Struct StreamingFitter 

pub struct StreamingFitter<M> { /* private fields */ }

Feeds batches from an iterator to a PartialFit model.

This adapter collects batches from an iterator and feeds them to a model that implements PartialFit. Multiple epochs can be specified, causing the batches to be replayed multiple times for convergence.

Type Parameters

• M: The model type, which must implement PartialFit<X, Y>.

Implementations

impl<M> StreamingFitter<M>

pub fn new(model: M) -> Self

Create a new StreamingFitter wrapping the given model.

The default number of epochs is 1 (a single pass over the data).

pub fn n_epochs(self, n_epochs: usize) -> Self

Set the number of epochs (passes over the data).

Each epoch replays all batches in order. More epochs can improve convergence for online learning algorithms.

Panics

This method does not panic, but fit_batches will return early with the initial model state if n_epochs is 0.

pub fn fit_batches<X, Y, I>(self, batches: I) -> Result<M::FitResult, M::Error>

where M: PartialFit<X, Y>, M::FitResult: PartialFit<X, Y, FitResult = M::FitResult, Error = M::Error>, I: IntoIterator<Item = (X, Y)>,

Feed all batches to the model, returning the fitted result.

The batches are collected into a Vec so they can be replayed across multiple epochs. For a single epoch with a non-cloneable iterator, use fit_batches_single_epoch.

Errors

Returns the first error encountered during any partial_fit call.

pub fn fit_batches_single_epoch<X, Y, I>( self, batches: I, ) -> Result<M::FitResult, M::Error>

where M: PartialFit<X, Y>, M::FitResult: PartialFit<X, Y, FitResult = M::FitResult, Error = M::Error>, I: IntoIterator<Item = (X, Y)>,

Feed batches from a single-pass iterator to the model.

Unlike fit_batches, this method does not collect the batches, so it only supports a single epoch. The n_epochs setting is ignored.

Errors

Returns the first error encountered during any partial_fit call.