BrotherHa
BrotherHa
Hello @MilesCranmer Thank you very much for the quick reply. Is the format ok like that? Best regards
I have searched a bit where the issue occurs and in fact for most individuals of the regression the loss values from the equations_-table and the recalculated ones are absolutely...
About the loss functions I am very sure. The only issues here might be problems caused by the precision of calculation, etc. And for more than 90% of all individuals...
> > 90% of all individuals both methods calculate exactly the same loss > > Do you see any pattern in the ones that don’t match? e.g., Are they very...
> Just to confirm, it is only when you load the model from a saved checkpoint that this happens? It’s not when you use `warm_start=True` and call `fit` a second...
The Problem occured in the first place when fitting a new model. Then directly after the regression finished I called predict() and calculated the loss for training, test and total...