Description
Currently, we write the code for serialization/deserialization in Python and Scala manually, it can not scale to the big number of MLlib API.
If the serialization could be done in pickle (using Pyrolite in JVM) in extensional way, then it should be much easy to add Python API for MLlib.
Attachments
Issue Links
- is related to
-
SPARK-4023 PySpark's stat.Statistics is broken
- Resolved
-
SPARK-3486 Add PySpark support for Word2Vec
- Resolved
- links to