.046969 / 0.424275 (-0.377306) 0.004874 / 0.007607 (-0.002733) 0.334028 / 0.226044 (0.107984) 3.269645 / 2.268929 (1.000717) 1.834528 / 55.444624 (-53.610096) 1.559883 / 6.876477 (-5.316594) 1.581380 / 2.142072 (-0.560693) 0.479580 / 4.805227 (-4.325647) 0.099077 / 6.500664 (-6.401587) 0.041166 / 0.075469 (-0.034303)

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) 0.918810 / 1.841788 (-0.922978) 11.505017 / 8.074308 (3.430709) 10.331934 / 10.191392 (0.140542) 0.128079 / 0.680424 (-0.552345) 0.013716 / 0.534201 (-0.520485) 0.271567 / 0.579283 (-0.307716) 0.264846 / 0.434364 (-0.169518) 0.305245 / 0.540337 (-0.235092) 0.401391 / 1.386936 (-0.985546)
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) 0.004860 / 0.011353 (-0.006493) 0.002854 / 0.011008 (-0.008155) 0.048327 / 0.038508 (0.009819) 0.051377 / 0.023109 (0.028268) 0.264344 / 0.275898 (-0.011554) 0.286800 / 0.323480 (-0.036680) 0.003969 / 0.007986 (-0.004016) 0.002415 / 0.004328 (-0.001914) 0.048498 / 0.004250 (0.044247) 0.040399 / 0.037052 (0.003347) 0.267254 / 0.258489 (0.008765) 0.292049 / 0.293841 (-0.001792) 0.024730 / 0.128546 (-0.103817) 0.007275 / 0.075646 (-0.068371) 0.053725 / 0.419271 (-0.365546) 0.033142 / 0.043533 (-0.010391) 0.265418 / 0.255139 (0.010279) 0.286242 / 0.283200 (0.003042) 0.017824 / 0.141683 (-0.123859) 1.135978 / 1.452155 (-0.316176) 1.192506 / 1.492716 (-0.300210)

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) 0.091907 / 0.018006 (0.073900) 0.307152 / 0.000490 (0.306663) 0.000223 / 0.000200 (0.000023) 0.000046 / 0.000054 (-0.000009)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.021909 / 0.037411 (-0.015502) 0.070676 / 0.014526 (0.056150) 0.081651 / 0.176557 (-0.094906) 0.120915 / 0.737135 (-0.616220) 0.085882 / 0.296338 (-0.210456)

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) 0.288008 / 0.215209 (0.072799) 2.861352 / 2.077655 (0.783697) 1.539045 / 1.504120 (0.034925) 1.412175 / 1.541195 (-0.129019) 1.421236 / 1.468490 (-0.047254) 0.404921 / 4.584777 (-4.179856) 2.480211 / 3.745712 (-1.265501) 2.473083 / 5.269862 (-2.796779) 1.558894 / 4.565676 (-3.006783) 0.046692 / 0.424275 (-0.377584) 0.004802 / 0.007607 (-0.002805) 0.346046 / 0.226044 (0.120001) 3.464387 / 2.268929 (1.195459) 1.937298 / 55.444624 (-53.507326) 1.593701 / 6.876477 (-5.282776) 1.730688 / 2.142072 (-0.411385) 0.481069 / 4.805227 (-4.324158) 0.098991 / 6.500664 (-6.401673) 0.040491 / 0.075469 (-0.034978)

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) 0.967809 / 1.841788 (-0.873979) 11.952335 / 8.074308 (3.878027) 10.616711 / 10.191392 (0.425319) 0.128938 / 0.680424 (-0.551486) 0.015455 / 0.534201 (-0.518746) 0.272100 / 0.579283 (-0.307183) 0.278275 / 0.434364 (-0.156089) 0.309711 / 0.540337 (-0.230627) 0.411026 / 1.386936 (-0.975910)

mariosasko
mariosasko approved these changes Nov 27, 2023
View reviewed changes
Copy link
Collaborator

@mariosasko mariosasko left a comment

Choose a reason for hiding this comment

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

LGTM, thanks!

@lhoestq lhoestq merged commit d7bbf34 into main Nov 28, 2023
@lhoestq lhoestq deleted the add-trust-remote-code branch November 28, 2023 16:03
@github-actions
Copy link

github-actions bot commented Nov 28, 2023

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) 0.008470 / 0.011353 (-0.002883) 0.003201 / 0.011008 (-0.007808) 0.063193 / 0.038508 (0.024685) 0.064174 / 0.023109 (0.041064) 0.248316 / 0.275898 (-0.027582) 0.281598 / 0.323480 (-0.041882) 0.004076 / 0.007986 (-0.003909) 0.002397 / 0.004328 (-0.001932) 0.048834 / 0.004250 (0.044584) 0.056517 / 0.037052 (0.019465) 0.254164 / 0.258489 (-0.004326) 0.289800 / 0.293841 (-0.004041) 0.031092 / 0.128546 (-0.097454) 0.010885 / 0.075646 (-0.064762) 0.219198 / 0.419271 (-0.200073) 0.040087 / 0.043533 (-0.003446) 0.250900 / 0.255139 (-0.004239) 0.267787 / 0.283200 (-0.015413) 0.019666 / 0.141683 (-0.122017) 1.114960 / 1.452155 (-0.337194) 1.266675 / 1.492716 (-0.226041)

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) 0.091429 / 0.018006 (0.073422) 0.301804 / 0.000490 (0.301314) 0.000212 / 0.000200 (0.000012) 0.000064 / 0.000054 (0.000009)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.021053 / 0.037411 (-0.016358) 0.062407 / 0.014526 (0.047881) 0.073166 / 0.176557 (-0.103391) 0.119642 / 0.737135 (-0.617493) 0.074771 / 0.296338 (-0.221567)

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) 0.278582 / 0.215209 (0.063373) 2.773023 / 2.077655 (0.695368) 1.459977 / 1.504120 (-0.044143) 1.330453 / 1.541195 (-0.210742) 1.372797 / 1.468490 (-0.095693) 0.628845 / 4.584777 (-3.955932) 3.428779 / 3.745712 (-0.316933) 3.138967 / 5.269862 (-2.130895) 2.126891 / 4.565676 (-2.438785) 0.062340 / 0.424275 (-0.361935) 0.004939 / 0.007607 (-0.002668) 0.336058 / 0.226044 (0.110014) 3.463741 / 2.268929 (1.194813) 1.847504 / 55.444624 (-53.597120) 1.984173 / 6.876477 (-4.892304) 1.602962 / 2.142072 (-0.539110) 0.637683 / 4.805227 (-4.167545) 0.117898 / 6.500664 (-6.382766) 0.043308 / 0.075469 (-0.032161)

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) 1.087773 / 1.841788 (-0.754014) 14.959526 / 8.074308 (6.885218) 10.886003 / 10.191392 (0.694611) 0.163385 / 0.680424 (-0.517039) 0.016679 / 0.534201 (-0.517522) 0.351913 / 0.579283 (-0.227370) 0.359007 / 0.434364 (-0.075357) 0.323824 / 0.540337 (-0.216513) 0.549268 / 1.386936 (-0.837668)
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) 0.005265 / 0.011353 (-0.006088) 0.003367 / 0.011008 (-0.007641) 0.062741 / 0.038508 (0.024233) 0.068463 / 0.023109 (0.045354) 0.258497 / 0.275898 (-0.017401) 0.355360 / 0.323480 (0.031880) 0.003910 / 0.007986 (-0.004075) 0.002399 / 0.004328 (-0.001929) 0.055564 / 0.004250 (0.051313) 0.039644 / 0.037052 (0.002591) 0.258313 / 0.258489 (-0.000176) 0.328927 / 0.293841 (0.035086) 0.035634 / 0.128546 (-0.092912) 0.010378 / 0.075646 (-0.065268) 0.073109 / 0.419271 (-0.346163) 0.039752 / 0.043533 (-0.003781) 0.258237 / 0.255139 (0.003098) 0.330329 / 0.283200 (0.047129) 0.023924 / 0.141683 (-0.117759) 1.198639 / 1.452155 (-0.253515) 1.202307 / 1.492716 (-0.290409)

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) 0.091297 / 0.018006 (0.073290) 0.298729 / 0.000490 (0.298240) 0.000210 / 0.000200 (0.000010) 0.000049 / 0.000054 (-0.000005)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.022381 / 0.037411 (-0.015030) 0.070226 / 0.014526 (0.055700) 0.080549 / 0.176557 (-0.096007) 0.119677 / 0.737135 (-0.617458) 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) 0.289270 / 0.215209 (0.074061) 2.853830 / 2.077655 (0.776175) 1.528938 / 1.504120 (0.024818) 1.398429 / 1.541195 (-0.142766) 1.472465 / 1.468490 (0.003975) 0.779015 / 4.584777 (-3.805762) 3.287724 / 3.745712 (-0.457988) 3.020908 / 5.269862 (-2.248953) 1.926094 / 4.565676 (-2.639583) 0.063163 / 0.424275 (-0.361112) 0.005175 / 0.007607 (-0.002432) 0.342884 / 0.226044 (0.116840) 3.476837 / 2.268929 (1.207908) 1.880683 / 55.444624 (-53.563942) 1.613845 / 6.876477 (-5.262632) 1.624734 / 2.142072 (-0.517338) 0.626220 / 4.805227 (-4.179007) 0.114976 / 6.500664 (-6.385689) 0.040670 / 0.075469 (-0.034799)

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) 1.116815 / 1.841788 (-0.724973) 15.388426 / 8.074308 (7.314118) 10.825276 / 10.191392 (0.633884) 0.172659 / 0.680424 (-0.507765) 0.015468 / 0.534201 (-0.518733) 0.285552 / 0.579283 (-0.293731) 0.346886 / 0.434364 (-0.087478) 0.348696 / 0.540337 (-0.191641) 0.729335 / 1.386936 (-0.657601)

@veekaybee veekaybee mentioned this pull request Feb 23, 2024
Closed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@albertvillanova albertvillanova albertvillanova left review comments

@mariosasko mariosasko mariosasko 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.

5 participants

@lhoestq @HuggingFaceDocBuilderDev @albertvillanova @mariosasko