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

canonicalize data dir in config ID hash #5899

Merged
merged 1 commit into from
Jun 2, 2023
Merged

canonicalize data dir in config ID hash #5899

merged 1 commit into from
Jun 2, 2023

Conversation

kylrth
Copy link
Contributor

@kylrth kylrth commented May 25, 2023

fixes #5871

The second commit is optional but improves readability.

@HuggingFaceDocBuilderDev
Copy link

HuggingFaceDocBuilderDev commented Jun 1, 2023 •
edited
Loading

The documentation is not available anymore as the PR was closed or merged.

mariosasko
mariosasko reviewed Jun 1, 2023
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.

Thanks!

src/datasets/builder.py Outdated Show resolved Hide resolved
src/datasets/builder.py Outdated Show resolved Hide resolved
@kylrth
canonicalize data dir in config ID hash
7789d90 
This leaves the hash unchanged when the data dir changes in
insubstantial ways, like adding a trailing slash or using a symlink.

fixes huggingface#5871
@mariosasko mariosasko merged commit 02ee418 into huggingface:main Jun 2, 2023
@github-actions
Copy link

github-actions bot commented Jun 2, 2023

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) 0.009137 / 0.011353 (-0.002216) 0.006119 / 0.011008 (-0.004889) 0.136530 / 0.038508 (0.098022) 0.038434 / 0.023109 (0.015325) 0.427900 / 0.275898 (0.152002) 0.449757 / 0.323480 (0.126277) 0.007673 / 0.007986 (-0.000313) 0.007147 / 0.004328 (0.002818) 0.108029 / 0.004250 (0.103778) 0.055072 / 0.037052 (0.018020) 0.439245 / 0.258489 (0.180756) 0.477285 / 0.293841 (0.183444) 0.044838 / 0.128546 (-0.083708) 0.020814 / 0.075646 (-0.054832) 0.436098 / 0.419271 (0.016826) 0.067459 / 0.043533 (0.023926) 0.427470 / 0.255139 (0.172331) 0.443260 / 0.283200 (0.160060) 0.125466 / 0.141683 (-0.016216) 1.996756 / 1.452155 (0.544601) 2.100679 / 1.492716 (0.607962)

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) 0.278407 / 0.018006 (0.260401) 0.625855 / 0.000490 (0.625365) 0.005544 / 0.000200 (0.005344) 0.000107 / 0.000054 (0.000053)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.033495 / 0.037411 (-0.003916) 0.134718 / 0.014526 (0.120192) 0.150151 / 0.176557 (-0.026406) 0.221385 / 0.737135 (-0.515751) 0.150932 / 0.296338 (-0.145406)

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) 0.668845 / 0.215209 (0.453636) 6.678436 / 2.077655 (4.600781) 2.714074 / 1.504120 (1.209954) 2.275784 / 1.541195 (0.734589) 2.332852 / 1.468490 (0.864361) 1.014877 / 4.584777 (-3.569900) 6.086455 / 3.745712 (2.340743) 2.990029 / 5.269862 (-2.279832) 1.862236 / 4.565676 (-2.703441) 0.122179 / 0.424275 (-0.302096) 0.015706 / 0.007607 (0.008099) 0.873473 / 0.226044 (0.647429) 8.580109 / 2.268929 (6.311180) 3.458360 / 55.444624 (-51.986264) 2.738801 / 6.876477 (-4.137676) 2.918428 / 2.142072 (0.776356) 1.224910 / 4.805227 (-3.580317) 0.243006 / 6.500664 (-6.257658) 0.087121 / 0.075469 (0.011652)

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) 1.757802 / 1.841788 (-0.083986) 19.447999 / 8.074308 (11.373691) 24.518157 / 10.191392 (14.326765) 0.245013 / 0.680424 (-0.435411) 0.032290 / 0.534201 (-0.501911) 0.542043 / 0.579283 (-0.037240) 0.708154 / 0.434364 (0.273790) 0.660584 / 0.540337 (0.120247) 0.794868 / 1.386936 (-0.592068)
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) 0.009496 / 0.011353 (-0.001857) 0.005842 / 0.011008 (-0.005166) 0.112813 / 0.038508 (0.074305) 0.039120 / 0.023109 (0.016011) 0.489717 / 0.275898 (0.213819) 0.532586 / 0.323480 (0.209107) 0.007681 / 0.007986 (-0.000304) 0.005337 / 0.004328 (0.001009) 0.107244 / 0.004250 (0.102994) 0.056847 / 0.037052 (0.019794) 0.499447 / 0.258489 (0.240958) 0.548995 / 0.293841 (0.255154) 0.058047 / 0.128546 (-0.070499) 0.015468 / 0.075646 (-0.060179) 0.124600 / 0.419271 (-0.294671) 0.060940 / 0.043533 (0.017407) 0.488370 / 0.255139 (0.233231) 0.518540 / 0.283200 (0.235341) 0.124147 / 0.141683 (-0.017536) 1.902922 / 1.452155 (0.450767) 2.033519 / 1.492716 (0.540803)

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) 0.319527 / 0.018006 (0.301521) 0.629641 / 0.000490 (0.629152) 0.000721 / 0.000200 (0.000521) 0.000101 / 0.000054 (0.000046)

Benchmark: benchmark_indices_mapping.json

metric select shard shuffle sort train_test_split
new / old (diff) 0.033150 / 0.037411 (-0.004262) 0.134250 / 0.014526 (0.119724) 0.161273 / 0.176557 (-0.015283) 0.211471 / 0.737135 (-0.525664) 0.155326 / 0.296338 (-0.141012)

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) 0.705244 / 0.215209 (0.490035) 7.043040 / 2.077655 (4.965386) 3.308948 / 1.504120 (1.804828) 2.885050 / 1.541195 (1.343855) 2.810260 / 1.468490 (1.341770) 1.027095 / 4.584777 (-3.557682) 6.111398 / 3.745712 (2.365686) 5.385545 / 5.269862 (0.115684) 2.521668 / 4.565676 (-2.044009) 0.122419 / 0.424275 (-0.301856) 0.016376 / 0.007607 (0.008768) 0.830856 / 0.226044 (0.604811) 8.952199 / 2.268929 (6.683271) 4.207875 / 55.444624 (-51.236749) 3.346624 / 6.876477 (-3.529853) 3.395316 / 2.142072 (1.253244) 1.351816 / 4.805227 (-3.453411) 0.303056 / 6.500664 (-6.197608) 0.098713 / 0.075469 (0.023244)

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) 1.841903 / 1.841788 (0.000116) 20.472125 / 8.074308 (12.397817) 23.433200 / 10.191392 (13.241808) 0.242599 / 0.680424 (-0.437825) 0.030701 / 0.534201 (-0.503500) 0.541614 / 0.579283 (-0.037669) 0.657827 / 0.434364 (0.223463) 0.652448 / 0.540337 (0.112111) 0.773743 / 1.386936 (-0.613193)

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.

data configuration hash suffix depends on uncanonicalized data_dir
3 participants
@kylrth @HuggingFaceDocBuilderDev @mariosasko