[
https://issues.apache.org/jira/browse/SPARK-61010?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Ruifeng Zheng updated SPARK-61010:
----------------------------------
Fix Version/s: connect-rust-4.3.0
> Fit native Rust LogisticRegression and predict with its cached model
> --------------------------------------------------------------------
>
> Key: SPARK-61010
> URL: https://issues.apache.org/jira/browse/SPARK-61010
> Project: Spark
> Issue Type: Bug
> Components: Connect
> Affects Versions: connect-rust-4.3.0
> Reporter: Ruifeng Zheng
> Priority: Major
> Fix For: connect-rust-4.3.0
>
>
> The native Rust LogisticRegression.fit implementation currently ignores its
> training DataFrame and constructs an unfitted model locally. Its transform
> serializes an estimator-like operator instead of the fitted model object
> reference expected by Spark Connect.
> Send MlCommand.Fit with the estimator, parameters and actual training
> relation. Validate and retain the returned cached model reference, use it for
> prediction, and reject prediction in another session. Test learned
> predictions on held-out rows from two fits with opposite labels, model
> cloning and deferred execution, and training errors. Exercise the Rust-backed
> Python binding against the same behavior as official PySpark's JVM
> LogisticRegression.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
