Jupyter Notebook¶
You can easily create a jupyter notebook session via Open Ondemand platform. You can find it in the Interactive Applications section. To launch the session you need to fill in the form:
Account - fill in your account number. You can find it out from the command line with command sprojects -a
or from the portal.
Partition - select partition from the drop down menu.
Number of hours - fill in wallclock time for the job.
Number of nodes - select number of required nodes. Usually one node is enough. In case you need more than one node you need to use tools such as MPI to fully use them.
Number of cores - select number of required cpu cores. Each node has 64 cores.
Num Gpus - select number of required GPUs. In case GPU is required, you need also to specify gpu partition, otherwise this parameter is ignored. Each GPU node has 4 graphic cards.
Memory - specify memory requirement in MB. Each node has 256GB of memory.
Modules - specify modules that will be loaded into environment. Use module avail
command in the command prompt to see all available modules. Use space as a separator. Use JupyterHub for testing purposes or as a default environment.
Conda Environment - If you have created your conda environment, specify the name of the environment here. Make sure the environment contains jupyter notebook.
Venv Path - If you have created your own python virtual environment, specify absolute path here. Make sure the environment contains jupyter notebook. If you used python version loaded via modules, specify that version in the Modules.