Jupyter
Jupyter provides a notebook interface/format for interactive calculation.[1]
Methods of running Jupyter
There are three methods to run Jupyter:
- as a job on a compute node (including GPUs and bigmem nodes) using Slurm, presenting in a browser running on your PC. See: Jupyter Hub
- using the web browser on your PC, with the Jupyter kernel running on
picotte001. See: Jupyter via VS Code - using remote display, running a browser on a compute node displaying to your PC. This is very laggy. See this demonstration video on Drexel Streams.
The second method is recommended unless you have a large computation to do.
Installing Packages Inside a Jupyter Notebook
DO NOT DO THE FOLLOWING
!conda install --yes some_package!pip install some_package%pip install some_package
For details, please see this blog post.
Instead, create a virtualenv in a normal terminal session. The virtualenv is like a Conda env.