STEPHY trained models: simulation benchmarks and the Denmark application

Trained model weights, normalisation parameters, held-out test predictions and training histories for STEPHY: A Graph Neural Inference Framework for Rapid Estimation of Regional Epidemic Dynamics from Large Viral Phylogenies. Contains the simulation benchmarks (one main 100k-tree experiment, a similar-population-size experiment, and three 5k training-set-size replicates, each with a STEPHY model and a CBLV ablation) together with the heads trained for the Denmark SARS-CoV-2 application. These artefacts are required to regenerate the figures produced by supportingFigures/ in the code repository; unpack the archive and set STEPHY_MODELS to the resulting directory.