Fix download for dict of dicts of URLs

[albertvillanova]

Fix download for a dict of dicts of URLs when batched (default), introduced by:

• Multithreaded downloads #6794

This PR also implements regression tests.

Fix #6869, fix #6850.

[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]

Once merged, I think a patch release is needed.

[albertvillanova]

Once the CI is green, I am merging this PR and making a patch release, @huggingface/datasets.

[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)	0.005352 / 0.011353 (-0.006001)	0.004140 / 0.011008 (-0.006868)	0.063844 / 0.038508 (0.025336)	0.030712 / 0.023109 (0.007603)	0.232790 / 0.275898 (-0.043108)	0.262334 / 0.323480 (-0.061145)	0.003264 / 0.007986 (-0.004721)	0.002654 / 0.004328 (-0.001674)	0.049775 / 0.004250 (0.045524)	0.046803 / 0.037052 (0.009751)	0.250667 / 0.258489 (-0.007822)	0.283581 / 0.293841 (-0.010260)	0.027660 / 0.128546 (-0.100886)	0.010560 / 0.075646 (-0.065087)	0.208676 / 0.419271 (-0.210596)	0.035415 / 0.043533 (-0.008118)	0.235380 / 0.255139 (-0.019759)	0.261220 / 0.283200 (-0.021980)	0.019551 / 0.141683 (-0.122132)	1.140196 / 1.452155 (-0.311959)	1.173021 / 1.492716 (-0.319696)

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)	0.092665 / 0.018006 (0.074659)	0.301524 / 0.000490 (0.301034)	0.000216 / 0.000200 (0.000016)	0.000043 / 0.000054 (-0.000011)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.018485 / 0.037411 (-0.018927)	0.061722 / 0.014526 (0.047196)	0.074701 / 0.176557 (-0.101855)	0.121443 / 0.737135 (-0.615692)	0.076268 / 0.296338 (-0.220070)

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)	0.284143 / 0.215209 (0.068934)	2.789979 / 2.077655 (0.712324)	1.501156 / 1.504120 (-0.002964)	1.379414 / 1.541195 (-0.161781)	1.419092 / 1.468490 (-0.049398)	0.554107 / 4.584777 (-4.030670)	2.365659 / 3.745712 (-1.380053)	2.763963 / 5.269862 (-2.505898)	1.712587 / 4.565676 (-2.853090)	0.060961 / 0.424275 (-0.363314)	0.005301 / 0.007607 (-0.002306)	0.346253 / 0.226044 (0.120209)	3.351833 / 2.268929 (1.082905)	1.831946 / 55.444624 (-53.612679)	1.556530 / 6.876477 (-5.319947)	1.574185 / 2.142072 (-0.567887)	0.630396 / 4.805227 (-4.174831)	0.116126 / 6.500664 (-6.384538)	0.042391 / 0.075469 (-0.033078)

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)	0.981430 / 1.841788 (-0.860358)	11.619671 / 8.074308 (3.545363)	9.718227 / 10.191392 (-0.473165)	0.130918 / 0.680424 (-0.549506)	0.014116 / 0.534201 (-0.520085)	0.288729 / 0.579283 (-0.290554)	0.259183 / 0.434364 (-0.175181)	0.323764 / 0.540337 (-0.216574)	0.420336 / 1.386936 (-0.966600)

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)	0.005255 / 0.011353 (-0.006098)	0.003664 / 0.011008 (-0.007344)	0.051376 / 0.038508 (0.012868)	0.030429 / 0.023109 (0.007320)	0.263090 / 0.275898 (-0.012808)	0.289959 / 0.323480 (-0.033521)	0.004214 / 0.007986 (-0.003772)	0.002782 / 0.004328 (-0.001546)	0.049043 / 0.004250 (0.044793)	0.041016 / 0.037052 (0.003964)	0.275616 / 0.258489 (0.017127)	0.303350 / 0.293841 (0.009509)	0.029484 / 0.128546 (-0.099062)	0.010329 / 0.075646 (-0.065317)	0.058680 / 0.419271 (-0.360591)	0.032818 / 0.043533 (-0.010715)	0.263368 / 0.255139 (0.008229)	0.286839 / 0.283200 (0.003640)	0.018029 / 0.141683 (-0.123654)	1.169207 / 1.452155 (-0.282948)	1.206568 / 1.492716 (-0.286148)

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)	0.101394 / 0.018006 (0.083387)	0.310414 / 0.000490 (0.309924)	0.000213 / 0.000200 (0.000013)	0.000053 / 0.000054 (-0.000002)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.021662 / 0.037411 (-0.015749)	0.075320 / 0.014526 (0.060794)	0.086607 / 0.176557 (-0.089949)	0.127268 / 0.737135 (-0.609867)	0.088244 / 0.296338 (-0.208095)

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)	0.293591 / 0.215209 (0.078382)	2.871845 / 2.077655 (0.794190)	1.543624 / 1.504120 (0.039504)	1.426698 / 1.541195 (-0.114497)	1.445348 / 1.468490 (-0.023142)	0.565156 / 4.584777 (-4.019621)	0.961782 / 3.745712 (-2.783930)	2.827904 / 5.269862 (-2.441958)	1.747728 / 4.565676 (-2.817949)	0.063275 / 0.424275 (-0.361000)	0.004987 / 0.007607 (-0.002620)	0.349652 / 0.226044 (0.123607)	3.448635 / 2.268929 (1.179707)	1.891734 / 55.444624 (-53.552890)	1.624274 / 6.876477 (-5.252202)	1.641531 / 2.142072 (-0.500541)	0.642081 / 4.805227 (-4.163146)	0.116136 / 6.500664 (-6.384528)	0.040807 / 0.075469 (-0.034662)

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)	1.002090 / 1.841788 (-0.839697)	12.401097 / 8.074308 (4.326788)	9.799316 / 10.191392 (-0.392076)	0.131770 / 0.680424 (-0.548654)	0.016817 / 0.534201 (-0.517384)	0.301136 / 0.579283 (-0.278147)	0.136810 / 0.434364 (-0.297554)	0.384740 / 0.540337 (-0.155598)	0.423779 / 1.386936 (-0.963157)