Apply FAIRChem UMA models with ASE integration for materials science simulations.
npx skills add https://github.com/Hongyu-yu/matsci-ai-skills --skill fairchem-hongyu-yu
Please help me install this Agent Skill. Skill: fairchem Source: https://github.com/Hongyu-yu/matsci-ai-skills/tree/main/skills/fairchem Command: npx skills add https://github.com/Hongyu-yu/matsci-ai-skills --skill fairchem-hongyu-yu
FAIRChem provides expert guidance for using UMA-based machine-learning potentials to accelerate materials science and quantum chemistry workflows, replacing expensive DFT with fast, accurate predictions across catalysis, materials, molecules, MOFs, and molecular crystals.
Load a pretrained UMA model with load_predict_unit, create a FAIRChemCalculator for your domain (e.g., oc20, omat, omol, odac, omc), and use it with ASE to perform an optimization or MD workflow.
High-intent search queries and answers about installing and using this skill.
Machine learning potentials can replace DFT by using FAIRChem UMA models to accelerate materials science simulations. You load a pretrained model via load_predict_unit, configure a FAIRChemCalculator with ASE, and run fast predictions across catalysis, bulk materials, and molecules.
ASE integration with multi-GPU support is available for large-scale molecular dynamics and optimization. You create a FAIRChemCalculator with the appropriate task_name for your domain, optionally enable turbo mode for speed, and execute the workflow directly within ASE.
UMA models support multiple domains for machine learning chemistry, including oc20 for catalysis, omat for bulk materials, omol for molecules, odac for MOFs, and omc for molecular crystals. Each domain requires specifying the correct task_name when creating the FAIRChemCalculator.
To start catalysis screening with FAIRChem UMA models, load a pretrained model using load_predict_unit, create a FAIRChemCalculator with the oc20 task_name, and integrate it with ASE to perform the end-to-end screening workflow.
Turbo mode should be enabled for materials science simulations when you need maximum computational speed for large-scale tasks. It is an optional setting in FAIRChemCalculator that accelerates predictions alongside multi-GPU support for intensive workflows.
Validating machine learning potentials for molecular crystals involves using the omc domain UMA model within the FAIRChem framework. Following best practices for model selection and domain-specific validation ensures accurate predictions when replacing expensive DFT workflows.
jkitchinRuns ML-driven chemistry predictions with UMA models via the FAIRChemCalculator for ASE workflows.
tondevrelBuild, manipulate, and analyze atomic structures for atomistic simulations.
liongkjAutomate molecular simulation workflows from natural language prompts using LangGraph and ASE.
learningmatter-mitCalculate phonon band structures and thermal properties using MLIPs.
