-
Notifications
You must be signed in to change notification settings - Fork 25
Open
Labels
Description
As introduced in #924 and to be extended in #906, the metrics will be returnable without reduction by mean across all dims.
For MSLL the metric returns through reshaping the output to the original shape.
It would be preferable to instead ensure the correct shape is returned.
An option for this is to use the similar approach with the ._fit() and ._predict() methods in Emulator and introduce a ._call() for metric that can be subclassesd by abstract but with the __call__ potentially calling another function to ensure correct shapes given aggregation method that is passed.
This issue might be best addressed as part of #906 but opening an issue in case it's implemented separately.
Metadata
Metadata
Assignees
Labels
Type
Projects
Status
📋 Product backlog