batch size and nz

[csgwon]

I was not able to get the CIFAR10 example to run with the defaults, since the expand_as call expects the expanding dimension to be greater or equal to the first dimension (and the defaults have the batch size as 64 and nz as 128). If I set the batch size and the nz sizes the same, it seems to work fine.

[blancaag]

Hi,

I encountered same issue. The problem is in the "dim" parameter of the normalisation functions (src/utils.py), when the noise == 'sphere':

def normalize(x, dim=1):
    '''
    Projects points to a sphere.
    '''
    return x.div(x.norm(2, dim=dim).expand_as(x))

def normalize_(x, dim=1):
    '''
    Projects points to a sphere inplace.
    '''
    x.div_(x.norm(2, dim=dim).expand_as(x))

It can be solved by setting dim=0, w/o altering the normalisation.

[ZuowenWang0000]

@DmitryUlyanov @blancaag @csgwon Does change the dim from =1 to 0 gives the correct solution?