OSV-q ValueError: axes don‘t match array
z 子问题报错 AttributeError: ‘numpy.ndarray‘ object has no attribute ‘clone‘
于是转类型，在代码里加了一行
z = torch.from_numpy(z)又回到损失函数了呀，这行代码好像见过，尝试按原来的改一下
torch.Size([1, 3, 320, 320]) torch.Size([1, 3, 320, 320])
Traceback (most recent call last):
File "C:/Users/shang/Desktop/STDN_LI/TVGnet/OSV_q.py", line 396, in
train(opt)
File "C:/Users/shang/Desktop/STDN_LI/TVGnet/OSV_q.py", line 303, in train
loss = loss_func(input_net, output_net, miu2, b2)
File "C:/Users/shang/Desktop/STDN_LI/TVGnet/OSV_q.py", line 99, in loss_func
b2 = b2.transpose(2, 0, 1)
ValueError: axes don't match array
Process finished with exit code 1
b2 = b2.permute(2, 0, 1)好吧，不对，和之前的错误不一样
torch.Size([1, 3, 320, 320]) torch.Size([1, 3, 320, 320])
Traceback (most recent call last):
File "C:/Users/shang/Desktop/STDN_LI/TVGnet/OSV_q.py", line 396, in
train(opt)
File "C:/Users/shang/Desktop/STDN_LI/TVGnet/OSV_q.py", line 303, in train
loss = loss_func(input_net, output_net, miu2, b2)
File "C:/Users/shang/Desktop/STDN_LI/TVGnet/OSV_q.py", line 99, in loss_func
b2 = b2.permute(2, 0, 1)
AttributeError: 'numpy.ndarray' object has no attribute 'permute'
Process finished with exit code 1 改回去，研究上面的问题
ValueError: axes don't match array，翻译过来是：轴不匹配数组
http://t.csdn.cn/el99Q 关于tranpose的解释
，因b2定义的是和f相等，维度已经不需要改变了，所以把关于b2的这几行代码注释掉了
# dive = divergence(zx, zy)
# b2 = b2.transpose(2, 0, 1)
# b2 = b2[np.newaxis,:,:,:] # 给数组增加维度
# b2 = np_to_torch(b2).to(device)
f1 = f1.cuda()
dive1 = dive1.cuda()
# b2 = b2.cuda()然后就有了新错误，cuda和cpu不统一
File "C:/Users/shang/Desktop/STDN_LI/TVGnet/OSV_q.py", line 303, in train
loss = loss_func(input_net, output_net, miu2, b2)
File "C:/Users/shang/Desktop/STDN_LI/TVGnet/OSV_q.py", line 108, in loss_func
loss1 = miu2 * mse(u1, u11)
File "D:\ProgramData\Anaconda3\envs\python36\lib\site-packages\torch\nn\modules\module.py", line 727, in _call_impl
result = self.forward(*input, **kwargs)
File "D:\ProgramData\Anaconda3\envs\python36\lib\site-packages\torch\nn\modules\loss.py", line 446, in forward
return F.mse_loss(input, target, reduction=self.reduction)
File "D:\ProgramData\Anaconda3\envs\python36\lib\site-packages\torch\nn\functional.py", line 2660, in mse_loss
return torch._C._nn.mse_loss(expanded_input, expanded_target, _Reduction.get_enum(reduction))
RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu!
Process finished with exit code 1
这一行的错误， loss1 = miu2 * mse(u1, u11)，然后调用这个函数时u1的值是网络输出。
input_net = input_net.to(device)
# 原来的
output_net = output_net
# 改为
output_net = output_net.to(device)还是不对，同样的报错
b2 = torch.tensor(b2).cuda()可以了，下面是数组类型问题
====>> Sun Mar 13 11:06:48 2022 Pass time: 0:00:17.732670
Traceback (most recent call last):
File "C:/Users/shang/Desktop/STDN_LI/TVGnet/OSV_q.py", line 396, in
train(opt)
File "C:/Users/shang/Desktop/STDN_LI/TVGnet/OSV_q.py", line 345, in train
b2 = b2 + f - output_net - divergence(zx, zy)
TypeError: unsupported operand type(s) for +: 'numpy.ndarray' and 'Tensor'
Process finished with exit code 1 z的类型是numpy, output_net 是 Tensor，梯度函数定义的是对 Tensor形式的.clone(),
于是在对z求梯度时使用了新的梯度函数 ,定义中使用numpy的函数 .copy()
还是不对，Debug发现，q1,q2,b1,b2，conjoDx，conjoDy，lap都是numpy.ndarry形式的，所以
# 原来
q1 = np.zeros(f.shape)
q2 = np.zeros(f.shape)
b1 = np.zeros(f.shape)
b2 = np.zeros(f.shape)
# 改为
q1 = torch.zeros(f.shape)
q2 = torch.zeros(f.shape)
b1 = torch.zeros(f.shape)
b2 = torch.zeros(f.shape)报错
====>> Sun Mar 13 11:37:09 2022 Pass time: 0:00:17.131145
Traceback (most recent call last):
File "C:/Users/shang/Desktop/STDN_LI/TVGnet/OSV_q.py", line 395, in
train(opt)
File "C:/Users/shang/Desktop/STDN_LI/TVGnet/OSV_q.py", line 317, in train
pdf1 = miu2 * (lap * fftn(f - output_net - b2)) - karma * (conjoDx * fftn(q1) + conjoDy * fftn(q2))
RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu!
Process finished with exit code 1 决定统一为tensor类型，所以把定义的fft,divgence函数内部都改为torch了，另外，定义的复共轭这里也改成torch，试一下结果吧
Dh_psf = np.array([[0, 0, 0], [1, -1, 0], [0, 0, 0]])
Dv_psf = np.array([[0, 1, 0], [0, -1, 0], [0, 0, 0]])
# 新加入的两行
Dh_psf = torch.from_numpy(Dh_psf)
Dh_psf = torch.from_numpy(Dh_psf)
otfDx = psf2otf(Dh_psf, [nh, nw])
otfDy = psf2otf(Dv_psf, [nh, nw])
conjoDx = conj(otfDx)
conjoDy = conj(otfDy)
还是不行
====>> Build Net
Traceback (most recent call last):
File "C:/Users/shang/Desktop/STDN_LI/TVGnet/OSV_q.py", line 397, in
train(opt)
File "C:/Users/shang/Desktop/STDN_LI/TVGnet/OSV_q.py", line 245, in train
Dh_psf = torch.from_numpy(Dh_psf)
TypeError: expected np.ndarray (got Tensor)
Process finished with exit code 1 http://t.csdn.cn/rFoeH
在q1,q2,b1,b2,后面都加了.to(device)
可以了，代码继续运行，还需要检查一下定义的散度函数
Traceback (most recent call last):
File "C:/Users/shang/Desktop/STDN_LI/TVGnet/OSV_q.py", line 396, in
train(opt)
File "C:/Users/shang/Desktop/STDN_LI/TVGnet/OSV_q.py", line 345, in train
b2 = b2 + f - output_net - divergence(zx, zy)
File "C:/Users/shang/Desktop/STDN_LI/TVGnet/OSV_q.py", line 88, in divergence
div = divx + divy
ValueError: operands could not be broadcast together with shapes (1,2,320,320) (1,3,319,320)
Process finished with exit code 1 重新定义了散度函数，然后主程序已经可以运行了，后续还有一点小问题
def divergence(zx, zy):
zxx, zxy = getGrd2(zx)
zyx, zyy = getGrd2(zy)
dive = zxx + zyy
return dive====>> Sun Mar 13 15:10:04 2022 Pass time: 0:00:16.247797
Traceback (most recent call last):
File "C:/Users/shang/Desktop/STDN_LI/TVGnet/OSV_q.py", line 387, in
train(opt)
File "C:/Users/shang/Desktop/STDN_LI/TVGnet/OSV_q.py", line 353, in train
img_v = (v.cpu().data[0].numpy().astype(np.float32)) * 255.0
AttributeError: 'numpy.ndarray' object has no attribute 'cpu'
Process finished with exit code 1 因为v是numpy数组，所以将v转为tensor类型的
v = torch.tensor(v)运行应该是可以了，运行时间也明显增加 ，但是又回到了一个之前的问题。
====>> Sun Mar 13 16:37:38 2022 Pass time: 0:00:31.486364
Traceback (most recent call last):
File "C:/Users/shang/Desktop/STDN_LI/TVGnet/OSV_q.py", line 390, in
train(opt)
File "C:/Users/shang/Desktop/STDN_LI/TVGnet/OSV_q.py", line 309, in train
pdf1 = miu2 * (lap * fftn(f - output_net - b2)) - karma * (conjoDx * fftn(q1) + conjoDy * fftn(q2))
RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu!
Process finished with exit code 1