This is a package for hyperparameter optimization of the ZH dark photon NN using ATLAS grid GPU/CPU resources.
It contains the following directories:
- hpogrid: the main work-horse
python: contains the training script to be usedhpo_scripts: contains the script needed to create the configurations to run/submit to the grid
- Setup the
ml-baseconda environment
source hpogrid/setupenv.sh
- Create the configurations and project needed to run/submit to grid. Here you will define the name of the project!
source hpo_scripts/DarkPhoton_NN.sh
- Run the hyperparameter optimization ...
- locally:
hpogrid run <project_name>
- on the grid:
hpogrid submit <project_name> --n_jobs <number of jobs>
- If you ran on the grid, you can monitor the progress of your jobs by
hpogrid tasks show -d <n days since jobs were submitted> -n *.<project_name>.*
- When your grid jobs are done, you can see the results by doing
hpogrid report <project_name> -d <n days since jobs were submitted> --to_html
There are options to save to html, csv (--to_csv) or mlflow (--to_mlflow)