CODEWITHSUNDEEP
pytorch
Zhen Peng
Zhen Peng

Reputation: 31

pytorch:RuntimeError: dimension out of range (expected to be in range of [-1, 0], but got 1)

When i using the cross-entropy loss as a loss function,i get this dimension out of range error.

This is my code:

self.ce = nn.CrossEntropyLoss()
    def forward(self, pred, y):
        loss = 0
        for w_, p_, y_ in zip(self.weights, pred, y):
            loss += w_ * self.ce(p_, y_)
        return loss

when i run this code :

the value of p_:tensor(1.00000e-02 *[-0.7625,  5.8737], device='cuda:0')
the value of w_:tensor(1., device='cuda:0')
the value of y_:tensor(0, device='cuda:0')

Upvotes: 1

Views: 2711

Answers (1)

Steven
Steven

Reputation: 5162

For cross entropy there should be the same number of labels as predictions.

In your specific case the dimensions of y_ and p_ should match which they don't as y_ is a 0 dimensional scalar and p_ is 1x2.

Upvotes: 1

Related Questions