multiple sample version of Annotating cell clusters by integrating single-cell RNA-seq data

hey kaizhang:
great tools for single cell chromatin analysis. especially the ['fragment_paired'] is integrated into adata.obsm make quantification on consensus extremely fast.

I recently using snapatac2 to analysis my dataset. which consists of scATAC and scRNA both with multiple sample (not paired multiome).
following the tutorial I transfer label from scRNA annotation to scATAC

here is my question, in this tutorial,
scvi.model.SCVI.setup_anndata(data, batch_key="batch")
adata was setup with batch_key 'batch' which is used to distinguish scRNA data and gene activity generated by scATAC. and now given my data is all consist of multiple sample. is it necessary to add this sample-batch to batch_key of scvi.model.SCVI.setup_anndata.
or only batch to distinguish scRNA and scATAC is all I need despite multiple samples?

and as I can search, the parameter batch_key of scvi.model.SCVI.setup_anndata can only accept one obs.
so may be a new batch = sample batch + scRNA/scATAC batch is feasible?
like

Start with the annotation tutorial and the linked scvi.model.SCVI.setup_anndata documentation, focusing on how batch_key is described for multiple scRNA and scATAC samples. Clarify the supported configuration and update the tutorial or its notes so the expected batch fields and completion criteria are explicit.