Struct SparseAdamW 

pub struct SparseAdamW { /* private fields */ }

Sparse AdamW optimizer

Implements AdamW optimization algorithm with decoupled weight decay for sparse tensors.

Implementations

impl SparseAdamW

pub fn new( lr: f32, beta1: f32, beta2: f32, eps: f32, weight_decay: f32, amsgrad: bool, ) -> Self

Create a new sparse AdamW optimizer

pub fn default(lr: f32) -> Self

Create AdamW optimizer with default settings

Trait Implementations

impl SparseOptimizer for SparseAdamW

fn step( &mut self, parameters: &mut [&mut CsrTensor], gradients: &[&CsrTensor], ) -> TorshResult<()>

Update sparse parameters with sparse gradients

fn lr(&self) -> f32

Get current learning rate

fn set_lr(&mut self, lr: f32)

Set learning rate

fn zero_grad(&mut self)

Zero gradients (if applicable)