Hello.
I am experimenting an issue that on previous versions I did not have.
When computing the cross entropy error with the parameter reduce=False I got this error:
RuntimeError: expand(torch.cuda.FloatTensor{[100]}, size=[]): the number of sizes provided (0) must be greater or equal to the number of dimensions in the tensor (1)
I am evaluating torch.nn.functional.cross_entropy(out,y,reduce=False)
where y has shape (100,) and out has shape (100,10)
Can anyone give me some feedback? I have seen on other posts that people experimented this issue when calling backward on a Loss.
Thanks.