Struct Linear 

pub struct Linear {
    pub weight: Tensor,
    pub lora: Option<LoRAAdapter>,
}

Bias-free linear layer y = x @ weight with an optional LoRA adapter.

Fields

weight: Tensor

Weight [in, out] (already transposed from the PyTorch [out, in]).

lora: Option<LoRAAdapter>

Optional low-rank adapter added to the output.

Implementations

impl Linear

pub fn new(weight: Tensor) -> Self

A layer without LoRA.

pub fn with_lora(weight: Tensor, lora: LoRAAdapter) -> Self

A layer with a LoRA adapter.

pub fn forward(&self, engine: &ComputeEngine, x: &Tensor) -> Tensor

Projects [tokens, in] or [batch, seq, in] to the same leading shape with out features.

pub fn params(&self) -> Vec<Tensor>

Trainable tensors (weights, then any LoRA adapters).

pub fn named_params(&self, prefix: &str) -> HashMap<String, Tensor>

Parameters keyed by Hugging Face-style names under prefix.

pub fn lora_params(&self) -> Vec<Tensor>

Only the LoRA adapter tensors (what LoRA fine-tuning updates).

Trait Implementations

impl Clone for Linear

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for Linear

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

Formats the value using the given formatter.