a=torch.rand(2,2,3,4)
a[a>0.5] = 1 ###a>0.5
a[a<=0.5] = 0
print(a)
index = torch.nonzero(a==0)
index = index[index[:,3] != (a.size(3)-1)]
index[:,3] = index[:,3]+1
a[index[:,0], index[:,1], index[:,2], index[:,3]] = 0
print(a)
a=torch.tensor(range(24)).view(2,2,3,2)
b=torch.tensor(range(24)).view(2,2,3,2)
c = torch.rand(2,2,3,2)
c[c>0.5] = 1 ###a>0.5
c[c<=0.5] = 0
b[c.byte()]
这里的b和c是一样的维度,所以b[c.byte()]就是把矩阵c中为1位置上的b的值取出来,b[c.byte()]的维度为[10],表示有10个满足条件的值