Struct VeraLayer 

pub struct VeraLayer { /* private fields */ }

VeRA layer implementing Vector-based Random Matrix Adaptation.

Uses frozen random matrices A and B with trainable scaling vectors: ΔW = B @ diag(d) @ A

Where:

• A: Frozen random matrix [r, in_features] (Kaiming initialization)
• B: Frozen random matrix [out_features, r] (zero initialization or small random)
• d: Trainable scaling vector [r]
• b: Optional trainable bias vector [out_features]

Implementations

impl VeraLayer

pub fn new( in_features: usize, out_features: usize, config: VeraConfig, device: &Device, ) -> Result<VeraLayer, PeftError>

Create a new VeRA layer.

Arguments

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

Errors

Returns an error if configuration validation fails or layer construction fails.

pub fn new_with_bias( in_features: usize, out_features: usize, config: VeraConfig, device: &Device, ) -> Result<VeraLayer, PeftError>

Create a new VeRA layer with trainable bias.

Arguments

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

Errors

Returns an error if configuration validation fails or layer construction fails.

pub fn scaling_vector(&self) -> &Tensor

Get the scaling vector d.

pub fn rank(&self) -> usize

Get the rank.

Trait Implementations

impl Adapter for VeraLayer

type Config = VeraConfig

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) -> &<VeraLayer as Adapter>::Config

Get the adapter’s configuration.

impl Mergeable for VeraLayer

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 Trainable for VeraLayer

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.