[ 
https://issues.apache.org/jira/browse/SPARK-61697?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Hyukjin Kwon reassigned SPARK-61697:
------------------------------------

    Assignee: Hyukjin Kwon

> Add PySpark's Params API, typed models, summaries and persistence to the 
> native ML classes
> ------------------------------------------------------------------------------------------
>
>                 Key: SPARK-61697
>                 URL: https://issues.apache.org/jira/browse/SPARK-61697
>             Project: Spark
>          Issue Type: Improvement
>          Components: Connect
>    Affects Versions: connect-rust-4.3.0
>            Reporter: Hyukjin Kwon
>            Assignee: Hyukjin Kwon
>            Priority: Major
>              Labels: pull-request-available
>             Fix For: connect-rust-4.3.0
>
>
> The native ML classes of the Python client (pyspark.ml.connect) are to follow 
> the API of pyspark.ml 4.3.0 over Spark Connect, class by class. This adds 
> what each class builds on, and completes LinearSVC and KMeans with their 
> models and summaries as the worked examples:
> - PySpark's Params API (params, getParam, hasParam, isSet, isDefined, 
> hasDefault, getOrDefault, explainParam(s), extractParamMap, set, clear, 
> copy), of PySpark's own Param objects, with the docs, converters and defaults 
> of each official class (captured from pyspark-client 4.3.0), and the defaults 
> of each instance as PySpark keeps them (seed, outputCol, the defaults that a 
> fit copies from the estimator to its model, and those that a load keeps);
> - typed model and summary classes, whose members read the server's model 
> objects;
> - persistence (write/save/read/load, MLWriter, MLReader), fits of param maps 
> and fitMultiple;
> - in the Rust API, ObjectModel, training_summary, summary_transform and 
> evaluated_summary.



--
This message was sent by Atlassian Jira
(v10.3.26#10030026)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to