-
Notifications
You must be signed in to change notification settings - Fork 4
Use Jupyter Notebooks on Orion
The Orion Jupyter notebook loads the module load python/3.7.5 module. While this will work for some things, it will not include tools needed for map plots, etc.
Run the following script:
/work/noaa/da/cmartin/Python/jupyter/add_conda_ufo_kernel.sh
This script will copy the necessary files needed for Jupyter Notebook to use Kevin Dougherty's UFO conda environment which contains all you need to do plotting and analysis in Python.
Run the following script:
/path/to/placeholder.sh
Because on Orion, the Jupyter notebooks run on compute nodes, there is no internet access, and thus the code cannot download shape files. This will stage them in the proper location for your user.
Orion has a nice web interface where you can request an interactive session of things like RStudio, MatLab, Jupyter Notebooks, or even a remote desktop.
Please visit https://orion-ood.hpc.msstate.edu/ and login like you normally would to Orion.
Once logged in, at the top, click 'Interactive Apps', then click 'Jupyter Notebook' (towards the bottom).
On this page, make sure your account, partition, QoS, etc. are what you are wanting to request. Note: debug only gets you 30 minutes max of runtime, but will spawn the fastest. For Jupyter, 1 node and 1 task are probably sufficient. Remember we are charged for these hours!
If you are satisfied with your configuration, click the Launch button at the bottom of the page. You will be taken to a page with a message such as this: please be patient while your request sits in the queue.
Once you see Connect to Jupyter, click that button and you will open the Jupyter interactive session. This will open a page that should show your home directory's contents. From here you can open a notebook, or create a new one.
To make sure you are using the correct Python kernel. In the notebook, go to the Kernel menu, then Change kernel, and select UFO. This will ensure that all of the necessary import statements should work as intended.
When finished, make sure to click the red Delete button on the Orion Dashboard page where you clicked Connect to Jupyter so that the instance stops and we are not charged for unused CPU hours.