Adjust how choose_classifier handles seed parameters #222
Conversation
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
We recently got automatically updated to Debian trixie with Java 21. But that seems to cause problems for the current version of XGBoost.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.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.
This closes #221.
It also closes #224, a CI/CD bug.
Previously, we were manually setting the seeds for Spark's built-in ML models but not for XGBoost and LightGBM. This inconsistency is an oversight I made while adding XGBoost and LightGBM. Since we weren't setting the seed for XGBoost or LightGBM, the models trained by these libraries were slightly different on each run of hlink. This caused some inconsistent results from matching.
Also, the manual setting of the seeds for the Spark models did not allow users to pass in their own seeds, so they were stuck with the single seed we had chosen.
Now all of these models are handled uniformly. We accept the seed set by the user if there is one. If there is no seed in the
paramsdictionary, then we add a"seed": 2133entry before passing the parameters to the classifier. This fixes both issues.