Struct LinearLayer 

pub struct LinearLayer { /* private fields */ }

Dense linear layer: y = x @ weight + bias.

Supports both graph-mode (autograd training) and inference-mode (raw tensors).

Implementations

impl LinearLayer

pub fn new( graph: &mut Graph, in_features: usize, out_features: usize, weight_init: Tensor, bias_init: Tensor, ) -> Result<Self, ModelError>

Creates a layer from explicit parameter tensors.

pub fn zero_init( graph: &mut Graph, in_features: usize, out_features: usize, ) -> Result<Self, ModelError>

Creates a zero-initialized layer.

pub fn sync_from_graph(&mut self, graph: &Graph) -> Result<(), ModelError>

Synchronizes owned tensors from the graph (e.g. after optimizer step).

pub fn forward( &self, graph: &mut Graph, input: NodeId, ) -> Result<NodeId, ModelError>

Graph-mode forward pass (for training with autograd).

pub fn forward_inference(&self, input: &Tensor) -> Result<Tensor, ModelError>

Inference-mode forward pass (no graph needed).

pub fn in_features(&self) -> usize

pub fn out_features(&self) -> usize

pub fn weight_node(&self) -> Option<NodeId>

pub fn bias_node(&self) -> Option<NodeId>

pub fn weight(&self) -> &Tensor

pub fn bias(&self) -> &Tensor

Trait Implementations

impl Clone for LinearLayer

fn clone(&self) -> LinearLayer

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for LinearLayer

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl PartialEq for LinearLayer

fn eq(&self, other: &LinearLayer) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl StructuralPartialEq for LinearLayer