Struct LoraLayer 

pub struct LoraLayer { /* private fields */ }

LoRA layer implementing low-rank adaptation.

Computes: output = base_output + (x @ A^T @ B^T) * scaling

Implementations

impl LoraLayer

pub fn new( in_features: usize, out_features: usize, config: LoraConfig, vb: VarBuilderArgs<'_, Box<dyn SimpleBackend + '_>>, ) -> Result<LoraLayer, PeftError>

Create a new LoRA layer.

Arguments

• in_features - Input dimension
• out_features - Output dimension
• config - LoRA configuration
• vb - Variable builder for weight initialization

Errors

Returns error if configuration is invalid or weight initialization fails.

pub fn new_with_zeros( in_features: usize, out_features: usize, config: LoraConfig, device: &Device, ) -> Result<LoraLayer, PeftError>

Create a new LoRA layer with zeros initialization for B.

Arguments

• in_features - Input dimension
• out_features - Output dimension
• config - LoRA configuration
• device - Device to create tensors on

Errors

Returns error if configuration is invalid or tensor initialization fails.

pub fn scaling(&self) -> f64

Get the scaling factor.

pub fn rank(&self) -> usize

Get the rank.

pub fn weights(&self) -> (&Tensor, &Tensor)

Get the LoRA A and B weight tensors.

Returns (lora_a, lora_b) where:

• lora_a has shape [r, in_features]
• lora_b has shape [out_features, r]

Trait Implementations

impl Adapter for LoraLayer

type Config = LoraConfig

The configuration type for this adapter.

fn forward( &self, input: &Tensor, base_output: Option<&Tensor>, ) -> Result<Tensor, PeftError>

Forward pass applying the adapter transformation.

fn num_parameters(&self) -> usize

Get the number of trainable parameters.

fn config(&self) -> &<LoraLayer as Adapter>::Config

Get the adapter’s configuration.

impl Mergeable for LoraLayer

fn merge(&self, base_weight: &Tensor) -> Result<Tensor, PeftError>

Merge adapter weights into base model weights.

fn unmerge(&self, merged_weight: &Tensor) -> Result<Tensor, PeftError>

Unmerge adapter weights from merged weights.

impl SaveLoad for LoraLayer

fn state_dict(&self) -> Result<HashMap<String, Tensor>, PeftError>

Get all adapter tensors as a map of name -> tensor.

fn load_state_dict( &mut self, state_dict: HashMap<String, Tensor>, ) -> Result<(), PeftError>

Load adapter tensors from a state dict.

impl Trainable for LoraLayer

fn register_parameters( &self, _var_map: &mut VarMap, _prefix: &str, ) -> Result<(), PeftError>

Register trainable parameters with the variable map.

fn freeze(&mut self)

Freeze all adapter parameters (disable gradients).

fn unfreeze(&mut self)

Unfreeze all adapter parameters (enable gradients).

fn is_frozen(&self) -> bool

Check if the adapter is frozen.