Bump huggingface-hub lower version to 0.21.2

[albertvillanova]

This should fix the version compatibility issue when using huggingface_hub < 0.21.2 and latest fsspec (>=2023.12.0).

See my comment: #6687 (comment)

EDIT: the fix has been released in huggingface_hub 0.21.2 - I removed my commits that were using huggingface_hub@main

Please note that people using huggingface_hub < 0.21.2 and latest fsspec will have issues when using datasets:

• https://github.com/huggingface/lighteval/actions/runs/8139147047/job/22241658122?pr=86
• Update huggingface-hub for compatibility with datasets 2.18 lighteval#84

CC: @clefourrier

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[lhoestq]

Thanks for the fix !

[albertvillanova]

@lhoestq if you agree, I could make a patch release tomorrow morning.

[lhoestq]

sure :)

[github-actions]

Show benchmarks

PyArrow==8.0.0

Show updated benchmarks!

Benchmark: benchmark_array_xd.json

metric	read_batch_formatted_as_numpy after write_array2d	read_batch_formatted_as_numpy after write_flattened_sequence	read_batch_formatted_as_numpy after write_nested_sequence	read_batch_unformated after write_array2d	read_batch_unformated after write_flattened_sequence	read_batch_unformated after write_nested_sequence	read_col_formatted_as_numpy after write_array2d	read_col_formatted_as_numpy after write_flattened_sequence	read_col_formatted_as_numpy after write_nested_sequence	read_col_unformated after write_array2d	read_col_unformated after write_flattened_sequence	read_col_unformated after write_nested_sequence	read_formatted_as_numpy after write_array2d	read_formatted_as_numpy after write_flattened_sequence	read_formatted_as_numpy after write_nested_sequence	read_unformated after write_array2d	read_unformated after write_flattened_sequence	read_unformated after write_nested_sequence	write_array2d	write_flattened_sequence	write_nested_sequence
new / old (diff)	0.005086 / 0.011353 (-0.006267)	0.003695 / 0.011008 (-0.007313)	0.063430 / 0.038508 (0.024922)	0.026798 / 0.023109 (0.003689)	0.253761 / 0.275898 (-0.022138)	0.301301 / 0.323480 (-0.022179)	0.004160 / 0.007986 (-0.003825)	0.002783 / 0.004328 (-0.001545)	0.050698 / 0.004250 (0.046448)	0.040899 / 0.037052 (0.003846)	0.269024 / 0.258489 (0.010535)	0.323467 / 0.293841 (0.029626)	0.027756 / 0.128546 (-0.100791)	0.010684 / 0.075646 (-0.064963)	0.207128 / 0.419271 (-0.212144)	0.035874 / 0.043533 (-0.007659)	0.251620 / 0.255139 (-0.003519)	0.268668 / 0.283200 (-0.014532)	0.017387 / 0.141683 (-0.124296)	1.139230 / 1.452155 (-0.312925)	1.183613 / 1.492716 (-0.309103)

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)	0.096337 / 0.018006 (0.078331)	0.305014 / 0.000490 (0.304524)	0.000219 / 0.000200 (0.000019)	0.000050 / 0.000054 (-0.000005)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.018086 / 0.037411 (-0.019325)	0.061626 / 0.014526 (0.047100)	0.072598 / 0.176557 (-0.103959)	0.119944 / 0.737135 (-0.617192)	0.074549 / 0.296338 (-0.221789)

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)	0.282661 / 0.215209 (0.067452)	2.804473 / 2.077655 (0.726818)	1.444602 / 1.504120 (-0.059517)	1.313977 / 1.541195 (-0.227217)	1.319426 / 1.468490 (-0.149064)	0.570176 / 4.584777 (-4.014601)	2.397895 / 3.745712 (-1.347818)	2.760208 / 5.269862 (-2.509654)	1.732457 / 4.565676 (-2.833220)	0.062743 / 0.424275 (-0.361533)	0.004950 / 0.007607 (-0.002657)	0.338500 / 0.226044 (0.112456)	3.287249 / 2.268929 (1.018320)	1.777495 / 55.444624 (-53.667130)	1.521255 / 6.876477 (-5.355222)	1.517317 / 2.142072 (-0.624756)	0.642202 / 4.805227 (-4.163025)	0.116501 / 6.500664 (-6.384163)	0.042418 / 0.075469 (-0.033052)

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)	0.968966 / 1.841788 (-0.872822)	11.490531 / 8.074308 (3.416223)	9.507803 / 10.191392 (-0.683589)	0.141570 / 0.680424 (-0.538854)	0.014000 / 0.534201 (-0.520201)	0.284237 / 0.579283 (-0.295046)	0.269341 / 0.434364 (-0.165022)	0.321654 / 0.540337 (-0.218683)	0.446914 / 1.386936 (-0.940022)

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)	0.005280 / 0.011353 (-0.006072)	0.003794 / 0.011008 (-0.007214)	0.050328 / 0.038508 (0.011820)	0.029756 / 0.023109 (0.006647)	0.273403 / 0.275898 (-0.002495)	0.297346 / 0.323480 (-0.026133)	0.004310 / 0.007986 (-0.003676)	0.002858 / 0.004328 (-0.001470)	0.048833 / 0.004250 (0.044583)	0.045696 / 0.037052 (0.008644)	0.291034 / 0.258489 (0.032545)	0.318899 / 0.293841 (0.025058)	0.029809 / 0.128546 (-0.098737)	0.010710 / 0.075646 (-0.064936)	0.058183 / 0.419271 (-0.361089)	0.051761 / 0.043533 (0.008228)	0.275022 / 0.255139 (0.019883)	0.291614 / 0.283200 (0.008414)	0.017975 / 0.141683 (-0.123708)	1.148489 / 1.452155 (-0.303666)	1.218111 / 1.492716 (-0.274605)

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 (0.073799)	0.299413 / 0.000490 (0.298923)	0.000219 / 0.000200 (0.000019)	0.000044 / 0.000054 (-0.000010)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.021506 / 0.037411 (-0.015905)	0.075537 / 0.014526 (0.061011)	0.087020 / 0.176557 (-0.089536)	0.125270 / 0.737135 (-0.611865)	0.088038 / 0.296338 (-0.208300)

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)	0.300401 / 0.215209 (0.085192)	2.932571 / 2.077655 (0.854916)	1.609502 / 1.504120 (0.105383)	1.480078 / 1.541195 (-0.061117)	1.514902 / 1.468490 (0.046412)	0.575591 / 4.584777 (-4.009186)	2.461873 / 3.745712 (-1.283839)	2.728099 / 5.269862 (-2.541762)	1.760054 / 4.565676 (-2.805622)	0.064371 / 0.424275 (-0.359904)	0.004990 / 0.007607 (-0.002617)	0.350134 / 0.226044 (0.124090)	3.453249 / 2.268929 (1.184321)	1.979760 / 55.444624 (-53.464865)	1.741128 / 6.876477 (-5.135348)	1.825734 / 2.142072 (-0.316339)	0.654902 / 4.805227 (-4.150325)	0.116989 / 6.500664 (-6.383676)	0.040800 / 0.075469 (-0.034669)

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)	1.033352 / 1.841788 (-0.808436)	12.196711 / 8.074308 (4.122403)	10.315114 / 10.191392 (0.123722)	0.132541 / 0.680424 (-0.547882)	0.016455 / 0.534201 (-0.517746)	0.289025 / 0.579283 (-0.290258)	0.281464 / 0.434364 (-0.152900)	0.325302 / 0.540337 (-0.215036)	0.428469 / 1.386936 (-0.958467)