Struct Sequential 

pub struct Sequential { /* private fields */ }

Sequential model that chains layers together.

Implementations

impl Sequential

pub fn new(name: &str, layers: Vec<Box<dyn Layer>>) -> Self

Create a new sequential model.

pub fn add_layer(&mut self, layer: Box<dyn Layer>)

Add a layer to the model.

pub fn forward( &self, inputs: &dyn ArrayProtocol, ) -> Result<Box<dyn ArrayProtocol>, OperationError>

Forward pass through the model.

pub fn parameters(&self) -> Vec<Box<dyn ArrayProtocol>>

Get all parameters in the model.

pub fn train(&mut self)

Set the model to training mode.

pub fn eval(&mut self)

Set the model to evaluation mode.

pub fn name(&self) -> &str

Get the model’s name.

pub fn layers(&self) -> &[Box<dyn Layer>]

Get the layers in the model.

pub fn layers_mut(&mut self) -> &mut [Box<dyn Layer>]

Get mutable access to the layers in the model — e.g. so a deserializer can restore saved parameter values in place.

pub fn backward( &self, input: &dyn ArrayProtocol, grad_output: &dyn ArrayProtocol, ) -> Result<GradientDict, CoreError>

Backward pass through the model: given the original input fed to forward and grad_output (the gradient of the loss with respect to the model’s final output — e.g. Loss::backward’s result), computes the gradient for every parameter in every layer via backpropagation. The returned dictionary is keyed "{layer_index}.{param_name}", matching Self::all_parameter_names and Self::update_parameter’s expected format.

Layer::forward doesn’t cache intermediate activations, so this first recomputes the forward pass (retaining each layer’s input) and then walks the layers in reverse, threading each layer’s grad_input to the previous layer as its grad_output.

pub fn update_parameter( &mut self, param_name: &str, gradient: &dyn ArrayProtocol, learningrate: f64, ) -> Result<(), CoreError>

Update a parameter in the model

pub fn all_parameter_names(&self) -> Vec<String>

Get all parameter names in the model with layer prefixes

pub fn all_parameters(&self) -> Vec<Box<dyn ArrayProtocol>>

Get all parameters in the model