There is an error in setting BCEWithLogitsLossFlat(pos_weight=dls.train.cws)

[aying0818]

Hello,

i am trying to use the BCEWithLogitsLossFlat(pos_weight=dls.train.cws) in my dataset(2 classes). I am getting the following error after passing the class weights to the loss function:

ValueError: Target size (torch.Size([128])) must be the same as input size (torch.Size([128, 2]))

The error code is:
learn = Learner(dls, model, metrics=accuracy, opt_func=Adam, loss_func=BCEWithLogitsLossFlat(pos_weight=dls.train.cws))
learn.lr_find()

Notably, the same error was reported when I tried to reproduce tutorial file '01a_Multi-class classification'. This is a 6-categorization problem.

ValueError: Target size (torch.Size([384])) must be the same as input size (torch.Size([2304]))
The error code is:
learn = ts_learner(dls, metrics=accuracy) # == ts_learner(dls, arch=InceptionTimePlus, metrics=accuracy) since InceptionTimePlus is the default arch
learn.lr_find()

I would appreciate your help in checking what is causing this. Looking forward to your reply.