Support for sparse matrices / AnnData?

[maarten-devries]

Hi, Thank you for the great tool. I was wondering whether there is a plan to support sparse matrices (e.g. as used in AnnData files) as inputs? Right now, using a T4 GPU with 16GB GPU memory, the largest dataset I can run scFEA with is about ~10,000 cells x ~500 genes. This is largely because scFEA reads in a dense counts matrix into GPU memory. If this could be a sparse matrix in the future, then I would assume scFEA could scale to more cells.