All dataset files including their variations have been uploaded on Google Drive, which can be found here:
You files you should find in each folder are dataset_name_default.h5ad, dataset_name_undersampled.h5ad, dataset_name_oversampled.h5ad, dataset_name_imputed.h5ad, and dataset_name_test.h5ad.
For scBERT, follow the preprocess.py code in this link before running the code.