[QST] Lazy loading + rsc

[emdann]

From this tutorial I understand I should be able to lazily load my anndata object and perform processing with rsc.

Then why does this throw an error?

adata_ntc = anndata.experimental.read_lazy('object.h5ad')
rsc.get.anndata_to_GPU(adata_ntc)
rsc.pp.highly_variable_genes(adata_ntc, n_top_genes=5000, flavor='seurat_v3')

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Cell In[13], line 1
----> 1 _check_gpu_X(adata_ntc.X)

File ~/miniforge3/envs/rapids_singlecell/lib/python3.11/site-packages/rapids_singlecell/preprocessing/_utils.py:300, in _check_gpu_X(X, require_cf, allow_dask, allow_csc)
    298         return _check_gpu_X(X._meta, allow_csc=False)
    299     else:
--> 300         raise TypeError(
    301             "The input is a DaskArray. "
    302             "Rapids-singlecell doesn't support DaskArray in this function, "
    303             "so your input must be a CuPy ndarray or a CuPy sparse matrix. "
    304         )
    305 elif isinstance(X, cp.ndarray):
    306     return True

TypeError: The input is a DaskArray. Rapids-singlecell doesn't support DaskArray in this function, so your input must be a CuPy ndarray or a CuPy sparse matrix.

(Using rapids-singlecell-0.13.4)

[Intron7]

For HVG I currently only support cell_ranger and and seurat

[Intron7]

https://rapids-singlecell.readthedocs.io/en/latest/Out_of_core.html