Data, model and prediction files to accompany "Evaluating 2D and 3D Graph Encoders for Infrared-Based Organic Structure Verification"

This deposit contains the GFN2-xTB and DFT calculation results, trained PASGeN model checkpoints and prediction outputs used in “Evaluating 2D and 3D Graph Encoders for Infrared-Based Organic Structure Verification”. The models are phase‑aware IR spectrum predictors with 2D MPNN and 3D PaiNN encoders, trained to output normalized spectra on the 500–3500 cm⁻¹ window. Training uses the curated NIST IR dataset (13,568 phase‑annotated spectra), including single‑split and 10‑fold cross‑validation runs. Results on the dataset for simple ensembles (MPNN+PaiNN) are also provided.