A general fusion hub for the Princeton cluster designed to standardize fusion machine learning processes at Princeton University.
FAITH serves as a centralized platform for fusion-related machine learning workflows, providing standardized tools, processes, and background optimization for plasma control research at Princeton. This way, you can worry less about dataset/gpu optimiation and standardization. This platform primarily serves as a fast machine-learning trainer, but can also be used for general data analysis for fusion data stored at Princeton.
- Clone this repository onto anywhere on Stellar
- Get the location of this repository (you can find it by typing
pwdin the terminal) - In your main project's virtual environment or anaconda environment, type
pip install </path/to/FusionAIHub> - You can find tutorials on how to use this package in
docs,exampleandnotebooks
TODO: We will include this as a pip package in the future
NOTE: If you want to reload the package (after changes are made), use pip install --reinstall </path/to/FusionAIHub>
Go to your scratch directory while you are on the HEAD node (so you need internet access, which computing nodes do not have).
We will be using Python 3.12 and uv as a package manager. Since uv isn't on Stellar, for now we will install it via pip. First head over to your scratch directory with
cd /scratch/gpfs/[username]In your scratch directory, run
git clone git@github.com:PlasmaControl/FusionAIHub.git
cd FusionAIHub
git switch foundation25
module load anaconda3/2024.10
python -m venv .venv
conda deactivate
source .venv/bin/activate
pip install --upgrade pip
pip install uv
uv syncFrom now on, whenever you go into the repo, all you need to do is to run
source .venv/bin/activateFor more information, please contact
- Peter Steiner (ps9551@princeton.edu)
- Nathaniel Chen (nathaniel@princeton.edu)
- Kouroche Bouchiat (bouchiat@princeton.edu)
- Azarakash Jalalvand (aj17@princeton.edu)
- Egemen Kolemen (ekolemen@princeton.edu)