ALIGNN Trained Models For JARVIS-DFT Materials Properties

1.7 GB of pre-trained ALIGNN models for predicting properties from the JARVIS-DFT materials database. The models were created by Kamal Choudhary and updated in March 2026. This resource is shared under a CC BY 4.0 license.

Use Cases

• Apply pre-trained ALIGNN models to predict JARVIS-DFT materials properties for new candidate structures.
• Fine-tune the provided ALIGNN models on specific subsets of JARVIS-DFT data for specialized property forecasting.
• Use the models as a benchmark for developing new graph neural network architectures on condensed matter data.

Strengths

• The 1.7 GB size indicates a substantial collection of model weights and configurations.
• Models are based on the established JARVIS-DFT dataset for materials science.
• Recent update in March 2026 suggests maintained relevance.

Limitations

• The specific number of models, their performance metrics, and the exact JARVIS-DFT properties they predict are not detailed.
• Without column or sample data, the direct input features and output targets for the models are unclear.

Provenance

Source

figshare, authored by Kamal Choudhary.

Collection Method

Models trained on the JARVIS-DFT dataset using the ALIGNN architecture.

Time Range

null

Freshness

Last updated March 2026.

Geography

null