add GPU sync tests #2194
Closed
add GPU sync tests #2194
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
facebook-github-bot
added
the
CLA Signed
|
This pull request was exported from Phabricator. Differential Revision: D59173140 |
|
This pull request was exported from Phabricator. Differential Revision: D59173140 |
iamzainhuda
force-pushed
the
export-D59173140
branch
from
ef38e07
to
7ecf930
Compare
iamzainhuda
added a commit
to iamzainhuda/torchrec
that referenced
this pull request
Sep 25, 2024
Summary: Pull Request resolved: pytorch#2194 Added GPU sync tests to simulate gathering metric states on to rank 0 and computing. Tests don't cover this case before, which has resulted in SEVs in the past as users aren't aware of how RecMetrics collects and computes metrics. Added numerical value tests, most metrics are do not have this which can result in issues down the line if metrics need to be changed/accomodate future changes. Also we've found inconsistencies sometimes from other methods, so always good to check here. We compare each metric to a reference implementation from literature to ensure the values are as expected. Reviewed By: henrylhtsang Differential Revision: D59173140
iamzainhuda
added a commit
to iamzainhuda/torchrec
that referenced
this pull request
Oct 15, 2024
Summary: Added GPU sync tests to simulate gathering metric states on to rank 0 and computing. Tests don't cover this case before, which has resulted in SEVs in the past as users aren't aware of how RecMetrics collects and computes metrics. Added numerical value tests, most metrics are do not have this which can result in issues down the line if metrics need to be changed/accomodate future changes. Also we've found inconsistencies sometimes from other methods, so always good to check here. We compare each metric to a reference implementation from literature to ensure the values are as expected. Reviewed By: henrylhtsang Differential Revision: D59173140
iamzainhuda
force-pushed
the
export-D59173140
branch
from
7ecf930
to
b15a18a
Compare
|
This pull request was exported from Phabricator. Differential Revision: D59173140 |
iamzainhuda
added a commit
to iamzainhuda/torchrec
that referenced
this pull request
Oct 15, 2024
Summary: Added GPU sync tests to simulate gathering metric states on to rank 0 and computing. Tests don't cover this case before, which has resulted in SEVs in the past as users aren't aware of how RecMetrics collects and computes metrics. Added numerical value tests, most metrics are do not have this which can result in issues down the line if metrics need to be changed/accommodate future changes. Also we've found inconsistencies sometimes from other methods, so always good to check here. We compare each metric to a reference implementation from literature to ensure the values are as expected. Fixed and cleaned up some tests too. Reviewed By: henrylhtsang Differential Revision: D59173140
iamzainhuda
force-pushed
the
export-D59173140
branch
from
b15a18a
to
b482bb5
Compare
|
This pull request was exported from Phabricator. Differential Revision: D59173140 |
iamzainhuda
force-pushed
the
export-D59173140
branch
from
b482bb5
to
b1707f5
Compare
iamzainhuda
added a commit
to iamzainhuda/torchrec
that referenced
this pull request
Oct 15, 2024
Summary: Added GPU sync tests to simulate gathering metric states on to rank 0 and computing. Tests don't cover this case before, which has resulted in SEVs in the past as users aren't aware of how RecMetrics collects and computes metrics. Added numerical value tests, most metrics are do not have this which can result in issues down the line if metrics need to be changed/accommodate future changes. Also we've found inconsistencies sometimes from other methods, so always good to check here. We compare each metric to a reference implementation from literature to ensure the values are as expected. Fixed and cleaned up some tests too. Reviewed By: henrylhtsang Differential Revision: D59173140
|
This pull request was exported from Phabricator. Differential Revision: D59173140 |
iamzainhuda
force-pushed
the
export-D59173140
branch
from
b1707f5
to
1356d1d
Compare
iamzainhuda
added a commit
to iamzainhuda/torchrec
that referenced
this pull request
Oct 15, 2024
Summary: Added GPU sync tests to simulate gathering metric states on to rank 0 and computing. Tests don't cover this case before, which has resulted in SEVs in the past as users aren't aware of how RecMetrics collects and computes metrics. Added numerical value tests, most metrics are do not have this which can result in issues down the line if metrics need to be changed/accommodate future changes. Also we've found inconsistencies sometimes from other methods, so always good to check here. We compare each metric to a reference implementation from literature to ensure the values are as expected. Fixed and cleaned up some tests too. Reviewed By: henrylhtsang Differential Revision: D59173140
|
This pull request was exported from Phabricator. Differential Revision: D59173140 |
iamzainhuda
force-pushed
the
export-D59173140
branch
from
1356d1d
to
65f08fe
Compare
iamzainhuda
added a commit
to iamzainhuda/torchrec
that referenced
this pull request
Oct 16, 2024
Summary: **TLDR: Significant clean up of recmetrics tests and fixing long standing testing issues with particular metrics. Because RecMetrics is community sourced, the quality of metric and test implementations can have high variance which leads to SEVs later on due to the high surface area RecMetrics operates on.** Added GPU sync tests to simulate gathering metric states on to rank 0 and computing. Tests don't cover this case before, which has resulted in SEVs in the past as users aren't aware of how RecMetrics collects and computes metrics. Added numerical value tests, most metrics are do not have this which can result in issues down the line if metrics need to be changed/accommodate future changes. Also we've found inconsistencies sometimes from other methods, so always good to check here. We compare each metric to a reference implementation from literature to ensure the values are as expected. Fixed and cleaned up some tests, particularly enforcing a standard of quality that can be referenced for future metric implementations and tests. Reviewed By: henrylhtsang Differential Revision: D59173140
|
This pull request was exported from Phabricator. Differential Revision: D59173140 |
Summary: **TLDR: Significant clean up of recmetrics tests and fixing long standing testing issues with particular metrics. Because RecMetrics is community sourced, the quality of metric and test implementations can have high variance which leads to SEVs later on due to the high surface area RecMetrics operates on.** Added GPU sync tests to simulate gathering metric states on to rank 0 and computing. Tests don't cover this case before, which has resulted in SEVs in the past as users aren't aware of how RecMetrics collects and computes metrics. Added numerical value tests, most metrics are do not have this which can result in issues down the line if metrics need to be changed/accommodate future changes. Also we've found inconsistencies sometimes from other methods, so always good to check here. We compare each metric to a reference implementation from literature to ensure the values are as expected. Fixed and cleaned up some tests, particularly enforcing a standard of quality that can be referenced for future metric implementations and tests. Reviewed By: henrylhtsang Differential Revision: D59173140
iamzainhuda
force-pushed
the
export-D59173140
branch
from
65f08fe
to
6c353a8
Compare
|
This pull request was exported from Phabricator. Differential Revision: D59173140 |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Summary:
TLDR: Significant clean up of recmetrics tests and fixing long standing testing issues with particular metrics. Because RecMetrics is community sourced, the quality of metric and test implementations can have high variance which leads to SEVs later on due to the high surface area RecMetrics operates on.
Added GPU sync tests to simulate gathering metric states on to rank 0 and computing. Tests don't cover this case before, which has resulted in SEVs in the past as users aren't aware of how RecMetrics collects and computes metrics.
Added numerical value tests, most metrics are do not have this which can result in issues down the line if metrics need to be changed/accommodate future changes. Also we've found inconsistencies sometimes from other methods, so always good to check here. We compare each metric to a reference implementation from literature to ensure the values are as expected.
Fixed and cleaned up some tests, particularly enforcing a standard of quality that can be referenced for future metric implementations and tests.
Reviewed By: henrylhtsang
Differential Revision: D59173140