Following https://pytorch-tabular.readthedocs.io/en/latest/tutorials/13-Using%20Model%20Sweep%20as%20an%20initial%20Model%20Selection%20Tool/ I was able to adapt the model sweep to my problem. But i would like to use a custom loss during model sweep. For tm = TabularModel(...) it was straight forward to leverage a custom loss via the fit()-interface: tm.fit(..., loss=my_custom_loss()). However, have the feeling that currently this is not possible for model_sweep().

Using pytorch_tabluar==1.1.1