Reorder default data splits to have validation before test

[albertvillanova]

This PR reorders data splits, so that by default validation appears before test.

The default order becomes: [train, validation, test] instead of [train, test, validation].

[HuggingFaceDocBuilderDev]

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

[albertvillanova]

After this CI error: https://github.com/huggingface/datasets/actions/runs/4639528358/jobs/8210492953?pr=5718

FAILED tests/test_data_files.py::test_get_data_files_patterns[data_file_per_split4] - AssertionError: assert ['random', 'train'] == ['train', 'random']
  At index 0 diff: 'random' != 'train'
  Full diff:
  - ['train', 'random']
  + ['random', 'train']

I have checked locally and found out that the data split order is nondeterministic. I am addressing this in a separate issue.

We should first address:

• The order of data split names is nondeterministic #5728
• Fix nondeterministic sharded data split order #5729

[lhoestq]

it's better this way indeed !

[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)	0.007728 / 0.011353 (-0.003624)	0.005275 / 0.011008 (-0.005734)	0.097708 / 0.038508 (0.059199)	0.039851 / 0.023109 (0.016741)	0.333360 / 0.275898 (0.057462)	0.376135 / 0.323480 (0.052655)	0.006355 / 0.007986 (-0.001630)	0.004193 / 0.004328 (-0.000135)	0.072882 / 0.004250 (0.068631)	0.052668 / 0.037052 (0.015615)	0.347359 / 0.258489 (0.088870)	0.382280 / 0.293841 (0.088440)	0.035996 / 0.128546 (-0.092550)	0.012517 / 0.075646 (-0.063129)	0.334520 / 0.419271 (-0.084751)	0.051969 / 0.043533 (0.008436)	0.335735 / 0.255139 (0.080596)	0.359921 / 0.283200 (0.076722)	0.113971 / 0.141683 (-0.027712)	1.465636 / 1.452155 (0.013481)	1.559824 / 1.492716 (0.067108)

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)	0.223997 / 0.018006 (0.205991)	0.499041 / 0.000490 (0.498551)	0.009697 / 0.000200 (0.009497)	0.000245 / 0.000054 (0.000190)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.027031 / 0.037411 (-0.010381)	0.110271 / 0.014526 (0.095745)	0.115848 / 0.176557 (-0.060709)	0.174253 / 0.737135 (-0.562883)	0.122616 / 0.296338 (-0.173723)

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)	0.417275 / 0.215209 (0.202066)	4.158678 / 2.077655 (2.081023)	1.917585 / 1.504120 (0.413465)	1.722219 / 1.541195 (0.181025)	1.813284 / 1.468490 (0.344793)	0.707193 / 4.584777 (-3.877584)	3.853545 / 3.745712 (0.107833)	3.369240 / 5.269862 (-1.900621)	1.820264 / 4.565676 (-2.745412)	0.087340 / 0.424275 (-0.336936)	0.012305 / 0.007607 (0.004698)	0.520326 / 0.226044 (0.294281)	5.107383 / 2.268929 (2.838455)	2.413977 / 55.444624 (-53.030647)	2.074356 / 6.876477 (-4.802121)	2.255959 / 2.142072 (0.113887)	0.849850 / 4.805227 (-3.955377)	0.170116 / 6.500664 (-6.330548)	0.067203 / 0.075469 (-0.008267)

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)	1.168158 / 1.841788 (-0.673629)	15.046312 / 8.074308 (6.972004)	15.113924 / 10.191392 (4.922532)	0.145288 / 0.680424 (-0.535136)	0.017959 / 0.534201 (-0.516242)	0.424666 / 0.579283 (-0.154617)	0.422560 / 0.434364 (-0.011804)	0.526386 / 0.540337 (-0.013952)	0.623755 / 1.386936 (-0.763181)

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)	0.007676 / 0.011353 (-0.003677)	0.005240 / 0.011008 (-0.005769)	0.074668 / 0.038508 (0.036160)	0.035570 / 0.023109 (0.012461)	0.348524 / 0.275898 (0.072626)	0.378157 / 0.323480 (0.054677)	0.006112 / 0.007986 (-0.001873)	0.005641 / 0.004328 (0.001312)	0.073536 / 0.004250 (0.069286)	0.048651 / 0.037052 (0.011599)	0.359282 / 0.258489 (0.100793)	0.385961 / 0.293841 (0.092120)	0.035417 / 0.128546 (-0.093129)	0.012227 / 0.075646 (-0.063419)	0.085725 / 0.419271 (-0.333546)	0.049651 / 0.043533 (0.006118)	0.344122 / 0.255139 (0.088983)	0.364795 / 0.283200 (0.081595)	0.112711 / 0.141683 (-0.028972)	1.426823 / 1.452155 (-0.025332)	1.534745 / 1.492716 (0.042029)

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)	0.201728 / 0.018006 (0.183721)	0.448533 / 0.000490 (0.448043)	0.003554 / 0.000200 (0.003354)	0.000092 / 0.000054 (0.000038)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.030917 / 0.037411 (-0.006494)	0.117966 / 0.014526 (0.103440)	0.125954 / 0.176557 (-0.050602)	0.176382 / 0.737135 (-0.560753)	0.130757 / 0.296338 (-0.165582)

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)	0.422167 / 0.215209 (0.206958)	4.213948 / 2.077655 (2.136294)	2.040049 / 1.504120 (0.535929)	1.858317 / 1.541195 (0.317122)	1.937108 / 1.468490 (0.468618)	0.707797 / 4.584777 (-3.876979)	3.831061 / 3.745712 (0.085349)	3.373711 / 5.269862 (-1.896151)	1.590343 / 4.565676 (-2.975333)	0.086672 / 0.424275 (-0.337603)	0.012429 / 0.007607 (0.004821)	0.520269 / 0.226044 (0.294225)	5.207285 / 2.268929 (2.938357)	2.518107 / 55.444624 (-52.926517)	2.230696 / 6.876477 (-4.645781)	2.363164 / 2.142072 (0.221091)	0.836749 / 4.805227 (-3.968479)	0.169676 / 6.500664 (-6.330988)	0.065766 / 0.075469 (-0.009703)

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)	1.251195 / 1.841788 (-0.590592)	15.196091 / 8.074308 (7.121782)	14.991600 / 10.191392 (4.800208)	0.165335 / 0.680424 (-0.515089)	0.017789 / 0.534201 (-0.516412)	0.433863 / 0.579283 (-0.145420)	0.428660 / 0.434364 (-0.005704)	0.527385 / 0.540337 (-0.012952)	0.628067 / 1.386936 (-0.758869)