Fix benchmarks CI - pin protobuf

[lhoestq]

fix https://github.com/huggingface/datasets/actions/runs/4156059127/jobs/7189576331

[HuggingFaceDocBuilderDev]

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

[github-actions]

Show benchmarks

PyArrow==6.0.0

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)	0.011142 / 0.011353 (-0.000211)	0.005885 / 0.011008 (-0.005123)	0.115374 / 0.038508 (0.076866)	0.041704 / 0.023109 (0.018594)	0.356996 / 0.275898 (0.081098)	0.395076 / 0.323480 (0.071596)	0.008726 / 0.007986 (0.000740)	0.005528 / 0.004328 (0.001199)	0.087817 / 0.004250 (0.083566)	0.049273 / 0.037052 (0.012221)	0.363778 / 0.258489 (0.105289)	0.408801 / 0.293841 (0.114960)	0.045232 / 0.128546 (-0.083314)	0.013788 / 0.075646 (-0.061859)	0.395634 / 0.419271 (-0.023637)	0.056583 / 0.043533 (0.013051)	0.360779 / 0.255139 (0.105640)	0.386843 / 0.283200 (0.103643)	0.116632 / 0.141683 (-0.025051)	1.830020 / 1.452155 (0.377865)	1.808720 / 1.492716 (0.316003)

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)	0.221029 / 0.018006 (0.203023)	0.489463 / 0.000490 (0.488973)	0.002104 / 0.000200 (0.001904)	0.000098 / 0.000054 (0.000043)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.032873 / 0.037411 (-0.004539)	0.129526 / 0.014526 (0.115000)	0.141446 / 0.176557 (-0.035111)	0.189222 / 0.737135 (-0.547913)	0.149329 / 0.296338 (-0.147010)

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)	0.471389 / 0.215209 (0.256180)	4.710174 / 2.077655 (2.632519)	2.239122 / 1.504120 (0.735002)	1.977789 / 1.541195 (0.436595)	2.107336 / 1.468490 (0.638846)	0.816852 / 4.584777 (-3.767925)	4.944056 / 3.745712 (1.198344)	4.637939 / 5.269862 (-0.631922)	2.355546 / 4.565676 (-2.210131)	0.099324 / 0.424275 (-0.324951)	0.014529 / 0.007607 (0.006922)	0.596322 / 0.226044 (0.370277)	5.972216 / 2.268929 (3.703287)	2.697281 / 55.444624 (-52.747344)	2.293836 / 6.876477 (-4.582641)	2.380271 / 2.142072 (0.238199)	1.001307 / 4.805227 (-3.803920)	0.196981 / 6.500664 (-6.303683)	0.074390 / 0.075469 (-0.001079)

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)	1.482915 / 1.841788 (-0.358872)	18.739511 / 8.074308 (10.665202)	16.768191 / 10.191392 (6.576799)	0.203163 / 0.680424 (-0.477261)	0.037514 / 0.534201 (-0.496687)	0.529017 / 0.579283 (-0.050266)	0.577591 / 0.434364 (0.143227)	0.634057 / 0.540337 (0.093720)	0.759812 / 1.386936 (-0.627124)

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)	0.008815 / 0.011353 (-0.002537)	0.005956 / 0.011008 (-0.005052)	0.087912 / 0.038508 (0.049404)	0.040291 / 0.023109 (0.017182)	0.404079 / 0.275898 (0.128181)	0.447309 / 0.323480 (0.123829)	0.006515 / 0.007986 (-0.001471)	0.005917 / 0.004328 (0.001588)	0.085560 / 0.004250 (0.081310)	0.057077 / 0.037052 (0.020025)	0.403349 / 0.258489 (0.144860)	0.465644 / 0.293841 (0.171803)	0.043530 / 0.128546 (-0.085016)	0.014234 / 0.075646 (-0.061412)	0.102203 / 0.419271 (-0.317068)	0.058335 / 0.043533 (0.014802)	0.398488 / 0.255139 (0.143349)	0.424127 / 0.283200 (0.140927)	0.119058 / 0.141683 (-0.022625)	1.748748 / 1.452155 (0.296593)	1.822190 / 1.492716 (0.329474)

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)	0.255782 / 0.018006 (0.237776)	0.496665 / 0.000490 (0.496176)	0.000471 / 0.000200 (0.000271)	0.000069 / 0.000054 (0.000014)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.034111 / 0.037411 (-0.003301)	0.131442 / 0.014526 (0.116917)	0.144660 / 0.176557 (-0.031897)	0.188156 / 0.737135 (-0.548979)	0.149875 / 0.296338 (-0.146463)

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)	0.502218 / 0.215209 (0.287009)	5.004486 / 2.077655 (2.926832)	2.420379 / 1.504120 (0.916259)	2.194671 / 1.541195 (0.653476)	2.306376 / 1.468490 (0.837886)	0.856623 / 4.584777 (-3.728154)	4.963211 / 3.745712 (1.217499)	2.517965 / 5.269862 (-2.751896)	1.743880 / 4.565676 (-2.821797)	0.105270 / 0.424275 (-0.319005)	0.014725 / 0.007607 (0.007118)	0.621934 / 0.226044 (0.395890)	6.183827 / 2.268929 (3.914898)	2.945868 / 55.444624 (-52.498757)	2.557676 / 6.876477 (-4.318801)	2.622282 / 2.142072 (0.480210)	1.011647 / 4.805227 (-3.793580)	0.199573 / 6.500664 (-6.301091)	0.076283 / 0.075469 (0.000814)

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)	1.518813 / 1.841788 (-0.322975)	18.833017 / 8.074308 (10.758709)	16.095249 / 10.191392 (5.903857)	0.196667 / 0.680424 (-0.483757)	0.022060 / 0.534201 (-0.512141)	0.537802 / 0.579283 (-0.041481)	0.523676 / 0.434364 (0.089312)	0.629387 / 0.540337 (0.089049)	0.738042 / 1.386936 (-0.648894)

[albertvillanova]

Thanks for the fix, @lhoestq: this pin indeed fixes our CI benchmarks.

I would merge it as it is, as a hotfix, so that our CI becomes green.

There is one thing that does not make sense to me: protobuf-4.21.12 was released on 14/12/2022 and we are getting our CI error just since 10/02/2023...

[github-actions]

Show benchmarks

PyArrow==6.0.0

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)	0.008608 / 0.011353 (-0.002745)	0.004553 / 0.011008 (-0.006455)	0.100031 / 0.038508 (0.061523)	0.029498 / 0.023109 (0.006389)	0.306913 / 0.275898 (0.031015)	0.367369 / 0.323480 (0.043889)	0.006883 / 0.007986 (-0.001103)	0.004768 / 0.004328 (0.000440)	0.077424 / 0.004250 (0.073173)	0.034005 / 0.037052 (-0.003047)	0.317772 / 0.258489 (0.059283)	0.356859 / 0.293841 (0.063018)	0.033717 / 0.128546 (-0.094829)	0.011386 / 0.075646 (-0.064260)	0.322832 / 0.419271 (-0.096439)	0.043930 / 0.043533 (0.000397)	0.308087 / 0.255139 (0.052948)	0.338349 / 0.283200 (0.055149)	0.094780 / 0.141683 (-0.046903)	1.463454 / 1.452155 (0.011300)	1.495055 / 1.492716 (0.002338)

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)	0.191039 / 0.018006 (0.173033)	0.414650 / 0.000490 (0.414160)	0.002435 / 0.000200 (0.002235)	0.000075 / 0.000054 (0.000021)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.023871 / 0.037411 (-0.013540)	0.097140 / 0.014526 (0.082614)	0.105914 / 0.176557 (-0.070643)	0.147375 / 0.737135 (-0.589760)	0.107985 / 0.296338 (-0.188354)

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)	0.420174 / 0.215209 (0.204965)	4.208354 / 2.077655 (2.130700)	1.904568 / 1.504120 (0.400448)	1.687406 / 1.541195 (0.146212)	1.723901 / 1.468490 (0.255411)	0.693554 / 4.584777 (-3.891223)	3.445474 / 3.745712 (-0.300238)	1.904919 / 5.269862 (-3.364943)	1.284378 / 4.565676 (-3.281298)	0.082539 / 0.424275 (-0.341736)	0.012490 / 0.007607 (0.004883)	0.527778 / 0.226044 (0.301733)	5.300766 / 2.268929 (3.031838)	2.324666 / 55.444624 (-53.119958)	1.977166 / 6.876477 (-4.899311)	2.054396 / 2.142072 (-0.087677)	0.820966 / 4.805227 (-3.984261)	0.148584 / 6.500664 (-6.352080)	0.063618 / 0.075469 (-0.011851)

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)	1.188075 / 1.841788 (-0.653712)	13.706950 / 8.074308 (5.632642)	13.725122 / 10.191392 (3.533730)	0.167379 / 0.680424 (-0.513045)	0.028729 / 0.534201 (-0.505472)	0.395373 / 0.579283 (-0.183910)	0.403604 / 0.434364 (-0.030760)	0.464290 / 0.540337 (-0.076047)	0.553792 / 1.386936 (-0.833144)

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)	0.006565 / 0.011353 (-0.004787)	0.004588 / 0.011008 (-0.006420)	0.077312 / 0.038508 (0.038804)	0.027348 / 0.023109 (0.004239)	0.367753 / 0.275898 (0.091855)	0.403250 / 0.323480 (0.079770)	0.005201 / 0.007986 (-0.002785)	0.004695 / 0.004328 (0.000366)	0.076203 / 0.004250 (0.071953)	0.039388 / 0.037052 (0.002336)	0.374418 / 0.258489 (0.115929)	0.413623 / 0.293841 (0.119782)	0.031731 / 0.128546 (-0.096815)	0.011644 / 0.075646 (-0.064002)	0.086339 / 0.419271 (-0.332932)	0.048902 / 0.043533 (0.005369)	0.352064 / 0.255139 (0.096925)	0.386637 / 0.283200 (0.103437)	0.093662 / 0.141683 (-0.048021)	1.479863 / 1.452155 (0.027709)	1.562475 / 1.492716 (0.069758)

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)	0.231874 / 0.018006 (0.213867)	0.402185 / 0.000490 (0.401695)	0.005252 / 0.000200 (0.005052)	0.000086 / 0.000054 (0.000032)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.025402 / 0.037411 (-0.012010)	0.099896 / 0.014526 (0.085370)	0.106365 / 0.176557 (-0.070192)	0.143309 / 0.737135 (-0.593827)	0.112311 / 0.296338 (-0.184027)

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)	0.447637 / 0.215209 (0.232428)	4.469337 / 2.077655 (2.391682)	2.164332 / 1.504120 (0.660212)	1.957826 / 1.541195 (0.416631)	1.984580 / 1.468490 (0.516090)	0.702909 / 4.584777 (-3.881868)	3.361725 / 3.745712 (-0.383987)	2.818102 / 5.269862 (-2.451760)	1.589815 / 4.565676 (-2.975862)	0.083647 / 0.424275 (-0.340628)	0.012502 / 0.007607 (0.004895)	0.545578 / 0.226044 (0.319534)	5.480894 / 2.268929 (3.211966)	2.605599 / 55.444624 (-52.839026)	2.253444 / 6.876477 (-4.623032)	2.289818 / 2.142072 (0.147746)	0.803680 / 4.805227 (-4.001547)	0.151870 / 6.500664 (-6.348794)	0.066610 / 0.075469 (-0.008859)

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)	1.327390 / 1.841788 (-0.514398)	14.046936 / 8.074308 (5.972628)	13.643169 / 10.191392 (3.451777)	0.128223 / 0.680424 (-0.552201)	0.016941 / 0.534201 (-0.517260)	0.383887 / 0.579283 (-0.195396)	0.383891 / 0.434364 (-0.050473)	0.440191 / 0.540337 (-0.100146)	0.525357 / 1.386936 (-0.861579)

[lhoestq]

Yea there must have been an update in another package that unconstrained the protobuf dependency - idk which one though

[albertvillanova]

It is tensorboard. I have reported the issue to tensorflow:

• Cannot import tensorflow after latest tensorboard release: TypeError: Descriptors cannot not be created directly tensorflow/tensorflow#59665