[Question] Time Frame of performance_over_time plot less than actual training time

[kefa23]

Hello everyone,

I tried to plot the 'performance_over_time_' and am puzzled because the time frame on the x-axis is much smaller than the actual training time I defined in the parameter 'time_left_for_this_task'.
The defined runtime was ~550min, but the chart (and also the underlying data frame) only shows a time frame of ~300min.
Can you tell me why this is the case? Does it have something to do with the offline and online phase of AutoSklearn?

PS: I noticed that in the documentation there is no mention of the 'performance_over_time_' attribute for the AutoSklearn2Classifier class, maybe that's the reason? I was just wondering because you can call the attribute even if it's not listed in the documentation.

Thanks for the help!