Fix Multi-processor in LinUCB training#684
Closed
alexnikulkov wants to merge 1 commit intofacebookresearch:mainfrom
Closed
Fix Multi-processor in LinUCB training#684alexnikulkov wants to merge 1 commit intofacebookresearch:mainfrom
alexnikulkov wants to merge 1 commit intofacebookresearch:mainfrom
Conversation
Summary: 1. We train LinUCB models on two hours (7am and 8am) separately and get covariance matrixes A1 and A2 for each. 2. We train a LinUCB model at 7am, enable recurring so it launched a recurring training at 8am. Then we get covariance matrix A. We found that A is around 8*A1 + A2 instead of A1 + A2, where we used NPROC_PER_NODE = 8 when training. The reason is the following: Current implementation aggregates A and b calculated in each training processor. However, if we do recurring training, then in the 2nd training child, it will aggregate the previous A NPROC_PER_NODE times and then add the new A aggregated from current epoch. This diff fixes the above issue. Reviewed By: alexnikulkov Differential Revision: D39902734 fbshipit-source-id: 73a9e9525e68d75e2dd558b75e1db8a3bea7255a
|
This pull request was exported from Phabricator. Differential Revision: D39902734 |
Codecov ReportBase: 87.52% // Head: 87.52% // Increases project coverage by
Additional details and impacted files@@ Coverage Diff @@
## main #684 +/- ##
=======================================
Coverage 87.52% 87.52%
=======================================
Files 365 365
Lines 23446 23454 +8
Branches 44 44
=======================================
+ Hits 20521 20529 +8
Misses 2899 2899
Partials 26 26
Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here. ☔ View full report at Codecov. |
xuruiyang
pushed a commit
that referenced
this pull request
Sep 20, 2025
Summary: Pull Request resolved: #684 1. We train LinUCB models on two hours (7am and 8am) separately and get covariance matrixes A1 and A2 for each. 2. We train a LinUCB model at 7am, enable recurring so it launched a recurring training at 8am. Then we get covariance matrix A. We found that A is around 8*A1 + A2 instead of A1 + A2, where we used NPROC_PER_NODE = 8 when training. The reason is the following: Current implementation aggregates A and b calculated in each training processor. However, if we do recurring training, then in the 2nd training child, it will aggregate the previous A NPROC_PER_NODE times and then add the new A aggregated from current epoch. This diff fixes the above issue. Differential Revision: D39902734 fbshipit-source-id: 58ab1fade9feebd6b0419b18dd085a3944736f23
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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:
We found that A is around 8*A1 + A2 instead of A1 + A2, where we used NPROC_PER_NODE = 8 when training.
The reason is the following:
Current implementation aggregates A and b calculated in each training processor. However, if we do recurring training, then in the 2nd training child, it will aggregate the previous A NPROC_PER_NODE times and then add the new A aggregated from current epoch.
This diff fixes the above issue.
Reviewed By: alexnikulkov
Differential Revision: D39902734