>>>x=np.arange(10)>>>x[::-2]array([9,7,5,3,1])>>>y=torch.arange(10)>>>y[::2].flip(dims=(0,))tensor([8,6,4,2,0])>>>y[1::2].flip(dims=(0,))tensor([9,7,5,3,1]) For numpy the negative step size starts from the last index (convenient and intuitive), yet ...