scvi.module.base.BaseModuleClass

class scvi.module.base.BaseModuleClass

Bases: Module

Abstract class for scvi-tools modules.

Notes

See further usage examples in the following tutorials:

1. Constructing a probabilistic module

Attributes table

device
training

Methods table

forward(tensors[, ...])	Forward pass through the network.
generative(*args, **kwargs)	Run the generative model.
inference(*args, **kwargs)	Run the recognition model.
loss(*args, **kwargs)	Compute the loss for a minibatch of data.
on_load(model, **kwargs)	Callback function run in load().
sample(*args, **kwargs)	Generate samples from the learned model.

Attributes

BaseModuleClass.device

BaseModuleClass.training: bool

Methods

BaseModuleClass.forward(tensors, get_inference_input_kwargs=None, get_generative_input_kwargs=None, inference_kwargs=None, generative_kwargs=None, loss_kwargs=None, compute_loss=True)

Forward pass through the network.

Parameters:

• tensors – tensors to pass through

• get_inference_input_kwargs (dict | None (default: None)) – Keyword args for _get_inference_input()

• get_generative_input_kwargs (dict | None (default: None)) – Keyword args for _get_generative_input()

• inference_kwargs (dict | None (default: None)) – Keyword args for inference()

• generative_kwargs (dict | None (default: None)) – Keyword args for generative()

• loss_kwargs (dict | None (default: None)) – Keyword args for loss()

• compute_loss (default: True) – Whether to compute loss on forward pass. This adds another return value.

Return type:

tuple[Tensor, Tensor] | tuple[Tensor, Tensor, LossOutput]

abstractmethod BaseModuleClass.generative(*args, **kwargs)

Run the generative model.

This function should return the parameters associated with the likelihood of the data. This is typically written as \(p(x|z)\).

This function should return a dictionary with str keys and Tensor values.

Return type:

dict[str, Tensor | Distribution]

abstractmethod BaseModuleClass.inference(*args, **kwargs)

Run the recognition model.

In the case of variational inference, this function will perform steps related to computing variational distribution parameters. In a VAE, this will involve running data through encoder networks.

This function should return a dictionary with str keys and Tensor values.

Return type:

dict[str, Tensor | Distribution]

abstractmethod BaseModuleClass.loss(*args, **kwargs)

Compute the loss for a minibatch of data.

This function uses the outputs of the inference and generative functions to compute a loss. This many optionally include other penalty terms, which should be computed here.

This function should return an object of type LossOutput.

Return type:

LossOutput

BaseModuleClass.on_load(model, **kwargs)

Callback function run in load().

abstractmethod BaseModuleClass.sample(*args, **kwargs)

Generate samples from the learned model.