ml-foundation-potentials

To choose a foundation MLIP for molecular dynamics, map your scientific requirements to model families based on chemistry type, desired fidelity level like r2SCAN-grade, and simulation mode, prioritizing cheaper smaller models for expensive dynamic workflows.

Selecting an MLIP for phase stability involves searching the local model registry for existing fine-tuned checkpoints for your chemical system first, then choosing an appropriate foundation model if no suitable checkpoint is available.

Yes, MACE and MatGL models can be evaluated for both organic and inorganic compositions by applying fallback selection logic using model cost, dynamic suitability, and chemistry-specific guidance to determine the best fit.

You do not need to fine-tune a foundation MLIP if the local model registry already contains an existing checkpoint for your chemical system, allowing you to reuse fine-tuned models instead of selecting a new foundation model.

Model cost matters for NEB, phonons, and diffusion calculations because these are expensive dynamic workflows where prioritizing cheaper smaller foundation MLIP models helps balance computational expense with accuracy requirements.

Yes, the ml-foundation-potentials Skill supports FAIRCHEM model selection by mapping scientific requirements across MatGL, FAIRCHEM, and MACE environments to recommend the most suitable model for your atomistic simulation task.