OSV-q grd_ x=grd_ x[:, :, 0:-1, :]-data_ in[:, :, 1:, :]IndexError: too many indices for tensor of d

 input_net, target_net = Variable(batch[0]), Variable(batch[1], requires_grad=False) 

torch.Size([1, 3, 356, 356]) torch.Size([1, 3, 356, 356])
torch.Size([1, 3, 356, 356]) torch.Size([1, 3, 356, 356])
Traceback (most recent call last):
  File "C:/Users/shang/Desktop/STDN_LI/TVGnet/OSV_q.py", line 359, in <module>
    train(opt)
  File "C:/Users/shang/Desktop/STDN_LI/TVGnet/OSV_q.py", line 286, in train
    grd_x, grd_y = getGrd(img_in)
  File "C:/Users/shang/Desktop/STDN_LI/TVGnet/OSV_q.py", line 37, in getGrd
    grd_x = grd_x[:, :, 0:-1, :] - data_in[:, :, 1:, :]
IndexError: too many indices for tensor of dimension 3
==>> Wed Mar  9 09:41:36 2022 [Iteration 1 Epoch 500]    Loss = 12718.471680
==>> Wed Mar  9 09:41:36 2022 [Iter 1/8]   loss = 12718.471680
====>> Wed Mar  9 09:41:36 2022   Pass time: 0:00:19.132388

Process finished with exit code 1

Change to True

input_net, target_net = Variable(batch[0]), Variable(batch[1], requires_grad=True) 

torch.Size([1, 3, 356, 356]) torch.Size([1, 3, 356, 356])
==>> Wed Mar  9 09:27:59 2022 [Iteration 1 Epoch 500]    Loss = 12005.107422
==>> Wed Mar  9 09:27:59 2022 [Iter 1/8]   loss = 12005.107422
====>> Wed Mar  9 09:27:59 2022   Pass time: 0:00:20.932553
Traceback (most recent call last):
  File "C:/Users/shang/Desktop/STDN_LI/TVGnet/OSV_q.py", line 359, in <module>
    train(opt)
  File "C:/Users/shang/Desktop/STDN_LI/TVGnet/OSV_q.py", line 286, in train
    grd_x, grd_y = getGrd(img_in)
  File "C:/Users/shang/Desktop/STDN_LI/TVGnet/OSV_q.py", line 37, in getGrd
    grd_x = grd_x[:, :, 0:-1, :] - data_in[:, :, 1:, :]
IndexError: too many indices for tensor of dimension 3

It seems different , But I don't know what's different

Decided to go from batch Functions are pushed up one by one

After trying many methods, I suddenly thought of a problem , Since it is all gradient function dimension error , The input image and network output dimensions are different , We define a new dimensional gradient function

Then this problem does not report an error, and the code goes forward one line , Bitter ......

==>> Wed Mar  9 11:11:56 2022 [Iter 1/8]   loss = 8833.374023
====>> Wed Mar  9 11:11:56 2022   Pass time: 0:00:17.003332
Traceback (most recent call last):
  File "C:/Users/shang/Desktop/STDN_LI/TVGnet/OSV_q.py", line 361, in <module>
    train(opt)
  File "C:/Users/shang/Desktop/STDN_LI/TVGnet/OSV_q.py", line 289, in train
    lap = -(cmul(fft2(grd_x), cconj(fft2(grd_x), inplace=False)) + cmul(fft2(grd_y), cconj(fft2(grd_y), inplace=False)))
  File "C:/Users/shang/Desktop/STDN_LI/TVGnet/OSV_q.py", line 138, in fft2
    return torch.fft.fft2(t, 2)
AttributeError: 'builtin_function_or_method' object has no attribute 'fft2'

Process finished with exit code 1