sc_reconstruction.models.ReconMLSCVI#

class sc_reconstruction.models.ReconMLSCVI(*args, **kwargs)[source]#

Bases: SCVI, BaseReconstructionModel

A structured SCVI model for reconstruction that adheres to the base class interface.

Parameters:

  • adata (AnnData | None)

  • registry (dict | None)

  • n_hidden (int)

  • n_latent (int)

  • n_layers (int)

  • dropout_rate (float)

  • dispersion (Literal['gene', 'gene-batch', 'gene-label', 'gene-cell'])

  • gene_likelihood (Literal['zinb', 'nb', 'poisson', 'normal'])

  • use_observed_lib_size (bool)

  • latent_distribution (Literal['normal', 'ln'])

  • library_log_means (np.ndarray | None)

  • library_log_vars (np.ndarray | None)

__init__(adata=None, registry=None, n_hidden=128, n_latent=10, n_layers=1, dropout_rate=0.1, dispersion='gene', gene_likelihood='zinb', use_observed_lib_size=True, latent_distribution='normal', library_log_means=None, library_log_vars=None, **kwargs)[source]#
Parameters:

  • adata (AnnData | None)

  • registry (dict | None)

  • n_hidden (int)

  • n_latent (int)

  • n_layers (int)

  • dropout_rate (float)

  • dispersion (Literal['gene', 'gene-batch', 'gene-label', 'gene-cell'])

  • gene_likelihood (Literal['zinb', 'nb', 'poisson', 'normal'])

  • use_observed_lib_size (bool)

  • latent_distribution (Literal['normal', 'ln'])

  • library_log_means (np.ndarray | None)

  • library_log_vars (np.ndarray | None)

Methods

__init__([adata, registry, n_hidden, ...])

load(path[, map_location])

Load the model

predict(X[, inference_kwargs, pred_type])

Parameters: X: np.ndarray - Raw input data.

prepare(adata, **kwargs)

Register adata with the SCVI data manager and cache the reference.

save(path, **kwargs)

Save the model module to path.

train(**train_kwargs)

Drop the unused save_path kwarg and delegate to SCVI.train.