lucidrains/tab-transformer-pytorch

Only continuous variables

TheodoreGalanos opened this issue · 3 comments

Hi,

Silly question, but is it possible to train on only continuous variables?

Thanks!

@TheodoreGalanos yea there is! but in which case, it would just reduce down to a simple MLP

there is no attention being used for the continuous variables (unless if you discretize them)

Aha got it thank you so much! I'll turn some to categorical and see what happens :)

Continuing with this, is it possible to train just using categorical features? Should I change the forward(self, x_categ, x_cont) of the class TabTransformer or it should be able to handle empty x_cont?