OSV_ Q write divergence operator div and Laplace stepped on the pit

  In order to write the image matrix of Laplace , You need to use the divergence operator 
 Start copying other people's code 

def divergence(x, y):
     divx = x - np.roll(x, 1, axis=1)
     divy = y - np.roll(y, 1, axis=-1)
     div = divx + divy
     return div

 Report errors ：  Because it's not supported Tensor and list Division operation of type 

# loss1 = miu2 * mse(u1, f1 - dive + b2)    #  Because it's not supported Tensor and list Division operation of type 
# loss1 = torch.optim.Adam(f1 - dive + b2, lr=0.1)      #  l2 Regularization 

Report errors ： Because it's not supported Tensor and Add Division operation of type

Write Laplace in the most primitive way

zxx, zxy = getGrd(zx)
zyx, zyy = getGrd(zy)
dive = zxx + zyy

.......

Although there are other new mistakes , To move forward

Report errors ：

dive1 = zxx + zyy
RuntimeError: The size of tensor a (320) must match the size of tensor b (318) at non-singleton dimension 3

Maybe I did two gradients , Dimensions are inconsistent , Choose to zero the boundary of the gradient ：

def getGrd(data_in):
    # [m, n] = data_in.shape
    # data_in1 = torch.Tensor(np.pad(data_in, ((0, 0), (0, 0), (0, 0), (0, 0)), 'edge'))
    grd_x = data_in.clone()
    grd_y = data_in.clone()
    grd_x = grd_x[:, :, 0:-1, :] - data_in[:, :, 1:, :]
    # grd_x = torch.Tensor(np.pad(grd_x, ((0, 0), (0, 0), (0, 1), (0, 0)), "constant"))  #  Zero compensation 
    grd_y = grd_y[:, :, :, 0:-1] - data_in[:, :, :, 1:]
    # grd_y = torch.Tensor(np.pad(grd_y, ((0, 0), (0, 0), (0, 0), (0, 1)), "constant"))
#  Zero compensation 
    grd = (grd_x, grd_y)
    return grd