To use Jupyter with a Python virtualenv
- On the login node, select a python version to use, activate your virtualenv and install the jupyter and jupyterlab modules using pip - e.g.
source ~/venvs/3.10.4/bin/activate to the "Commands to run before Jupyter"
To use Jupyter with Anaconda
- On the login node, load the Anaconda module, activate your conda env, and install the Jupyter and Jupyterlab modules using pip - e.g.
conda activate Your_Conda_Environment in "Commands to run before Jupyter" section.
To use Jupyter with GPU, select "Jupyter/3.5.0" in the Modules section, put "module load CUDA/11.7.0" in "Commands to run before Jupyter" section, and change the "Number of GPUs" to be 1 or more
Login to spartan via ssh and activate your existing virtual env (venv-3.10.4)
Install ipykernel pip package
Now integrate that with jupyter notebook. You can choose a different kernel name, but I just went with the same name "venv"
NOTE: some users may end up needing to install ipyparallel package as well in their venv
Check the list of kernels. We now have "venv" also listed
The ~/.local/share/jupyter/kernels/spartan-venv/kernel.json points to the python executable inside your vitual environment. Jupyter will automatically pick this up when you select the "spartan-venv" option.
In opendemand, create a jupyter notebooks session with the following input values:
Add the dependency "Modules" field to ensure that the right python modules are loaded.
Remove the activate command from "Commands to run before Jupyter"
Start the jupyter session and select the spartan-venv from the available kernels and start a new notebook. You may select "spartan-venv" by clicking the drop-down-box (labeled as "New") near the top left corner to open a new notebook. For existing notebooks, you may change the kernel by selecting "Kernel -> > Change Kernel -> spartan-venv"