-
Notifications
You must be signed in to change notification settings - Fork 1.9k
Updated libmf and corresponding MatrixFactorizationSimpleTrainAndPredict() baselines per build #5121
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.
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
Conversation
…ict() baselines per build
test/Microsoft.ML.Tests/TrainerEstimators/MatrixFactorizationTests.cs
Outdated
Show resolved
Hide resolved
frank-dong-ms-zz
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
![]()
| @@ -54,7 +54,6 @@ public void MatrixFactorization_Estimator() | |||
| } | |||
|
|
|||
| [MatrixFactorizationFact] | |||
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Is MatrixFactorizationFact needed? #Resolved
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Actually it is not! Currently, MatrixFactorizationFact is just an EnvironmentSpecificFactAttribute that always returns true with IsEnvironmentSupported(). In a seperate PR, I'll be marking these MatrixFactorizationFact marked tests as Fact, and if possible, remove the MatrixFactorizationFact from the codebase all together. #Resolved
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Codecov Report
@@ Coverage Diff @@
## master #5121 +/- ##
=======================================
Coverage 75.55% 75.55%
=======================================
Files 995 995
Lines 179316 179318 +2
Branches 19298 19298
=======================================
+ Hits 135479 135482 +3
- Misses 38572 38575 +3
+ Partials 5265 5261 -4
|
Fixes #4874
This PR updates the libmf submodule, where its recent changes in libmf PR #41 and PR #42 address the shuffling of values in a given matrix.
The C++ function
void random_shuffle(_RanIt _First, _RanIt _Last)is implemented differently on Windows vs. MacOS vs. Linux. This resulted in inconsistent results on MacOS and more-predictable yet still inconsistent results on Linux. As a result, a given matrix factorization problem, even with a constant seed, produced differing MSEs between each run on each system. The libmf codebase has been updated to prevent this, and baseline MSE values in the unit testMatrixFactorizationSimpleTrainAndPredict()have been updated to reflect this.