Garden's Almanac of Matter Models Model availability by cluster
These machine-learned interatomic potentials are pre-installed on shared HPC clusters. Rootstock lets you run them all through the same drop-in ASE calculator.
Polaris
ALCF - GPU types
- NVIDIA A100 40GB
● 0 verified ○ 40 installed // 22 not installed
Sophia
ALCF - GPU types
- NVIDIA A100 40GB
● 37 verified ○ 6 installed // 19 not installed
Perlmutter
NERSC - GPU types
- NVIDIA A100 40GB
● 39 verified ○ 20 installed // 3 not installed
Delta
NCSA - GPU types
- NVIDIA A100 40GB
● 47 verified ○ 4 installed // 11 not installed
Frontier
OLCF - GPU types
- AMD MI250X
● 15 verified ○ 0 installed // 47 not installed
Della
Princeton - GPU types
- NVIDIA A100 40GB
● 28 verified ○ 1 installed // 33 not installed
verified — ran in last 30 days installed — not recently verified not installed — not on this cluster
Running a model
Cluster admins pre-install each MLIP with its own dependencies.
You just install the lightweight rootstock package in your working environment and pick a supported checkpoint.
Behind the ASE calculator, Rootstock starts the model as a subprocess in the model’s pre-installed environment.
That lets you switch MLIPs without rebuilding your project dependencies.
Get started
# In a Perlmutter job/session
from ase.build import bulk
from rootstock import RootstockCalculator
atoms = bulk("Cu", "fcc", a=3.6) * (5, 5, 5)
with RootstockCalculator(
cluster="perlmutter",
checkpoint="mace-mp-0-medium",
device="cuda",
) as calc:
atoms.calc = calc
print(atoms.get_potential_energy())