You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Marek Novotny commented: Alternative solution is to throw exception if H2OMOJOModel is being serialized:
H2OMOJOModel is currently Java serializable and serves as a closure for Spark UDFs that are responsible for model scoring. However, H2OMOJOModel which has been serialized, moved to different Spark application and deserialized can't be used for scoring. Such a H2OMOJOModel will get disconnected from H2O-3 mojo artifacts that were located on Spark executors of the original cluster.
So the solution would be to create a class that will play role of a closure for Spark UDFs and make H2OMOJOModel unserializable.
Marek Novotny commented: [~accountid:5fa438f822f3990076aa232d] I’ve created the PR [https://github.com//pull/2742|https://github.com//pull/2742|smart-link] to cover that.
For details, refer to support ticket: https://support.h2o.ai/a/tickets/100534
The text was updated successfully, but these errors were encountered: