Struct MicroLoRA 

pub struct MicroLoRA { /* private fields */ }

Micro-LoRA for per-request adaptation

Uses rank 1-2 for ultra-low latency updates. Forward pass: output += scale * (input @ down) @ up

Performance notes (from benchmarks):

• Rank-2 is ~5% faster than Rank-1 due to better SIMD vectorization
• Batch size 32 optimal: 0.447ms per-vector, 2,236 ops/sec throughput
• SIMD-enabled: +10% speedup over scalar

Implementations

impl MicroLoRA

pub fn new(hidden_dim: usize, rank: usize) -> Self

Create new Micro-LoRA adapter

Arguments

• hidden_dim - Model hidden dimension
• rank - LoRA rank (must be 1-2)

Panics

Panics if rank > 2

pub fn forward_scalar(&self, input: &[f32], output: &mut [f32])

Scalar forward pass (fallback)

pub fn forward(&self, input: &[f32], output: &mut [f32])

Forward pass with automatic SIMD detection

pub fn accumulate_gradient(&mut self, signal: &LearningSignal)

Accumulate gradient from learning signal

pub fn apply_accumulated(&mut self, learning_rate: f32)

Apply accumulated gradients with learning rate

pub fn reset(&mut self)

Reset adapter to initial state

pub fn rank(&self) -> usize

Get rank

pub fn hidden_dim(&self) -> usize

Get hidden dimension

pub fn param_count(&self) -> usize

Get parameter count

pub fn scale(&self) -> f32

Get scale factor

pub fn set_scale(&mut self, scale: f32)

Set scale factor

pub fn pending_updates(&self) -> usize

Get pending update count

pub fn get_weights(&self) -> (&Vec<f32>, &Vec<f32>)

Get LoRA weights for export (lora_a, lora_b)

Trait Implementations

impl Clone for MicroLoRA

fn clone(&self) -> MicroLoRA

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for MicroLoRA

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for MicroLoRA

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for MicroLoRA

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.