Datasets for OpenPyStruct

StructDataLite contains 100,000 samples and serves as a compact resource for initial training, debugging, and exploratory analysis. It balances coverage of geometric configurations and boundary conditions.

StructDataHeavy contains 2,000,000 samples, designed for high-fidelity model training. This dataset provides extensive variability across structural configurations and load cases, ensuring realistic and robust training scenarios.

Both lite and meta datasets include features such as roller positions, force locations and magnitudes, beam node coordinates, and derived physics outputs containing deflections, rotations, shear and moments, supporting potential for advanced structural analysis and optimization.