Fix cache path to snakecase for CachedDatasetModuleFactory and Cache

[izhx]

Fix #6750 (comment)

I didn't find a guideline on how to run the tests, so i just run the following steps to make sure that this bug is fixed.

1. python test.py,
2. then HF_DATASETS_OFFLINE=1 python test.py

The test.py is

  import datasets

  datasets.utils.logging.set_verbosity_info()

  ds = datasets.load_dataset('izhx/STS17-debug')
  print(ds)

  ds = datasets.load_dataset('C-MTEB/AFQMC', revision='b44c3b011063adb25877c13823db83bb193913c4')
  print(ds)

[izhx]

@lhoestq hi 😃, is there something else I need to do to check this change?

[lhoestq]

Hi ! Thanks for the fix :) I added a comment

I also created https://huggingface.co/datasets/hf-internal-testing/DatasetWithCapitalLetters if you want to use it to add a test in test_cache.py (similar to test_cache_single_config for example)

[izhx]

I added two tests and passed them on my server.

pytest tests/packaged_modules/test_cache.py 
========================================================================== test session starts ==========================================================================
platform linux -- Python 3.11.5, pytest-8.1.1, pluggy-1.4.0
rootdir: /mnt/nas/datasets
configfile: pyproject.toml
plugins: xdist-3.5.0, datadir-1.5.0
collected 8 items                                                                                                                                                       

tests/packaged_modules/test_cache.py ........                                                                                                                     [100%]

========================================================================== 8 passed in 50.71s ===========================================================================

pytest tests/test_load.py
========================================================================== test session starts ==========================================================================
platform linux -- Python 3.11.5, pytest-8.1.1, pluggy-1.4.0
rootdir: /mnt/nas/datasets
configfile: pyproject.toml
plugins: xdist-3.5.0, datadir-1.5.0
collected 151 items                                                                                                                                                     

tests/test_load.py .............................................................................................................................................. [ 94%]
.........                                                                                                                                                         [100%]

...

============================================================= 151 passed, 29 warnings in 578.36s (0:09:38) ==============================================================

[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.

[minotru]

Hi @izhx! I have also faced this issue, happy to see it already addressed, looking forward for PR merge :)

[izhx]

@lhoestq What do you think of these tests? 😀

[lhoestq]

LGTM ! thanks :)

[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)	0.005060 / 0.011353 (-0.006293)	0.003251 / 0.011008 (-0.007757)	0.063538 / 0.038508 (0.025030)	0.031178 / 0.023109 (0.008069)	0.249971 / 0.275898 (-0.025927)	0.284828 / 0.323480 (-0.038652)	0.004183 / 0.007986 (-0.003802)	0.002656 / 0.004328 (-0.001673)	0.049585 / 0.004250 (0.045335)	0.042656 / 0.037052 (0.005604)	0.270962 / 0.258489 (0.012473)	0.296091 / 0.293841 (0.002250)	0.028065 / 0.128546 (-0.100482)	0.010545 / 0.075646 (-0.065102)	0.207323 / 0.419271 (-0.211948)	0.035977 / 0.043533 (-0.007556)	0.257315 / 0.255139 (0.002176)	0.272238 / 0.283200 (-0.010962)	0.017984 / 0.141683 (-0.123699)	1.131314 / 1.452155 (-0.320840)	1.180259 / 1.492716 (-0.312457)

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)	0.090977 / 0.018006 (0.072971)	0.284021 / 0.000490 (0.283531)	0.000264 / 0.000200 (0.000065)	0.000044 / 0.000054 (-0.000011)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.017852 / 0.037411 (-0.019559)	0.061288 / 0.014526 (0.046762)	0.073844 / 0.176557 (-0.102713)	0.121371 / 0.737135 (-0.615764)	0.075036 / 0.296338 (-0.221303)

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)	0.287599 / 0.215209 (0.072390)	2.821172 / 2.077655 (0.743517)	1.459904 / 1.504120 (-0.044216)	1.340224 / 1.541195 (-0.200970)	1.357350 / 1.468490 (-0.111140)	0.557344 / 4.584777 (-4.027433)	2.412177 / 3.745712 (-1.333535)	2.745126 / 5.269862 (-2.524735)	1.754600 / 4.565676 (-2.811077)	0.062487 / 0.424275 (-0.361788)	0.005306 / 0.007607 (-0.002301)	0.338856 / 0.226044 (0.112811)	3.354953 / 2.268929 (1.086024)	1.803208 / 55.444624 (-53.641417)	1.553051 / 6.876477 (-5.323426)	1.554790 / 2.142072 (-0.587282)	0.651380 / 4.805227 (-4.153847)	0.117777 / 6.500664 (-6.382887)	0.041992 / 0.075469 (-0.033477)

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)	0.977588 / 1.841788 (-0.864200)	11.363058 / 8.074308 (3.288750)	9.791770 / 10.191392 (-0.399622)	0.130708 / 0.680424 (-0.549716)	0.013798 / 0.534201 (-0.520403)	0.288313 / 0.579283 (-0.290970)	0.268170 / 0.434364 (-0.166194)	0.324815 / 0.540337 (-0.215522)	0.419260 / 1.386936 (-0.967676)

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)	0.005187 / 0.011353 (-0.006166)	0.003348 / 0.011008 (-0.007660)	0.050309 / 0.038508 (0.011801)	0.031334 / 0.023109 (0.008225)	0.279542 / 0.275898 (0.003644)	0.299608 / 0.323480 (-0.023872)	0.004202 / 0.007986 (-0.003784)	0.002735 / 0.004328 (-0.001593)	0.050321 / 0.004250 (0.046070)	0.039793 / 0.037052 (0.002740)	0.289972 / 0.258489 (0.031483)	0.313887 / 0.293841 (0.020046)	0.028797 / 0.128546 (-0.099750)	0.010166 / 0.075646 (-0.065480)	0.059228 / 0.419271 (-0.360044)	0.032667 / 0.043533 (-0.010866)	0.278409 / 0.255139 (0.023270)	0.292208 / 0.283200 (0.009008)	0.017577 / 0.141683 (-0.124106)	1.175046 / 1.452155 (-0.277109)	1.200766 / 1.492716 (-0.291950)

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)	0.092236 / 0.018006 (0.074230)	0.298860 / 0.000490 (0.298370)	0.000211 / 0.000200 (0.000011)	0.000043 / 0.000054 (-0.000011)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.021475 / 0.037411 (-0.015936)	0.074414 / 0.014526 (0.059888)	0.087746 / 0.176557 (-0.088811)	0.124757 / 0.737135 (-0.612378)	0.088513 / 0.296338 (-0.207826)

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)	0.296583 / 0.215209 (0.081374)	2.894978 / 2.077655 (0.817323)	1.590806 / 1.504120 (0.086686)	1.463251 / 1.541195 (-0.077944)	1.478751 / 1.468490 (0.010261)	0.571724 / 4.584777 (-4.013053)	2.454356 / 3.745712 (-1.291356)	2.789275 / 5.269862 (-2.480586)	1.753866 / 4.565676 (-2.811811)	0.064787 / 0.424275 (-0.359488)	0.005321 / 0.007607 (-0.002287)	0.348454 / 0.226044 (0.122410)	3.453052 / 2.268929 (1.184124)	1.972237 / 55.444624 (-53.472388)	1.677822 / 6.876477 (-5.198655)	1.674750 / 2.142072 (-0.467322)	0.649353 / 4.805227 (-4.155874)	0.117135 / 6.500664 (-6.383529)	0.040018 / 0.075469 (-0.035451)

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)	1.029812 / 1.841788 (-0.811976)	11.945063 / 8.074308 (3.870755)	10.238380 / 10.191392 (0.046988)	0.146225 / 0.680424 (-0.534199)	0.015262 / 0.534201 (-0.518939)	0.286632 / 0.579283 (-0.292651)	0.272952 / 0.434364 (-0.161412)	0.323098 / 0.540337 (-0.217239)	0.423549 / 1.386936 (-0.963387)