AtomGen & vector-inference

Open foundation-model tooling from the Vector Institute engineering team — an atomistic-materials FM and its widely reused HPC serving stack.

Outputs 2

model

<p>An encoder-transformer atomistic-materials foundation model trained on the aggregated S2EF-15M dataset (OC20/22, MPtrj, SPICE).</p>

Architecture DENSE

library

<p>Efficient LLM inference and serving on Slurm HPC clusters; 106★ and actively maintained.</p>

Language Python

sciencematerialsinfrastructureopen-weight