[SPARK-58509][ML][CONNECT] Exclude OneVsRest classifier from model size estimation - #57715
Closed
zhengruifeng wants to merge 1 commit into
Closed
[SPARK-58509][ML][CONNECT] Exclude OneVsRest classifier from model size estimation#57715zhengruifeng wants to merge 1 commit into
zhengruifeng wants to merge 1 commit into
Conversation
uros-b
approved these changes
Aug 3, 2026
Member
|
Thank you @zhengruifeng! |
zhengruifeng
added a commit
that referenced
this pull request
Aug 3, 2026
…ze estimation ### What changes were proposed in this pull request? This follow-up to SPARK-58250 excludes `OneVsRestModel.classifier` from parameter-metadata sizing. It continues to account explicitly for label metadata and the fitted child models. The regression test initializes the source classifier logger before fitting OneVsRest, ensuring the estimate does not retain shared runtime state through the classifier param. ### Why are the changes needed? The fitted OneVsRestModel retains the base classifier estimator in its param map. That estimator can retain shared runtime state, including SparkSession-related state, which must not contribute to Spark Connect ML model-cache accounting. ### Does this PR introduce _any_ user-facing change? No. ### How was this patch tested? Updated the OneVsRestModel estimated-size regression test. Static checks: `git diff --check`, non-ASCII scan, and changed-Scala line-length scan. The targeted Scala suite was not run locally. ### Was this patch authored or co-authored using generative AI tooling? Generated-by: Codex GPT-5 Closes #57715 from zhengruifeng/one-vs-rest-model-size-dev3. Authored-by: Ruifeng Zheng <ruifengz@apache.org> Signed-off-by: Ruifeng Zheng <ruifengz@foxmail.com> (cherry picked from commit 51f1b64) Signed-off-by: Ruifeng Zheng <ruifengz@foxmail.com>
Contributor
Author
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
2 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.
What changes were proposed in this pull request?
This follow-up to SPARK-58250 excludes
OneVsRestModel.classifierfrom parameter-metadata sizing. It continues to account explicitly for label metadata and the fitted child models.The regression test initializes the source classifier logger before fitting OneVsRest, ensuring the estimate does not retain shared runtime state through the classifier param.
Why are the changes needed?
The fitted OneVsRestModel retains the base classifier estimator in its param map. That estimator can retain shared runtime state, including SparkSession-related state, which must not contribute to Spark Connect ML model-cache accounting.
Does this PR introduce any user-facing change?
No.
How was this patch tested?
Updated the OneVsRestModel estimated-size regression test.
Static checks:
git diff --check, non-ASCII scan, and changed-Scala line-length scan.The targeted Scala suite was not run locally.
Was this patch authored or co-authored using generative AI tooling?
Generated-by: Codex GPT-5