scvi.module.base.BaseModuleClass

class scvi.module.base.BaseModuleClass

Bases: torch.nn.modules.module.Module

Abstract class for scvi-tools modules.

Attributes table

device

Methods table

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

Attributes

T_destination

BaseModuleClass.T_destination

alias of TypeVar(‘T_destination’, bound=Mapping[str, torch.Tensor])

alias of TypeVar(‘T_destination’, bound=Mapping[str, torch.Tensor]) .. autoattribute:: BaseModuleClass.T_destination device ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

BaseModuleClass.device

dump_patches

BaseModuleClass.dump_patches: bool = False

This allows better BC support for load_state_dict(). In state_dict(), the version number will be saved as in the attribute _metadata of the returned state dict, and thus pickled. _metadata is a dictionary with keys that follow the naming convention of state dict. See _load_from_state_dict on how to use this information in loading.

If new parameters/buffers are added/removed from a module, this number shall be bumped, and the module’s _load_from_state_dict method can compare the version number and do appropriate changes if the state dict is from before the change.

training

BaseModuleClass.training: bool

Methods

forward

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 | NoneOptional[dict] (default: None)

Keyword args for _get_inference_input()

get_generative_input_kwargs : dict | NoneOptional[dict] (default: None)

Keyword args for _get_generative_input()

inference_kwargs : dict | NoneOptional[dict] (default: None)

Keyword args for inference()

generative_kwargs : dict | NoneOptional[dict] (default: None)

Keyword args for generative()

loss_kwargs : dict | NoneOptional[dict] (default: None)

Keyword args for loss()

compute_loss

Whether to compute loss on forward pass. This adds another return value.

Return type

Tuple[Tensor, Tensor] | Tuple[Tensor, Tensor, LossRecorder]Union[Tuple[Tensor, Tensor], Tuple[Tensor, Tensor, LossRecorder]]

generative

abstract 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

{str: Tensor | Distribution}Dict[str, Union[Tensor, Distribution]]

inference

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

Run the inference (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

{str: Tensor | Distribution}Dict[str, Union[Tensor, Distribution]]

loss

abstract 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 LossRecorder.

Return type

LossRecorder

sample

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

Generate samples from the learned model.