Is there a way to use MAPE instead of MAE?

[aegek]

I was searching for it but couldn't find anything related

[cbergmeir]

Hi, you mean using MAPE as a loss function to optimise? That tends to not work very well.

[aegek]

Yes I was, doesn't it differ from dataset to dataset?

[ourownstory]

@aegek you can use any loss function defined in torch.nn.modules.loss._Loss
You can even define your own, inheriting from torch.nn.modules.loss._Loss.

If you simply want to record MAPE as a metric, this will be supported when #185 gets merged.