Hyukjin Kwon created SPARK-61760:
------------------------------------
Summary: Read the params of the object of a native ML writer at
its save, as PySpark's RemoteMLWriter does
Key: SPARK-61760
URL: https://issues.apache.org/jira/browse/SPARK-61760
Project: Spark
Issue Type: Improvement
Components: Connect
Affects Versions: connect-rust-4.3.0
Reporter: Hyukjin Kwon
A native ML writer (the MLWriter of write() of an estimator, a model or an
evaluator of pyspark.ml.connect) takes the params of its object at write(),
where PySpark's RemoteMLWriter reads the params of its _instance at its save:
so w = lr.write(); lr.setMaxIter(7); w.save(path) saves maxIter=7 with PySpark
and the old value with this client.
This makes the native writer hold its object (exposed as PySpark's _instance)
and read its params at the save, of the native writer of its class after the
object's param maps are synced; a writer of PySpark's Python format (a
Pipeline, a validator) syncs its object at the save too.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]