-
Notifications
You must be signed in to change notification settings - Fork 4
Using CSC Taito GPU

Then input your university account and password, and login.
In your dashborad

Click the Sign Up in the red box. After submitting your application, the "Add Service" icon will disappear. Then you are able to login to the CSC server.
Open you putty

Then click Open and give the authentication information you just signed up
Various software has been prepared by the server and well packed so that you can use them easily. e.g.
module avail
Then you will see available packages, like

You can see caffe has already been installed.
If you want to clean packages have been loaded, you can use
´module purge´
if you want to load a specific package, you can use
´module load caffe´
You can use
´which caffe´
to check the caffe you have loaded. This caffe was built with OpenCV 2.4.10 and Python 3.4.5. But you can build your own caffe if you want to use Python 2.7. We will talk about this later.
Load python environment
´module load python-env/2.7.10´
Check right python has been loaded
´python -V´
Down load your own caffe and place it in the right place
Modifying the Makefile.config

´make all -j32´ ´make pycaffe´
The CSC taito-GPU server has already provided
module purge module avail module avail ##e.g. module avail caffe module load caffe
This caffe was built with OpenCV 2.4.10 and Python 3.4.5. If you want build your own caffe based on Python 2.7
module purge module load python-env/2.7.10 python -V
Requirements: Putty, Xming


Then you connect the Taito-GPU server with Putty again. Then put following command in the terminal
´srun -n 1 -N 1 --gres=gpu:k80:1 --x11=first -p gpu nsight´ Then a Nsight window will appear on your screen.