Unable to define `categorical_covariate_keys` for `scvi.model.LinearSCVI`

[marcfranquesa]

I am able to set categorical_covariate_keys for SCVI but not for LinearSCVI. Is this the intended behavior?

Thank you!

Model = scvi.model.LinearSCVI 

Model.setup_anndata(
    adata,
    layer="counts",
    categorical_covariate_keys=["gene"], # works for SCVI
)

Traceback (most recent call last):
  File ".../src/train.py", line 67, in <module>
    fire.Fire(main)
  File ".../.venv/lib/python3.10/site-packages/fire/core.py", line 135, in Fire
    component_trace = _Fire(component, args, parsed_flag_args, context, name)
  File ".../.venv/lib/python3.10/site-packages/fire/core.py", line 468, in _Fire
    component, remaining_args = _CallAndUpdateTrace(
  File ".../.venv/lib/python3.10/site-packages/fire/core.py", line 684, in _CallAndUpdateTrace
    component = fn(*varargs, **kwargs)
  File ".../src/train.py", line 54, in main
    Model.setup_anndata(
  File ".../.venv/lib/python3.10/site-packages/scvi/model/_linear_scvi.py", line 180, in setup_anndata
    adata_manager.register_fields(adata, **kwargs)
  File ".../.venv/lib/python3.10/site-packages/scvi/data/_manager.py", line 177, in register_fields
    raise TypeError(
TypeError: register_fields() got unexpected keyword arguments {'categorical_covariate_keys': ['gene']} passed without a source_registry.

Versions:

1.3.3

VERSION

[ori-kron-wis]

Yes. It was never implemented.
LinearSCVI is strictly linear: its latent representation is a linear projection of the input gene expression matrix.
The model does not provide hooks to ingest or adjust for covariates the way SCVI does.
You can go ahead and try to implement it
Why not using SCVI in the meantime?