GPU
The CSE department currently have 10 Ubuntu GPU severs, i.e. GPU7-GPU16. GPU7-GPU9 each has 8 Titanx GPU cards. And GPU10-GPU16 each has 4 GPU cards. All the servers have the basic (Caffe, Torch, Tensorflow, Theano) deep learning tools intalled. But before you using the servers, you should first apply for the GPU servers refering the following link: http://corner.cse.cuhk.edu.hk/fac/gpgpu-hpc.html
Suppose the CSE account name is xxx (e.g. xjqi) and we use GPU8 as an example.
How to login to CSE Server?
step1: ssh [email protected]
(Use terminal in Ubuntu system and ssh software, e.g. Putty in Windows system http://www.cse.cuhk.edu.hk/~byu/doc/windows2server.pdf )
step2: The system will require your password, type your CSE password which should be the same as you login your own computer.
step3: cd to your work folder cd /data/ssd/public
Where can I store the data?
After creating your account, you should have the access to the server. Then you cd to the data folder. In each gpu server, the path is "/data/ssd/public".
step1: cd /data/ssd/public
step2: mkdir xxx
Then you can store your files here.
How can I use Tensorflow?
Before using any deeplearning platform, make sure the cuda path is included in your environment variable by checking the .bashrc in your login folder.
Tensorflow is installed in all the servers and what you should do is to use python to execute the code.
The command is "python a.py"