Support data_dir parameter in push_to_hub

[albertvillanova]

Support data_dir parameter in push_to_hub.

This allows users to organize the data files according to their specific needs. For example, "wikimedia/wikipedia" files could be organized by year and/or date, e.g. "2024/20240101/20240101.en".

[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.

[albertvillanova]

@huggingface/datasets, feel free to review this PR so that it can be included in the next release.

[lhoestq]

LGTM !

[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)	0.005125 / 0.011353 (-0.006228)	0.003772 / 0.011008 (-0.007236)	0.063258 / 0.038508 (0.024750)	0.029479 / 0.023109 (0.006370)	0.245554 / 0.275898 (-0.030344)	0.266395 / 0.323480 (-0.057085)	0.003063 / 0.007986 (-0.004922)	0.003298 / 0.004328 (-0.001031)	0.049242 / 0.004250 (0.044991)	0.042390 / 0.037052 (0.005338)	0.258176 / 0.258489 (-0.000313)	0.279935 / 0.293841 (-0.013906)	0.027910 / 0.128546 (-0.100637)	0.011033 / 0.075646 (-0.064613)	0.207763 / 0.419271 (-0.211509)	0.036127 / 0.043533 (-0.007405)	0.247363 / 0.255139 (-0.007776)	0.261309 / 0.283200 (-0.021890)	0.020259 / 0.141683 (-0.121424)	1.152760 / 1.452155 (-0.299395)	1.194853 / 1.492716 (-0.297863)

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)	0.088936 / 0.018006 (0.070930)	0.298396 / 0.000490 (0.297906)	0.000211 / 0.000200 (0.000011)	0.000050 / 0.000054 (-0.000004)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.018434 / 0.037411 (-0.018977)	0.061991 / 0.014526 (0.047466)	0.072786 / 0.176557 (-0.103771)	0.120437 / 0.737135 (-0.616698)	0.078375 / 0.296338 (-0.217964)

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)	0.275821 / 0.215209 (0.060612)	2.703358 / 2.077655 (0.625703)	1.446783 / 1.504120 (-0.057337)	1.333556 / 1.541195 (-0.207639)	1.325753 / 1.468490 (-0.142737)	0.565196 / 4.584777 (-4.019581)	2.411193 / 3.745712 (-1.334520)	2.702764 / 5.269862 (-2.567098)	1.727425 / 4.565676 (-2.838252)	0.062966 / 0.424275 (-0.361309)	0.004985 / 0.007607 (-0.002622)	0.333473 / 0.226044 (0.107428)	3.270615 / 2.268929 (1.001687)	1.822213 / 55.444624 (-53.622411)	1.546572 / 6.876477 (-5.329905)	1.568767 / 2.142072 (-0.573305)	0.655907 / 4.805227 (-4.149321)	0.117173 / 6.500664 (-6.383491)	0.042415 / 0.075469 (-0.033054)

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)	0.987966 / 1.841788 (-0.853822)	11.851206 / 8.074308 (3.776898)	10.327751 / 10.191392 (0.136359)	0.127929 / 0.680424 (-0.552494)	0.013781 / 0.534201 (-0.520420)	0.286910 / 0.579283 (-0.292373)	0.273615 / 0.434364 (-0.160749)	0.323373 / 0.540337 (-0.216965)	0.426407 / 1.386936 (-0.960529)

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)	0.005412 / 0.011353 (-0.005941)	0.003619 / 0.011008 (-0.007389)	0.049603 / 0.038508 (0.011095)	0.031246 / 0.023109 (0.008136)	0.279723 / 0.275898 (0.003825)	0.298557 / 0.323480 (-0.024923)	0.004253 / 0.007986 (-0.003733)	0.002758 / 0.004328 (-0.001570)	0.048931 / 0.004250 (0.044680)	0.044245 / 0.037052 (0.007193)	0.295876 / 0.258489 (0.037387)	0.322720 / 0.293841 (0.028879)	0.046746 / 0.128546 (-0.081800)	0.010841 / 0.075646 (-0.064805)	0.058528 / 0.419271 (-0.360744)	0.034224 / 0.043533 (-0.009308)	0.279192 / 0.255139 (0.024053)	0.299775 / 0.283200 (0.016576)	0.017862 / 0.141683 (-0.123820)	1.154478 / 1.452155 (-0.297677)	1.190483 / 1.492716 (-0.302234)

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)	0.088717 / 0.018006 (0.070710)	0.297905 / 0.000490 (0.297415)	0.000209 / 0.000200 (0.000009)	0.000050 / 0.000054 (-0.000004)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.021458 / 0.037411 (-0.015953)	0.075616 / 0.014526 (0.061090)	0.087080 / 0.176557 (-0.089476)	0.125315 / 0.737135 (-0.611821)	0.088958 / 0.296338 (-0.207381)

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)	0.287085 / 0.215209 (0.071876)	2.807798 / 2.077655 (0.730143)	1.552201 / 1.504120 (0.048081)	1.422374 / 1.541195 (-0.118820)	1.437908 / 1.468490 (-0.030582)	0.569738 / 4.584777 (-4.015039)	2.493921 / 3.745712 (-1.251791)	2.648376 / 5.269862 (-2.621486)	1.741721 / 4.565676 (-2.823955)	0.063023 / 0.424275 (-0.361253)	0.005166 / 0.007607 (-0.002441)	0.336927 / 0.226044 (0.110882)	3.384517 / 2.268929 (1.115588)	1.909888 / 55.444624 (-53.534736)	1.641879 / 6.876477 (-5.234597)	1.727734 / 2.142072 (-0.414338)	0.647127 / 4.805227 (-4.158100)	0.115831 / 6.500664 (-6.384833)	0.041161 / 0.075469 (-0.034309)

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)	1.016310 / 1.841788 (-0.825477)	12.088500 / 8.074308 (4.014192)	10.799730 / 10.191392 (0.608338)	0.129049 / 0.680424 (-0.551375)	0.015379 / 0.534201 (-0.518822)	0.291352 / 0.579283 (-0.287931)	0.284579 / 0.434364 (-0.149785)	0.331214 / 0.540337 (-0.209124)	0.422902 / 1.386936 (-0.964034)