When to pass a number to nn.CrossEntropyLoss()?

What does nn.CrossEntropyLoss(10) mean?

Could I just use nn.CrossEntropyLoss()?

Using nn.CrossEntropyLoss() is to create a object of CrossEntropyLoss, you could only pass the parameter once when initialized it. If you wanna dynamic adjust the weight, you can use torch.nn.functional.cross_entropy() method while training

1 Like