Automatically create cache_dir from cache_file_name

[ringohoffman]

You get a pretty unhelpful error message when specifying a cache_file_name in a directory that doesn't exist, e.g. cache_file_name="./cache/data.map"

import datasets

cache_file_name="./cache/train.map"
dataset = datasets.load_dataset("ylecun/mnist")
dataset["train"].map(lambda x: x, cache_file_name=cache_file_name)

FileNotFoundError: [Errno 2] No such file or directory: '/.../cache/tmp48r61siw'

It is simple enough to create and I was expecting that this would have been the case.

cc: @albertvillanova @lhoestq

[ringohoffman]

Hi @albertvillanova, is this PR looking okay to you? Anything else you'd like to see?

[lhoestq]

cool thanks !

[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)	0.005278 / 0.011353 (-0.006075)	0.003536 / 0.011008 (-0.007472)	0.062604 / 0.038508 (0.024096)	0.030704 / 0.023109 (0.007595)	0.242178 / 0.275898 (-0.033720)	0.264335 / 0.323480 (-0.059145)	0.004118 / 0.007986 (-0.003868)	0.002789 / 0.004328 (-0.001539)	0.048813 / 0.004250 (0.044563)	0.041787 / 0.037052 (0.004735)	0.252369 / 0.258489 (-0.006120)	0.280981 / 0.293841 (-0.012859)	0.029646 / 0.128546 (-0.098900)	0.012093 / 0.075646 (-0.063553)	0.203036 / 0.419271 (-0.216235)	0.035814 / 0.043533 (-0.007719)	0.248929 / 0.255139 (-0.006210)	0.266568 / 0.283200 (-0.016632)	0.018761 / 0.141683 (-0.122922)	1.188443 / 1.452155 (-0.263712)	1.219324 / 1.492716 (-0.273392)

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)	0.095256 / 0.018006 (0.077250)	0.301069 / 0.000490 (0.300579)	0.000219 / 0.000200 (0.000019)	0.000054 / 0.000054 (-0.000001)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.018541 / 0.037411 (-0.018870)	0.067333 / 0.014526 (0.052807)	0.075483 / 0.176557 (-0.101073)	0.121301 / 0.737135 (-0.615834)	0.076924 / 0.296338 (-0.219414)

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)	0.284722 / 0.215209 (0.069513)	2.817656 / 2.077655 (0.740001)	1.483827 / 1.504120 (-0.020293)	1.363072 / 1.541195 (-0.178123)	1.380472 / 1.468490 (-0.088018)	0.739543 / 4.584777 (-3.845234)	2.390699 / 3.745712 (-1.355013)	2.980347 / 5.269862 (-2.289515)	1.897881 / 4.565676 (-2.667795)	0.078827 / 0.424275 (-0.345448)	0.005193 / 0.007607 (-0.002414)	0.342739 / 0.226044 (0.116695)	3.370871 / 2.268929 (1.101942)	1.846475 / 55.444624 (-53.598150)	1.577860 / 6.876477 (-5.298617)	1.628606 / 2.142072 (-0.513466)	0.815686 / 4.805227 (-3.989541)	0.134985 / 6.500664 (-6.365679)	0.042330 / 0.075469 (-0.033139)

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)	0.962530 / 1.841788 (-0.879258)	11.271449 / 8.074308 (3.197141)	9.615452 / 10.191392 (-0.575940)	0.140322 / 0.680424 (-0.540101)	0.014057 / 0.534201 (-0.520144)	0.306212 / 0.579283 (-0.273071)	0.266758 / 0.434364 (-0.167606)	0.341229 / 0.540337 (-0.199108)	0.428974 / 1.386936 (-0.957962)

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)	0.005980 / 0.011353 (-0.005373)	0.003831 / 0.011008 (-0.007177)	0.049837 / 0.038508 (0.011329)	0.030602 / 0.023109 (0.007493)	0.274107 / 0.275898 (-0.001791)	0.298175 / 0.323480 (-0.025305)	0.004492 / 0.007986 (-0.003494)	0.002840 / 0.004328 (-0.001489)	0.048984 / 0.004250 (0.044733)	0.040001 / 0.037052 (0.002949)	0.286130 / 0.258489 (0.027641)	0.321546 / 0.293841 (0.027705)	0.032675 / 0.128546 (-0.095871)	0.012222 / 0.075646 (-0.063424)	0.060321 / 0.419271 (-0.358950)	0.034456 / 0.043533 (-0.009077)	0.272408 / 0.255139 (0.017269)	0.294714 / 0.283200 (0.011515)	0.018568 / 0.141683 (-0.123115)	1.169826 / 1.452155 (-0.282329)	1.223906 / 1.492716 (-0.268810)

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)	0.093734 / 0.018006 (0.075727)	0.305915 / 0.000490 (0.305425)	0.000210 / 0.000200 (0.000010)	0.000052 / 0.000054 (-0.000003)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.022389 / 0.037411 (-0.015022)	0.076640 / 0.014526 (0.062114)	0.088660 / 0.176557 (-0.087897)	0.128998 / 0.737135 (-0.608137)	0.090346 / 0.296338 (-0.205992)

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)	0.291642 / 0.215209 (0.076433)	2.897270 / 2.077655 (0.819615)	1.571564 / 1.504120 (0.067444)	1.449533 / 1.541195 (-0.091662)	1.458744 / 1.468490 (-0.009746)	0.725465 / 4.584777 (-3.859312)	0.962597 / 3.745712 (-2.783115)	3.035056 / 5.269862 (-2.234806)	1.902542 / 4.565676 (-2.663135)	0.079869 / 0.424275 (-0.344407)	0.005172 / 0.007607 (-0.002435)	0.352099 / 0.226044 (0.126055)	3.469058 / 2.268929 (1.200129)	1.953402 / 55.444624 (-53.491222)	1.647182 / 6.876477 (-5.229294)	1.686473 / 2.142072 (-0.455599)	0.797218 / 4.805227 (-4.008009)	0.134161 / 6.500664 (-6.366503)	0.041563 / 0.075469 (-0.033906)

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)	1.045855 / 1.841788 (-0.795933)	12.271390 / 8.074308 (4.197082)	10.186889 / 10.191392 (-0.004503)	0.141141 / 0.680424 (-0.539283)	0.015482 / 0.534201 (-0.518719)	0.305699 / 0.579283 (-0.273584)	0.128539 / 0.434364 (-0.305825)	0.348492 / 0.540337 (-0.191845)	0.444867 / 1.386936 (-0.942069)