Rootstock¶

Rootstock is a Python library that lets you run many machine-learned interatomic potentials (MLIPs) on an HPC cluster from a single ASE-compatible calculator. Each MLIP family runs in its own pre-built, isolated Python environment that a maintainer has already installed and verified on the cluster, so you never resolve conflicting Python or library versions yourself. Swapping models is a one-line change to the checkpoint argument.

How do I use Rootstock?¶

Browse the Matter Model Almanac to find a model that is installed and verified on your cluster, then point a RootstockCalculator at it. Each model is already installed in an isolated environment, so changing models does not change your own environment.

How does Rootstock run a model?¶

You call the calculator from the lightweight rootstock library, which carries no model dependencies. Rootstock starts the model in a managed subprocess on the same GPU node, in the environment built for that cluster, loading weights from a cluster-local cache. Positions and forces are exchanged over a local Unix socket using the i-PI protocol. The model is loaded once and kept warm across calls.

How does Rootstock add models?¶

Maintainers define each model family as a Python file: a PEP 723 dependency list, a declaration of CHECKPOINTS in this family, and a setup() loader that returns an ASE calculator. They build the isolated environment and verify it on a GPU node. Automated testing re-verifies checkpoints periodically to catch regressions.

How to run¶