See publication for explanation of theory and architecture as well as experimental results on AiMOS. Source code and experiment run scripts both for local, single-node, execution and distributed, multi-node, execution using slurm.
LoDAdaC/
├── config/ # configuration scripts for model setup
├── DOCS/ # further documentation
├── models/ # model definitions
├── src/ # source code
├── scripts/ # job scripts / utilities
├── data/ # input training data
├── requirements.txt # core Python dependencies
├── INSTALL.md # detailed setup instructions
└── README.md
- Python ≥ 3.9
- Linux environment (tested on HPC clusters)
- MPI implementation (e.g., OpenMPI)
git clone https://github.com/DecentralizedMethods/LoDAdaC
cd LoDAdaC
Then follow INSTALL.md. Once that is done, modify scripts/runscript.sh to load modules and the correct environment, and scripts/experiment.py to set the input parameters.
sbatch scripts/runscript.sh
Alternatively, for single-node execution, simply call:
mpirun -np [np] python3 -u -m scripts.experiment