Pickle support for torch.Generator objects

[mariosasko]

Fix for https://discuss.huggingface.co/t/caching-a-dataset-processed-with-randomness/65616

[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.

[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)	0.005472 / 0.011353 (-0.005881)	0.003715 / 0.011008 (-0.007293)	0.063257 / 0.038508 (0.024749)	0.060683 / 0.023109 (0.037574)	0.250885 / 0.275898 (-0.025013)	0.271685 / 0.323480 (-0.051795)	0.003051 / 0.007986 (-0.004934)	0.002799 / 0.004328 (-0.001530)	0.049113 / 0.004250 (0.044863)	0.038965 / 0.037052 (0.001912)	0.252688 / 0.258489 (-0.005801)	0.282536 / 0.293841 (-0.011305)	0.028722 / 0.128546 (-0.099824)	0.010586 / 0.075646 (-0.065060)	0.205145 / 0.419271 (-0.214127)	0.036996 / 0.043533 (-0.006537)	0.248874 / 0.255139 (-0.006265)	0.266148 / 0.283200 (-0.017051)	0.018540 / 0.141683 (-0.123143)	1.120216 / 1.452155 (-0.331938)	1.191072 / 1.492716 (-0.301644)

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)	0.095721 / 0.018006 (0.077714)	0.313401 / 0.000490 (0.312911)	0.000234 / 0.000200 (0.000034)	0.000053 / 0.000054 (-0.000002)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.018604 / 0.037411 (-0.018807)	0.061571 / 0.014526 (0.047045)	0.075343 / 0.176557 (-0.101213)	0.121272 / 0.737135 (-0.615864)	0.076448 / 0.296338 (-0.219890)

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)	0.286885 / 0.215209 (0.071676)	2.809100 / 2.077655 (0.731445)	1.485365 / 1.504120 (-0.018755)	1.367672 / 1.541195 (-0.173523)	1.423570 / 1.468490 (-0.044920)	0.571063 / 4.584777 (-4.013714)	2.385248 / 3.745712 (-1.360464)	2.855251 / 5.269862 (-2.414610)	1.799371 / 4.565676 (-2.766306)	0.063491 / 0.424275 (-0.360784)	0.004942 / 0.007607 (-0.002665)	0.346181 / 0.226044 (0.120137)	3.388123 / 2.268929 (1.119195)	1.819093 / 55.444624 (-53.625532)	1.552998 / 6.876477 (-5.323479)	1.627930 / 2.142072 (-0.514143)	0.653438 / 4.805227 (-4.151789)	0.123831 / 6.500664 (-6.376833)	0.043340 / 0.075469 (-0.032129)

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)	0.952167 / 1.841788 (-0.889621)	12.149515 / 8.074308 (4.075207)	10.665085 / 10.191392 (0.473693)	0.127768 / 0.680424 (-0.552656)	0.014022 / 0.534201 (-0.520179)	0.285959 / 0.579283 (-0.293324)	0.269727 / 0.434364 (-0.164637)	0.336646 / 0.540337 (-0.203692)	0.442932 / 1.386936 (-0.944005)

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)	0.005351 / 0.011353 (-0.006002)	0.003561 / 0.011008 (-0.007448)	0.048890 / 0.038508 (0.010382)	0.054093 / 0.023109 (0.030984)	0.274397 / 0.275898 (-0.001501)	0.296980 / 0.323480 (-0.026500)	0.004126 / 0.007986 (-0.003860)	0.002751 / 0.004328 (-0.001578)	0.049131 / 0.004250 (0.044880)	0.040769 / 0.037052 (0.003716)	0.279147 / 0.258489 (0.020658)	0.302014 / 0.293841 (0.008173)	0.029847 / 0.128546 (-0.098699)	0.010710 / 0.075646 (-0.064936)	0.057626 / 0.419271 (-0.361645)	0.032801 / 0.043533 (-0.010732)	0.272698 / 0.255139 (0.017559)	0.289238 / 0.283200 (0.006039)	0.017876 / 0.141683 (-0.123807)	1.152059 / 1.452155 (-0.300096)	1.212289 / 1.492716 (-0.280427)

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)	0.092914 / 0.018006 (0.074908)	0.303092 / 0.000490 (0.302603)	0.000214 / 0.000200 (0.000014)	0.000058 / 0.000054 (0.000004)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.022074 / 0.037411 (-0.015337)	0.070109 / 0.014526 (0.055583)	0.083360 / 0.176557 (-0.093196)	0.122445 / 0.737135 (-0.614690)	0.083625 / 0.296338 (-0.212714)

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)	0.282788 / 0.215209 (0.067579)	2.789229 / 2.077655 (0.711574)	1.571077 / 1.504120 (0.066957)	1.452627 / 1.541195 (-0.088567)	1.493176 / 1.468490 (0.024686)	0.556892 / 4.584777 (-4.027885)	2.442771 / 3.745712 (-1.302941)	2.826316 / 5.269862 (-2.443545)	1.758276 / 4.565676 (-2.807401)	0.063039 / 0.424275 (-0.361236)	0.004928 / 0.007607 (-0.002679)	0.338247 / 0.226044 (0.112202)	3.346344 / 2.268929 (1.077416)	1.952520 / 55.444624 (-53.492104)	1.664520 / 6.876477 (-5.211956)	1.701528 / 2.142072 (-0.440544)	0.634746 / 4.805227 (-4.170481)	0.116879 / 6.500664 (-6.383786)	0.040990 / 0.075469 (-0.034479)

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)	0.969521 / 1.841788 (-0.872267)	12.431395 / 8.074308 (4.357087)	10.907503 / 10.191392 (0.716111)	0.131028 / 0.680424 (-0.549396)	0.015239 / 0.534201 (-0.518962)	0.290793 / 0.579283 (-0.288490)	0.275072 / 0.434364 (-0.159292)	0.331036 / 0.540337 (-0.209301)	0.567858 / 1.386936 (-0.819078)