Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Fix concurrent script loading with force_redownload #6718

Merged
merged 6 commits into from
Mar 7, 2024
Merged

Fix concurrent script loading with force_redownload #6718

merged 6 commits into from
Mar 7, 2024

Conversation

lhoestq
Copy link
Member

@lhoestq lhoestq commented Mar 5, 2024 •
edited
Loading

I added lock_importable_file in get_dataset_builder_class and extend_dataset_builder_for_streaming to fix the issue, and I also added a test

cc @clefourrier

@HuggingFaceDocBuilderDev
Copy link

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.

lhoestq
lhoestq commented Mar 6, 2024
View reviewed changes
src/datasets/load.py
# Define a directory with a unique name in our dataset or metric folder
# path is: ./datasets|metrics/dataset|metric_name/hash_from_code/script.py
# we use a hash as subdirectory_name to be able to have multiple versions of a dataset/metric processing file together
importable_subdirectory = os.path.join(importable_directory_path, subdirectory_name)
importable_local_file = os.path.join(importable_subdirectory, name + ".py")
importable_file = os.path.join(importable_subdirectory, name + ".py")
Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

just a variable name change for consistency

@lhoestq lhoestq requested a review from mariosasko March 6, 2024 15:21
mariosasko
mariosasko approved these changes Mar 6, 2024
View reviewed changes
Copy link
Collaborator

@mariosasko mariosasko left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM!

tests/test_load.py Outdated Show resolved Hide resolved
@lhoestq @mariosasko
Update tests/test_load.py
21f2d8c 
Co-authored-by: Mario Šaško <mariosasko777@gmail.com>
@lhoestq lhoestq merged commit f45bc6c into main Mar 7, 2024
12 checks passed
@lhoestq lhoestq deleted the fix-concurrent-script-loading-with-force_redownload branch March 7, 2024 13:58
@github-actions GitHub Actions
Copy link

github-actions bot commented Mar 7, 2024

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) 0.005074 / 0.011353 (-0.006279) 0.003505 / 0.011008 (-0.007503) 0.063683 / 0.038508 (0.025175) 0.029308 / 0.023109 (0.006199) 0.246648 / 0.275898 (-0.029250) 0.265546 / 0.323480 (-0.057933) 0.004108 / 0.007986 (-0.003878) 0.002683 / 0.004328 (-0.001646) 0.048634 / 0.004250 (0.044383) 0.043786 / 0.037052 (0.006733) 0.262197 / 0.258489 (0.003708) 0.291582 / 0.293841 (-0.002259) 0.027472 / 0.128546 (-0.101074) 0.010213 / 0.075646 (-0.065434) 0.206744 / 0.419271 (-0.212527) 0.036195 / 0.043533 (-0.007337) 0.249090 / 0.255139 (-0.006049) 0.280002 / 0.283200 (-0.003198) 0.018568 / 0.141683 (-0.123115) 1.124844 / 1.452155 (-0.327311) 1.159358 / 1.492716 (-0.333359)

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) 0.093186 / 0.018006 (0.075180) 0.302331 / 0.000490 (0.301842) 0.000217 / 0.000200 (0.000017) 0.000046 / 0.000054 (-0.000008)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.018727 / 0.037411 (-0.018684) 0.061730 / 0.014526 (0.047204) 0.074330 / 0.176557 (-0.102226) 0.119769 / 0.737135 (-0.617366) 0.075611 / 0.296338 (-0.220727)

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) 0.285063 / 0.215209 (0.069854) 2.824809 / 2.077655 (0.747155) 1.481858 / 1.504120 (-0.022262) 1.350193 / 1.541195 (-0.191002) 1.358012 / 1.468490 (-0.110478) 0.557842 / 4.584777 (-4.026935) 2.380729 / 3.745712 (-1.364983) 2.798891 / 5.269862 (-2.470970) 1.719288 / 4.565676 (-2.846388) 0.061705 / 0.424275 (-0.362570) 0.005431 / 0.007607 (-0.002176) 0.343233 / 0.226044 (0.117189) 3.375223 / 2.268929 (1.106295) 1.838188 / 55.444624 (-53.606436) 1.570015 / 6.876477 (-5.306461) 1.573157 / 2.142072 (-0.568915) 0.650678 / 4.805227 (-4.154549) 0.116412 / 6.500664 (-6.384252) 0.041754 / 0.075469 (-0.033715)

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) 0.970431 / 1.841788 (-0.871357) 11.317128 / 8.074308 (3.242819) 9.691240 / 10.191392 (-0.500152) 0.142260 / 0.680424 (-0.538164) 0.014131 / 0.534201 (-0.520070) 0.289910 / 0.579283 (-0.289373) 0.265648 / 0.434364 (-0.168715) 0.323130 / 0.540337 (-0.217208) 0.447005 / 1.386936 (-0.939931)
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) 0.005322 / 0.011353 (-0.006031) 0.003755 / 0.011008 (-0.007253) 0.049646 / 0.038508 (0.011138) 0.029669 / 0.023109 (0.006560) 0.284151 / 0.275898 (0.008253) 0.298351 / 0.323480 (-0.025128) 0.004183 / 0.007986 (-0.003803) 0.002683 / 0.004328 (-0.001645) 0.048814 / 0.004250 (0.044563) 0.045017 / 0.037052 (0.007965) 0.287358 / 0.258489 (0.028869) 0.317394 / 0.293841 (0.023553) 0.030025 / 0.128546 (-0.098521) 0.010854 / 0.075646 (-0.064793) 0.058694 / 0.419271 (-0.360578) 0.052287 / 0.043533 (0.008754) 0.279038 / 0.255139 (0.023899) 0.295442 / 0.283200 (0.012242) 0.019413 / 0.141683 (-0.122270) 1.146106 / 1.452155 (-0.306048) 1.197777 / 1.492716 (-0.294939)

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) 0.092191 / 0.018006 (0.074184) 0.302672 / 0.000490 (0.302182) 0.000623 / 0.000200 (0.000423) 0.000048 / 0.000054 (-0.000006)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.022067 / 0.037411 (-0.015345) 0.081760 / 0.014526 (0.067235) 0.087548 / 0.176557 (-0.089009) 0.126405 / 0.737135 (-0.610730) 0.089331 / 0.296338 (-0.207008)

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) 0.295821 / 0.215209 (0.080612) 2.897930 / 2.077655 (0.820276) 1.604500 / 1.504120 (0.100380) 1.471502 / 1.541195 (-0.069692) 1.497918 / 1.468490 (0.029428) 0.576179 / 4.584777 (-4.008598) 2.452103 / 3.745712 (-1.293609) 2.668043 / 5.269862 (-2.601818) 1.753544 / 4.565676 (-2.812133) 0.064410 / 0.424275 (-0.359865) 0.005027 / 0.007607 (-0.002580) 0.351509 / 0.226044 (0.125465) 3.479208 / 2.268929 (1.210280) 1.990356 / 55.444624 (-53.454269) 1.684920 / 6.876477 (-5.191556) 1.794251 / 2.142072 (-0.347821) 0.662692 / 4.805227 (-4.142535) 0.118589 / 6.500664 (-6.382076) 0.040813 / 0.075469 (-0.034656)

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) 1.002390 / 1.841788 (-0.839398) 12.004617 / 8.074308 (3.930309) 10.216005 / 10.191392 (0.024613) 0.154354 / 0.680424 (-0.526070) 0.015554 / 0.534201 (-0.518647) 0.288741 / 0.579283 (-0.290542) 0.276774 / 0.434364 (-0.157590) 0.327055 / 0.540337 (-0.213282) 0.435121 / 1.386936 (-0.951815)

@clefourrier clefourrier mentioned this pull request Mar 27, 2024
Closed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@mariosasko mariosasko mariosasko approved these changes

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
3 participants
@lhoestq @HuggingFaceDocBuilderDev @mariosasko