Struct SparseAdamTable 

pub struct SparseAdamTable { /* private fields */ }

Sparse / lazy Adam optimizer over a 2-D embedding table.

Mirrors SparseAdam but operates on an Array2 with a CsrGradient. Moment state is stored flat, indexed by row * cols + col.

Implementations

impl SparseAdamTable

pub fn new(config: SparseAdamConfig) -> Self

Create a new table optimizer. State is allocated lazily on the first SparseAdamTable::step.

pub fn global_step(&self) -> usize

The number of optimizer steps performed so far.

pub fn step( &mut self, params: &mut Array2<f64>, grad: &CsrGradient, ) -> Result<(), GpuOptimError>

Perform one sparse Adam update on the embedding table in place.

Errors

Returns GpuOptimError::DimensionMismatch on any shape disagreement and propagates gradient validation errors.

Trait Implementations

impl Clone for SparseAdamTable

fn clone(&self) -> SparseAdamTable

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for SparseAdamTable

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

Formats the value using the given formatter.