MatterGen is a generative model for inorganic materials design that can be fine-tuned for property-constrained generation across the periodic table. The dataset includes pre-trained model checkpoints for unconditional generation and property-conditioned generation, with conditioning on properties such as chemical system, space group, magnetic density, band gap, bulk modulus, and energy above hull. It also contains reference datasets for evaluation, including Alex-MP-20 and MP-20 datasets with correction schemes, provided via Git LFS. The dataset is designed for generating, evaluating, and training inorganic materials structures, with applications in materials discovery and property-directed generation.