Skip to content
New issue

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

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

Misc improvements #6004

Merged
merged 4 commits into from
Jul 6, 2023
Merged

Misc improvements #6004

merged 4 commits into from
Jul 6, 2023

Conversation

mariosasko
Copy link
Collaborator

Contains the following improvements:

  • fixes a "share dataset" link in README and modifies the "hosting" part in the disclaimer section
  • updates Makefile to also run the style checks on utils and setup.py
  • deletes a test for GH-hosted datasets (no longer supported)
  • deletes convert_dataset.sh (outdated)
  • aligns utils/release.py with transformers (the current version is outdated)

@github-actions
Copy link

github-actions bot commented Jul 3, 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) 0.006897 / 0.011353 (-0.004456) 0.004207 / 0.011008 (-0.006802) 0.104828 / 0.038508 (0.066320) 0.048054 / 0.023109 (0.024945) 0.373991 / 0.275898 (0.098093) 0.426740 / 0.323480 (0.103260) 0.005540 / 0.007986 (-0.002446) 0.003531 / 0.004328 (-0.000797) 0.079304 / 0.004250 (0.075053) 0.066996 / 0.037052 (0.029944) 0.370675 / 0.258489 (0.112186) 0.414154 / 0.293841 (0.120313) 0.031567 / 0.128546 (-0.096979) 0.008843 / 0.075646 (-0.066803) 0.357426 / 0.419271 (-0.061845) 0.067040 / 0.043533 (0.023508) 0.362384 / 0.255139 (0.107245) 0.376056 / 0.283200 (0.092856) 0.032985 / 0.141683 (-0.108697) 1.560603 / 1.452155 (0.108448) 1.619024 / 1.492716 (0.126308)

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) 0.229059 / 0.018006 (0.211053) 0.440513 / 0.000490 (0.440023) 0.004647 / 0.000200 (0.004447) 0.000085 / 0.000054 (0.000030)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.029517 / 0.037411 (-0.007894) 0.120974 / 0.014526 (0.106448) 0.125070 / 0.176557 (-0.051486) 0.184695 / 0.737135 (-0.552441) 0.130244 / 0.296338 (-0.166095)

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) 0.436930 / 0.215209 (0.221721) 4.356118 / 2.077655 (2.278463) 2.049169 / 1.504120 (0.545049) 1.842898 / 1.541195 (0.301703) 1.918948 / 1.468490 (0.450458) 0.553573 / 4.584777 (-4.031204) 3.883195 / 3.745712 (0.137483) 3.209780 / 5.269862 (-2.060081) 1.551707 / 4.565676 (-3.013970) 0.068181 / 0.424275 (-0.356094) 0.012370 / 0.007607 (0.004762) 0.539899 / 0.226044 (0.313854) 5.380008 / 2.268929 (3.111079) 2.518178 / 55.444624 (-52.926446) 2.174190 / 6.876477 (-4.702286) 2.317812 / 2.142072 (0.175740) 0.674154 / 4.805227 (-4.131073) 0.149313 / 6.500664 (-6.351351) 0.068297 / 0.075469 (-0.007172)

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) 1.261426 / 1.841788 (-0.580362) 15.316378 / 8.074308 (7.242070) 13.573512 / 10.191392 (3.382120) 0.190022 / 0.680424 (-0.490401) 0.018697 / 0.534201 (-0.515504) 0.448122 / 0.579283 (-0.131161) 0.435044 / 0.434364 (0.000681) 0.550065 / 0.540337 (0.009728) 0.653547 / 1.386936 (-0.733389)
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) 0.007116 / 0.011353 (-0.004237) 0.004375 / 0.011008 (-0.006633) 0.081793 / 0.038508 (0.043285) 0.047980 / 0.023109 (0.024871) 0.392185 / 0.275898 (0.116287) 0.462263 / 0.323480 (0.138783) 0.005574 / 0.007986 (-0.002412) 0.003552 / 0.004328 (-0.000776) 0.080413 / 0.004250 (0.076162) 0.065539 / 0.037052 (0.028487) 0.413137 / 0.258489 (0.154648) 0.467377 / 0.293841 (0.173536) 0.034386 / 0.128546 (-0.094160) 0.009183 / 0.075646 (-0.066464) 0.087542 / 0.419271 (-0.331730) 0.053954 / 0.043533 (0.010421) 0.385096 / 0.255139 (0.129957) 0.404900 / 0.283200 (0.121701) 0.025908 / 0.141683 (-0.115775) 1.550159 / 1.452155 (0.098005) 1.598794 / 1.492716 (0.106078)

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) 0.246222 / 0.018006 (0.228216) 0.441095 / 0.000490 (0.440605) 0.006863 / 0.000200 (0.006663) 0.000109 / 0.000054 (0.000055)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.032179 / 0.037411 (-0.005233) 0.120112 / 0.014526 (0.105586) 0.129326 / 0.176557 (-0.047230) 0.184542 / 0.737135 (-0.552593) 0.135038 / 0.296338 (-0.161300)

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) 0.459002 / 0.215209 (0.243793) 4.580258 / 2.077655 (2.502604) 2.296689 / 1.504120 (0.792569) 2.104338 / 1.541195 (0.563143) 2.182896 / 1.468490 (0.714406) 0.546447 / 4.584777 (-4.038330) 3.854047 / 3.745712 (0.108335) 1.873829 / 5.269862 (-3.396032) 1.116484 / 4.565676 (-3.449193) 0.067158 / 0.424275 (-0.357117) 0.012035 / 0.007607 (0.004428) 0.556642 / 0.226044 (0.330597) 5.574436 / 2.268929 (3.305508) 2.828223 / 55.444624 (-52.616402) 2.519851 / 6.876477 (-4.356626) 2.668594 / 2.142072 (0.526521) 0.675989 / 4.805227 (-4.129238) 0.146075 / 6.500664 (-6.354589) 0.067788 / 0.075469 (-0.007681)

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) 1.345958 / 1.841788 (-0.495830) 15.672748 / 8.074308 (7.598440) 14.937583 / 10.191392 (4.746191) 0.163479 / 0.680424 (-0.516945) 0.018364 / 0.534201 (-0.515837) 0.433296 / 0.579283 (-0.145987) 0.432463 / 0.434364 (-0.001901) 0.512000 / 0.540337 (-0.028338) 0.619397 / 1.386936 (-0.767539)

@HuggingFaceDocBuilderDev
Copy link

HuggingFaceDocBuilderDev commented Jul 3, 2023

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

@mariosasko mariosasko requested a review from lhoestq July 6, 2023 15:23
Copy link
Member

@lhoestq lhoestq left a comment

Choose a reason for hiding this comment

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

LGTM

utils/release.py Outdated Show resolved Hide resolved
Co-authored-by: Quentin Lhoest <42851186+lhoestq@users.noreply.github.com>
@github-actions
Copy link

github-actions bot commented Jul 6, 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) 0.010097 / 0.011353 (-0.001256) 0.005070 / 0.011008 (-0.005939) 0.118638 / 0.038508 (0.080130) 0.043651 / 0.023109 (0.020542) 0.356074 / 0.275898 (0.080176) 0.414578 / 0.323480 (0.091098) 0.005939 / 0.007986 (-0.002046) 0.004927 / 0.004328 (0.000598) 0.089545 / 0.004250 (0.085294) 0.067533 / 0.037052 (0.030481) 0.371550 / 0.258489 (0.113061) 0.417808 / 0.293841 (0.123967) 0.045186 / 0.128546 (-0.083361) 0.015763 / 0.075646 (-0.059883) 0.393304 / 0.419271 (-0.025967) 0.065123 / 0.043533 (0.021591) 0.345057 / 0.255139 (0.089918) 0.378809 / 0.283200 (0.095610) 0.033243 / 0.141683 (-0.108440) 1.679956 / 1.452155 (0.227802) 1.775456 / 1.492716 (0.282739)

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) 0.229723 / 0.018006 (0.211717) 0.554630 / 0.000490 (0.554140) 0.008729 / 0.000200 (0.008529) 0.000183 / 0.000054 (0.000129)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.027284 / 0.037411 (-0.010128) 0.114741 / 0.014526 (0.100215) 0.129188 / 0.176557 (-0.047369) 0.189270 / 0.737135 (-0.547866) 0.126000 / 0.296338 (-0.170339)

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) 0.580417 / 0.215209 (0.365208) 5.829337 / 2.077655 (3.751683) 2.421191 / 1.504120 (0.917071) 2.063673 / 1.541195 (0.522479) 2.133427 / 1.468490 (0.664937) 0.830964 / 4.584777 (-3.753813) 5.107139 / 3.745712 (1.361427) 4.599451 / 5.269862 (-0.670410) 2.406502 / 4.565676 (-2.159175) 0.100422 / 0.424275 (-0.323853) 0.011850 / 0.007607 (0.004243) 0.741881 / 0.226044 (0.515836) 7.425689 / 2.268929 (5.156760) 3.068948 / 55.444624 (-52.375676) 2.496292 / 6.876477 (-4.380184) 2.566420 / 2.142072 (0.424348) 1.093084 / 4.805227 (-3.712144) 0.224106 / 6.500664 (-6.276558) 0.084549 / 0.075469 (0.009080)

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) 1.416315 / 1.841788 (-0.425473) 16.306901 / 8.074308 (8.232593) 19.792419 / 10.191392 (9.601027) 0.224223 / 0.680424 (-0.456201) 0.026385 / 0.534201 (-0.507816) 0.463460 / 0.579283 (-0.115823) 0.598385 / 0.434364 (0.164021) 0.543981 / 0.540337 (0.003644) 0.647454 / 1.386936 (-0.739482)
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) 0.009470 / 0.011353 (-0.001883) 0.004800 / 0.011008 (-0.006208) 0.094276 / 0.038508 (0.055768) 0.045157 / 0.023109 (0.022048) 0.397302 / 0.275898 (0.121404) 0.474213 / 0.323480 (0.150733) 0.005826 / 0.007986 (-0.002160) 0.003724 / 0.004328 (-0.000605) 0.090060 / 0.004250 (0.085809) 0.066671 / 0.037052 (0.029618) 0.439560 / 0.258489 (0.181071) 0.468598 / 0.293841 (0.174757) 0.044549 / 0.128546 (-0.083997) 0.014000 / 0.075646 (-0.061646) 0.110457 / 0.419271 (-0.308815) 0.065898 / 0.043533 (0.022365) 0.408101 / 0.255139 (0.152962) 0.433473 / 0.283200 (0.150273) 0.038438 / 0.141683 (-0.103245) 1.767781 / 1.452155 (0.315626) 1.791575 / 1.492716 (0.298859)

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) 0.230257 / 0.018006 (0.212251) 0.492280 / 0.000490 (0.491790) 0.005110 / 0.000200 (0.004910) 0.000119 / 0.000054 (0.000065)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.028854 / 0.037411 (-0.008557) 0.111702 / 0.014526 (0.097176) 0.122040 / 0.176557 (-0.054517) 0.179103 / 0.737135 (-0.558032) 0.128869 / 0.296338 (-0.167470)

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) 0.634795 / 0.215209 (0.419586) 6.204760 / 2.077655 (4.127105) 2.692479 / 1.504120 (1.188359) 2.324260 / 1.541195 (0.783066) 2.380640 / 1.468490 (0.912149) 0.887827 / 4.584777 (-3.696950) 5.251648 / 3.745712 (1.505935) 2.632767 / 5.269862 (-2.637095) 1.745721 / 4.565676 (-2.819955) 0.108364 / 0.424275 (-0.315911) 0.013409 / 0.007607 (0.005802) 0.783427 / 0.226044 (0.557383) 7.765144 / 2.268929 (5.496216) 3.340686 / 55.444624 (-52.103938) 2.715340 / 6.876477 (-4.161137) 2.768604 / 2.142072 (0.626531) 1.119746 / 4.805227 (-3.685481) 0.210804 / 6.500664 (-6.289860) 0.072600 / 0.075469 (-0.002869)

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) 1.517334 / 1.841788 (-0.324454) 17.046837 / 8.074308 (8.972529) 19.371090 / 10.191392 (9.179698) 0.194275 / 0.680424 (-0.486148) 0.026712 / 0.534201 (-0.507488) 0.462731 / 0.579283 (-0.116552) 0.568958 / 0.434364 (0.134595) 0.555707 / 0.540337 (0.015370) 0.663654 / 1.386936 (-0.723283)

@mariosasko mariosasko merged commit 3e34d06 into main Jul 6, 2023
@mariosasko mariosasko deleted the misc-improvement branch July 6, 2023 16:55
@github-actions
Copy link

github-actions bot commented Jul 6, 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) 0.006423 / 0.011353 (-0.004930) 0.003882 / 0.011008 (-0.007126) 0.082976 / 0.038508 (0.044468) 0.071281 / 0.023109 (0.048171) 0.311367 / 0.275898 (0.035469) 0.348228 / 0.323480 (0.024748) 0.005315 / 0.007986 (-0.002671) 0.003326 / 0.004328 (-0.001003) 0.064641 / 0.004250 (0.060391) 0.056134 / 0.037052 (0.019081) 0.314071 / 0.258489 (0.055582) 0.360534 / 0.293841 (0.066693) 0.030642 / 0.128546 (-0.097904) 0.008301 / 0.075646 (-0.067345) 0.285820 / 0.419271 (-0.133451) 0.069241 / 0.043533 (0.025708) 0.313995 / 0.255139 (0.058856) 0.336656 / 0.283200 (0.053457) 0.031686 / 0.141683 (-0.109997) 1.467627 / 1.452155 (0.015472) 1.536493 / 1.492716 (0.043777)

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) 0.196518 / 0.018006 (0.178512) 0.458235 / 0.000490 (0.457745) 0.005599 / 0.000200 (0.005399) 0.000088 / 0.000054 (0.000034)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.027371 / 0.037411 (-0.010040) 0.080986 / 0.014526 (0.066460) 0.093296 / 0.176557 (-0.083260) 0.150592 / 0.737135 (-0.586543) 0.094150 / 0.296338 (-0.202188)

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) 0.379412 / 0.215209 (0.164202) 3.797927 / 2.077655 (1.720272) 1.830654 / 1.504120 (0.326534) 1.669569 / 1.541195 (0.128374) 1.746738 / 1.468490 (0.278248) 0.479536 / 4.584777 (-4.105241) 3.592867 / 3.745712 (-0.152845) 5.468098 / 5.269862 (0.198237) 3.268013 / 4.565676 (-1.297663) 0.056635 / 0.424275 (-0.367640) 0.007224 / 0.007607 (-0.000383) 0.456681 / 0.226044 (0.230636) 4.566736 / 2.268929 (2.297807) 2.362831 / 55.444624 (-53.081793) 1.965141 / 6.876477 (-4.911336) 2.156905 / 2.142072 (0.014833) 0.572543 / 4.805227 (-4.232684) 0.132203 / 6.500664 (-6.368461) 0.059254 / 0.075469 (-0.016215)

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) 1.256134 / 1.841788 (-0.585654) 19.905438 / 8.074308 (11.831130) 14.179556 / 10.191392 (3.988164) 0.168043 / 0.680424 (-0.512381) 0.018215 / 0.534201 (-0.515986) 0.392740 / 0.579283 (-0.186543) 0.398397 / 0.434364 (-0.035967) 0.463806 / 0.540337 (-0.076531) 0.616248 / 1.386936 (-0.770688)
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) 0.006564 / 0.011353 (-0.004789) 0.003923 / 0.011008 (-0.007085) 0.063929 / 0.038508 (0.025421) 0.073780 / 0.023109 (0.050671) 0.360242 / 0.275898 (0.084344) 0.395078 / 0.323480 (0.071598) 0.005265 / 0.007986 (-0.002720) 0.003229 / 0.004328 (-0.001100) 0.064094 / 0.004250 (0.059843) 0.057468 / 0.037052 (0.020416) 0.369530 / 0.258489 (0.111041) 0.411159 / 0.293841 (0.117318) 0.031278 / 0.128546 (-0.097268) 0.008424 / 0.075646 (-0.067222) 0.070411 / 0.419271 (-0.348860) 0.048714 / 0.043533 (0.005181) 0.361280 / 0.255139 (0.106141) 0.382468 / 0.283200 (0.099269) 0.023059 / 0.141683 (-0.118624) 1.452369 / 1.452155 (0.000215) 1.519192 / 1.492716 (0.026475)

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) 0.223745 / 0.018006 (0.205739) 0.442086 / 0.000490 (0.441596) 0.000379 / 0.000200 (0.000179) 0.000055 / 0.000054 (0.000001)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.030919 / 0.037411 (-0.006493) 0.088483 / 0.014526 (0.073958) 0.101165 / 0.176557 (-0.075391) 0.154332 / 0.737135 (-0.582804) 0.103030 / 0.296338 (-0.193309)

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) 0.414520 / 0.215209 (0.199311) 4.126754 / 2.077655 (2.049099) 2.142677 / 1.504120 (0.638557) 1.995300 / 1.541195 (0.454106) 2.101678 / 1.468490 (0.633188) 0.481099 / 4.584777 (-4.103678) 3.562813 / 3.745712 (-0.182900) 3.392463 / 5.269862 (-1.877399) 1.983943 / 4.565676 (-2.581734) 0.056594 / 0.424275 (-0.367681) 0.007216 / 0.007607 (-0.000391) 0.495085 / 0.226044 (0.269041) 4.955640 / 2.268929 (2.686712) 2.629434 / 55.444624 (-52.815191) 2.269577 / 6.876477 (-4.606900) 2.357708 / 2.142072 (0.215635) 0.612370 / 4.805227 (-4.192857) 0.131169 / 6.500664 (-6.369495) 0.061029 / 0.075469 (-0.014440)

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) 1.339438 / 1.841788 (-0.502350) 19.757611 / 8.074308 (11.683303) 14.246254 / 10.191392 (4.054862) 0.170750 / 0.680424 (-0.509674) 0.018192 / 0.534201 (-0.516009) 0.395693 / 0.579283 (-0.183590) 0.411003 / 0.434364 (-0.023361) 0.478531 / 0.540337 (-0.061806) 0.650291 / 1.386936 (-0.736645)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants