Partitioned model is not calculating explanations when scoring new data

Options
dataikuer07
dataikuer07 Registered Posts: 1 ✭✭✭

Hello everybody,

I created a partitioned model based on geographic regions and then deploy it to the project's flow. When I use a predict recipe to get predictions on new not partitioned data that contains the column of geographic region, it calculates the predictions but not the explanations and then a warning message is shown:

PARTITION_MODE_UNSUPPORTED occurred 1 times
Computing explanations is not supported with partition redispatch

My doubt is: does anyone know how to calculate predictions and their corresponding explanations when using a partitioned model ? (image attached of the warning message)

Thanks a lot !

Answers

  • louisplt
    louisplt Dataiker Posts: 21 Dataiker
    Options

    Hello @dataikuer07
    ,

    Indeed, you cannot compute predictions with their explanations on unpartitioned data using a partitioned model.

    If you want to be able to compute them on your new data, you would need to first partition this data using a sync recipe.

    Hope this helps,

    Louis

Setup Info
    Tags
      Help me…