about the loss in train_binclass.py about icpr2020dfdc HOT 1 CLOSED

Hey @algorithmlixuan ,

Hi, thanks for the great work.

thank you for checking out!

Would you like to tell me why you write your BCEloss code like that? thank you very much

The reason is that we use as criterion the BCEWithLogitsLoss, which computes for us the sigmoid and then the BCE: therefore, we can pass to the function directly the output out of the network and not the vector pred of predictions.
You can alternatively use as criterion = BCELoss() (BCELoss) and then compute loss = criterion(pred, labels), passing in this case to the loss function the pred vector.
Hope this answers your question!
Cheers :)

Edoardo

from icpr2020dfdc.