Some time ago, I have been connected to another server in the server, not to mention the shortage of computing resources. A group of people shared an account environment and changed, and mmdetection could not be installed according to the official tutorial, and various errors could not be solved. Today, my brother suddenly said that we installed putty on our account on this server, so we can remotely connect to the one we usually use on this server. I tried it and found that it works, but the interface is ugly and the function is relatively low. According to the tutorial, mmdetection will be installed in a while.
It’s the first time I heard about putty, but I’m still wondering why we can remotely connect to the one we usually use through putty on this server, but can’t connect to that one directly with xshell?
1. The following error occurred during the installation process, which is huge, and I didn't cut the very long section in the middle. There are different opinions on the Internet. I tried the following method and it succeeded once.
Method: Install cython
although it reminds me that I already have it, but when I run it again, it is inexplicable and does not report an error.
2. During the installation process, there was a problem about cuda, the specific error is not important. It's like this:
the cuda soft link of the server is connected to version 10.1, and cuda 9.0 is connected to cuda under my own account. But the nvcc-V command view is version 7.5. After wondering for a long time, I went to my bashrc to check the cuda path, and found that it was a soft link of 9.0 under my own account, and the source was over, and the nvcc check was still 7.5. Try to change the path in the environment variable back to /usr/local/bin. After source, nvcc becomes 10.1. From this point of view, the cuda of my own account is invalid. I haven't figured out the reason yet. I will use version 10.1 for now, mmdetection is supported.