Ruifeng Zheng created SPARK-60608:
-------------------------------------
Summary: Apply the class allowlist to nested model readers
Key: SPARK-60608
URL: https://issues.apache.org/jira/browse/SPARK-60608
Project: Spark
Issue Type: Bug
Components: MLlib
Affects Versions: 5.0.0
Reporter: Ruifeng Zheng
Nested model readers should apply the same ML class-loading policy as top-level
model loading in Spark Connect.
DefaultParamsReader.loadParamsInstanceReader currently resolves the class
recorded in nested model metadata directly. Route that resolution through
MLAllowListedLoader, which is already used by default parameter loading.
Add a separate MLClassLoadingSuite with checked-in model fixtures. A minimal
RFormulaModel containing an allowlisted VectorAssembler should load
successfully; the corresponding fixture naming an unregistered nested stage
should fail with the allowlist error. The unregistered class is an inert
placeholder with no executable payload.
Pull request: https://github.com/apache/spark/pull/59854
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
