Hi everyone,
I am applying the GRU model to the benchmark tasks and noticed that the MAE appears to be on a different scale than the original data. When inspecting the targets and predictions passed to the MAE computation, they also appear scaled.
It seems the preprocessing step scales the features but the inverse transform may not be applied before evaluation. Could you confirm whether the targets and predictions are being unscaled prior to MAE calculation?
Thank you for developing such a useful package!
Best,
Xenia
Hi everyone,
I am applying the GRU model to the benchmark tasks and noticed that the MAE appears to be on a different scale than the original data. When inspecting the targets and predictions passed to the MAE computation, they also appear scaled.
It seems the preprocessing step scales the features but the inverse transform may not be applied before evaluation. Could you confirm whether the targets and predictions are being unscaled prior to MAE calculation?
Thank you for developing such a useful package!
Best,
Xenia