base

DeepEstimator(module, loss_fn='mse', optimizer_fn='sgd', lr=0.001, is_feature_incremental=False, device='cpu', seed=42, **kwargs)

Bases: Estimator

Abstract base class that implements basic functionality of River-compatible PyTorch wrappers.

PARAMETER	DESCRIPTION
module	Torch Module that builds the autoencoder to be wrapped. The Module should accept parameter n_features so that the returned model's input shape can be determined based on the number of features in the initial training example. TYPE: Type[Module]
loss_fn	Loss function to be used for training the wrapped model. Can be a loss function provided by torch.nn.functional or one of the following: 'mse', 'l1', 'cross_entropy', 'binary_crossentropy', 'smooth_l1', 'kl_div'. TYPE: Union[str, Callable] DEFAULT: 'mse'
optimizer_fn	Optimizer to be used for training the wrapped model. Can be an optimizer class provided by torch.optim or one of the following: "adam", "adam_w", "sgd", "rmsprop", "lbfgs". TYPE: Union[str, Callable] DEFAULT: 'sgd'
lr	Learning rate of the optimizer. TYPE: float DEFAULT: 0.001
device	Device to run the wrapped model on. Can be "cpu" or "cuda". TYPE: str DEFAULT: 'cpu'
seed	Random seed to be used for training the wrapped model. TYPE: int DEFAULT: 42
**kwargs	Parameters to be passed to the Module or the optimizer. DEFAULT: {}

clone(new_params=None, include_attributes=False)

Clones the estimator.

PARAMETER	DESCRIPTION
new_params	New parameters to be passed to the cloned estimator. TYPE: dict[Any, Any] | None DEFAULT: None
include_attributes	If True, the attributes of the estimator will be copied to the cloned estimator. This is useful when the estimator is a transformer and the attributes are the learned parameters. DEFAULT: False

RETURNS	DESCRIPTION
DeepEstimator	The cloned estimator.

draw()

Draws the wrapped model.

initialize_module(x, **kwargs)

PARAMETER	DESCRIPTION
module	The instance or class or callable to be initialized, e.g. self.module.
kwargs	The keyword arguments to initialize the instance or class. Can be an empty dict. TYPE: dict DEFAULT: {}

RETURNS	DESCRIPTION
instance	The initialized component.

DeepEstimatorInitialized(module, loss_fn='mse', optimizer_fn='sgd', lr=0.001, device='cpu', seed=42, is_feature_incremental=False, **kwargs)

Bases: Estimator

Enhances PyTorch modules with dynamic adaptability to evolving features.

The class extends the functionality of a base estimator by dynamically updating and expanding neural network layers to handle incremental changes in feature space. It supports feature set discovery, input size adjustments, weight expansion, and varied learning procedures. This makes it suitable for evolving input spaces while maintaining neural network integrity.

ATTRIBUTE	DESCRIPTION
module	The PyTorch model that serves as the backbone of this class's functionality. TYPE: Module
lr	Learning rate for model optimization. TYPE: float
loss_fn	The loss function used for computing training error. TYPE: Union[str, Callable]
loss_func	The compiled loss function produced via get_loss_fn. TYPE: Callable
optimizer	The compiled optimizer used for updating model weights. TYPE: Optimizer
optimizer_fn	The optimizer function or class used for training. TYPE: Union[str, Callable]
device	The computational device (e.g., "cpu", "cuda") used for training. TYPE: str
seed	The random seed for ensuring reproducible operations. TYPE: int
is_feature_incremental	Indicates whether the model should automatically expand based on new features. TYPE: bool
kwargs	Additional arguments passed to the model and utilities. TYPE: dict
input_layer	The input layer of the PyTorch model, determined dynamically. TYPE: Module
output_layer	The output layer of the PyTorch model, determined dynamically. TYPE: Module
observed_features	Tracks all observed input features dynamically, allowing for feature incrementation. TYPE: SortedSet

RollingDeepEstimator(module, loss_fn='mse', optimizer_fn='sgd', lr=0.001, device='cpu', seed=42, window_size=10, append_predict=False, **kwargs)

Bases: DeepEstimator

Abstract base class that implements basic functionality of River-compatible PyTorch wrappers including a rolling window to allow the model to make predictions based on multiple previous examples.

PARAMETER	DESCRIPTION
module	Torch Module that builds the autoencoder to be wrapped. The Module should accept parameter n_features so that the returned model's input shape can be determined based on the number of features in the initial training example. TYPE: Type[Module]
loss_fn	Loss function to be used for training the wrapped model. Can be a loss function provided by torch.nn.functional or one of the following: 'mse', 'l1', 'cross_entropy', 'binary_crossentropy', 'smooth_l1', 'kl_div'. TYPE: Union[str, Callable] DEFAULT: 'mse'
optimizer_fn	Optimizer to be used for training the wrapped model. Can be an optimizer class provided by torch.optim or one of the following: "adam", "adam_w", "sgd", "rmsprop", "lbfgs". TYPE: Union[str, Callable] DEFAULT: 'sgd'
lr	Learning rate of the optimizer. TYPE: float DEFAULT: 0.001
device	Device to run the wrapped model on. Can be "cpu" or "cuda". TYPE: str DEFAULT: 'cpu'
seed	Random seed to be used for training the wrapped model. TYPE: int DEFAULT: 42
window_size	Size of the rolling window used for storing previous examples. TYPE: int DEFAULT: 10
append_predict	Whether to append inputs passed for prediction to the rolling window. TYPE: bool DEFAULT: False
**kwargs	Parameters to be passed to the Module or the optimizer. DEFAULT: {}

RollingDeepEstimatorInitialized(module, loss_fn='mse', optimizer_fn='sgd', lr=0.001, device='cpu', seed=42, window_size=10, append_predict=False, **kwargs)

Bases: DeepEstimatorInitialized

RollingDeepEstimatorInitialized class for rolling window-based deep learning model estimation.

This class extends the functionality of the DeepEstimatorInitialized class to support training and prediction using a rolling window. It maintains a fixed-size deque to store a rolling window of input data. It can optionally append predictions to the input window to facilitate iterative prediction workflows. This class is designed for advanced users who need rolling window functionality in their deep learning estimation pipelines.

ATTRIBUTE	DESCRIPTION
window_size	The size of the rolling window used for training and prediction. TYPE: int
append_predict	Flag to indicate whether to append predictions into the rolling window. TYPE: bool
_x_window	A fixed-size deque object, which stores the most recent input window data. TYPE: Deque
_batch_i	The internal counter for batch index tracking during training or prediction. TYPE: int