Having more than one mixer can lead to improved predictions.
How
There was a check/guard in place to force NnMixer usage for time series tasks which this PR removes.
Additionally, this fixes a small bug in the confidence range reported for the default case in numerical tasks (when we're not able to find a bound with a specified width tolerance).
Why
Having more than one mixer can lead to improved predictions.
How
There was a check/guard in place to force
NnMixer
usage for time series tasks which this PR removes.Additionally, this fixes a small bug in the confidence range reported for the default case in numerical tasks (when we're not able to find a bound with a specified width tolerance).