Binary Cross Entropy loss
- typically used for binary classification problems learn more about cross entropy loss at cross-entropy loss
related: BCEWithLogitsLoss (which is basically BCELoss but with a sigmoid applied to the logits first). You should use it instead in pytorch (cause it has stable gradients)