This is the code for this paper.
This code is written to run on a slurm cluster. It also runs elsewhere but there will be differences.
Before running build the singularity container:
apptainer build --nv container.sif container.def
You can then run a script using, for instance:
apptainer run --nv container.sif python ./scripts/train.py
Make sure you bind the needed directories (data etc.)
To override parameters use hydra CLI syntax, e.g.:
apptainer run --nv container.sif python ./scripts/train.py train.lr=0.0001 train.total_steps=100000
See this paper on how to get the QCML dataset in array_record format.