Conditional Generative Modeling for Amorphous Multi-Element Materials

"data.zip" : the training dataset. It includes : "poscar": 10000 random structures in POSCAR format "cif": the random structures in CIF format "train_set_scaled.csv", "test_set_scaled.csv", "val_set_scaled.csv": the standardized PDMs(pair distribution matrices) of the training set, test set, and validation set, respectively "with_cif_train.csv", "with_cif_test.csv", "with_cif_val.csv" : the training set, test set, and validation set, respectively, which saves both the structure information(CIF) and PDMs. "train.feather","test.feather", "val.feather": the "feather" format of "with_cif_train.csv", "with_cif_test.csv", "with_cif_val.csv" "train.pkl","test.pkl","val.pkl": the serialized (pickle) format of "with_cif_train.csv", "with_cif_test.csv", "with_cif_val.csv" "scaler_stats.txt" : the mean values and standard deviations of the training set "models.zip“ : the trained Cond-CDVAE model checkpoints used for structure generation and evaluation "apollox.yaml": the configuration YAML file matching the setup used in the paper "eval_recon.zip": the reconstructed 1000 structures used to compute reconstruction errors, uniqueness and coverage "final_generated_structures.zip": the newly generated 500 structures used to compute novelty fraction