Error when using scvi_mode to speed computation.

[Ethan-Earlie]

When setting scvi_mode=True for running Mefisto with muon.tl.mofa(), errors pop up that define_mini_batch is not defined for multiple nodes.

AttributeError: 'Z_GP_Node_mv' object has no attribute 'define_mini_batch'

After looking into this, it appears that many nodes defined in mofapy2/core/ do not have a define_mini_batch method defined.

Would you please be able to help with this? I am not sure if these nodes need define_mini_batch() to be defined, or if the Sigma_node.py code should not call this method in the first place?

Without scvi_mode=True, Mefisto is running for days using a GPU for a dataset of 11k cells and 6k features, across 2 groups and 5 time points - without ever completing. Is there a better way to speed up this computation?

Thanks,

Ethan