跑DID-MDN代码遇到的问题

《Density-aware Single Image De-raining using a Multi-stream Dense Network》
我的环境：win10，python3.6，pytorch1.2.0

1.

ImportError: bad magic number in 'myutils': b'\x03\xf3\r\n'

解决办法：

删除myutils文件夹中的.pyc文件

2.

ModuleNotFoundError: No module named 'torch.utils.serialization'

因为新版torch中删掉了load_lua

解决办法：

安装torchfile：pip install torchfile（虚拟环境中请使用conda install torchfile）

将utils.py第7行中的from torch.utils.serialization import load_lua改为import torchfile,

将utils.py第90行中的load_lua改为torchfile.load

3.

RuntimeError: DataLoader worker (pid(s) 10740) exited unexpectedly

解决办法：

test.py中57行：parser.add_argument('--workers', type=int, help='number of data loading workers', default=1)

将default=1改为default=0
(Windows系统中num_workers>0时会出现bug，所以只能设置为0)

4.

  File "D:\1AA\DID_MDN\models\derain_dense.py", line 484, in forward
    label_d11.data.resize_((1, 8, sizePatchGAN, sizePatchGAN)).fill_(label_result)
RuntimeError: set_sizes_contiguous is not allowed on a Tensor created from .data or .detach().
If your intent is to change the metadata of a Tensor (such as sizes / strides / storage / storage_offset)
without autograd tracking the change, remove the .data / .detach() call and wrap the change in a `with torch.no_grad():` block.
For example, change:
    x.data.set_(y)
to:
    with torch.no_grad():
        x.set_(y)

解决办法：

将derain_residual.py和derain_dense.py中的：label_d11.data.resize_改为：label_d11.resize_

两个文件中大概有三处，代码具体位置我记不太清楚了，可以根据报错提示查找

Reference:
https://blog.csdn.net/kent_xyli/article/details/89187558