ALIGNN/ALIGNN-FF: Pure PyTorch Models for Materials Property Prediction

Pure PyTorch implementations of the ALIGNN and ALIGNN-FF models, released under a CC-BY-4.0 license. The 384.0 MB ZIP file contains machine learning models for predicting material properties, authored by Kamal Choudhary. It was last updated on May 20, 2026.

Use Cases

• Predicting material properties based on atomic graph representations.
• Fine-tuning or deploying the ALIGNN-FF force field model for molecular simulations.
• Benchmarking pure PyTorch implementations against other deep learning frameworks for materials informatics.

Strengths

• Model files are provided in a 384.0 MB ZIP archive, indicating substantial pre-trained parameters.
• The description explicitly states the models are 'Pure Pytorch (DGL-free)', a specific technical feature.

Limitations

• Column-level documentation is absent; model architecture and input/output specifications must be inferred after download.
• The dataset's row count and specific training data are unknown, which may limit suitability assessment.

Provenance

Source

Kamal Choudhary

Freshness

Last updated 2026-05-20 17:40:36.