skills.rest
learningmatter-mit
learningmatter-mitOFFICIAL
Explore all skills by this author →
WebsiteGitHub

ml-property-predict-scd

Train atomistic property prediction models from SelfConditionedDenoisingAtoms checkpoints.

144|21|Updated Jan 8, 2026
Search Tags:#finetuning#materials science#property prediction#selfconditioneddenoisingatoms#molecular ML#graph embeddings#dataset onboarding
One-click install
npx skills add https://github.com/learningmatter-mit/AtomisticSkills --skill ml-property-predict-scd
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: ml-property-predict-scd
Source: https://github.com/learningmatter-mit/AtomisticSkills/tree/main/.agents/skills/ml-property-predict-scd
Command: npx skills add https://github.com/learningmatter-mit/AtomisticSkills --skill ml-property-predict-scd

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

Enables training or transfer of property-prediction models from SelfConditionedDenoisingAtoms (SCD) foundation checkpoints, reducing the effort to build ML pipelines for atomistic materials and molecular property datasets.

Core Features & Use Cases

  • Frozen SCD encoder embeddings: reuse pretrained SCD to generate mol_emb graph-level features (and optionally atom_embs) for downstream ML.
  • Lightweight head adaptation: train scalar_head, atom_emb_mlp, or mol_emb_mlp while keeping the SCD backbone frozen for faster iteration.
  • Full-model finetuning or pretraining: run upstream train.py for full finetuning or pretraining from scratch on new datasets.
  • Dataset onboarding guidance: provides a dataset contract for returning torch_geometric.data.Data with required fields for scalar/energy-force/periodic tasks.

Quick Start

Use the SCD frozen-backbone embedder to produce mol_emb features for a new molecular dataset task.

Frequently Asked Questions about ml-property-predict-scd

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I predict molecular properties using pretrained graph embeddings?▼

You can predict molecular properties by using frozen SelfConditionedDenoisingAtoms encoder embeddings to generate mol_emb graph-level features for training lightweight downstream ML heads.

Can I finetune atomistic models for both molecules and periodic materials?▼

Yes, full-model finetuning supports both molecules and periodic materials, requiring correct checkpoint selection for each domain and enabling allow_periodic for periodic graph handling.

What is the dataset contract for training energy and force prediction models?▼

The dataset contract requires returning torch_geometric.data.Data objects containing the specific fields necessary for scalar, energy-force, or periodic property prediction tasks.

Do I need a specific Conda environment to train SCD property prediction models?▼

Yes, you must configure a scd-agent Conda environment to run the SCD foundation checkpoints for atomistic property prediction, lightweight head training, or pretraining from scratch.

What is the best way to adapt a pretrained materials science model for a new dataset?▼

The best way is lightweight head adaptation: train a scalar_head, atom_emb_mlp, or mol_emb_mlp while keeping the SCD backbone frozen to iterate faster on new datasets.

Why does my periodic material property prediction fail during dataset onboarding?▼

Periodic material prediction fails if the dataset contract is violated or if allow_periodic and noise_in_loader settings are not correctly configured for periodic graph handling.

Related Skills

Explore More Education & Research Skills→
learningmatter-mitlearningmatter-mit
144

ml-matgl-finetune

Fine-tune MatGL interatomic potentials on labeled atomistic datasets with PyTorch Lightning.

Official
Advanced
jkitchinjkitchin
34

idaes

Model, simulate, and optimize chemical processes with IDAES flowsheets.

Community
Advanced
jasrajtulsijasrajtulsi

torch-geometric

Builds PyTorch Geometric workflows for node, link, and graph tasks with sampling and explainability.

Community
Advanced
Hung-3008Hung-3008
1

deepchem

Run end-to-end molecular property prediction workflows with DeepChem.

Community
Advanced