Functionality to save the trained model

[ShayaNovelis]

Is there a way to save the trained model or its weights when using MirroredStrategyRunner?

[marcialf]

Only chief node can save the model.

https://keras.io/guides/distributed_training/

[alexperry-shifu]

Is there an example of saving the model when starting with your simple Spark example?

[rahulnatarajan98]

Are there any updates on this?

[tomersagi]

Only chief node can save the model.

https://keras.io/guides/distributed_training/

ok, but how do you tell the train function to save the model only on the chief node in the Spark API?