gene number

[XiejcSYSU]

thanks for sharing the codes.
I have a question about the number of genes. In adata_train.parquet, there are 20000+ genes, but in train_or_control_bulk_by_cell_type_adata.h5ad and de_train.parquet, there are 18000+ genes. Why the number of genes is different? When I dropped the genes in bulk_adata that not in the fixed_bulk_adata, the code: assert np.allclose(fixed_bulk_adata.X, bulk_adata.X) still throw an error.