INSTALL The interpretation document of
Help file
-
there pytorch=1.0.1 torchvision=0.2.2 cudatoolkit=9.0, This ATSS Can run .
-
It's better to have a look at cuda Version of :
1.cat /usr/local/cuda/version.txt
2. perhaps nvcc -v
- no module named cv2:
resolvent :
Use command
pip install opencv-python
At the same time, when downloading the package , Often encounter in the existing channel It is not found inside or the server is very slow in foreign countries , This is the time to add channels 了 .
In China, we recommend Tsinghua or 、 CUHK or aliyuan , Run the following code :
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/msys2/
And then according to anconada Or follow the code below to view the existing channels:
conda config --set show_channel_urls yes
reference :Anaconda In other countries ( Tsinghua source 、 China University of science and technology )
How to create Python A virtual environment (Pytorch)
function python When the script appears no module named cv2 How to solve
problem RuntimeError: cannot perform reduction function min on tensor with no elements because the operation
terms of settlement : The label problem , I'm using voc Of xml label . stay labelimg Open the label and find that the upper right corner has been checked difficult This option , After removing it, save it and use it .
In fact, the image has no corresponding tag file , No load on labels file , So there's no way to calculate loss, You can see if there are any loaded into the corresponding image labels file
At this time , I'll be rude kill process , And then restart the program .
Here are some possible solutions :
Help file : Be sure to run servers with large memory and disk space deep learning Code for
- No usable temporary directory found in ['/tmp', '/var/tmp', '/usr/tmp', .....
The essential problem is that there is no disk space , In the end, even the records couldn't be written .
Answer
That's python on the server side saying "Hey!, none of those 4 directories are places I can write temporary files!". Check to make sure at least one of them exists, can be written to by the user mercurial is running as on the server side, and make sure the disk isn't full.
The solution is a word : Delete !!
References : How to solve linux newspaper No space left on device error
No space left on device solve Linux How to make the system disk full