Struct SparseWeights 

pub struct SparseWeights {
    pub entries: Vec<(usize, f32)>,
}

Sparse lexical weight representation (SPLADE-like).

Stores non-zero (dimension_index, weight) pairs, sorted by descending absolute weight. Used for sparse retrieval ranking.

Fields

entries: Vec<(usize, f32)>

(index, weight) pairs, sorted by descending absolute weight.

Implementations

impl SparseWeights

pub fn from_dense(vec: &[f32], top_k: usize, min_weight: f32) -> Self

Build a sparse representation from a dense vector by keeping the top-k dimensions with the largest absolute values, filtered by a minimum threshold.

This is a pragmatic derivation from the dense embedding. When a native SPLADE-style sparse output is available from the model, use SparseWeights::from_entries instead.

pub fn from_entries(entries: Vec<(usize, f32)>) -> Self

Build directly from pre-computed (index, weight) entries.

pub fn len(&self) -> usize

Number of non-zero entries.

pub fn is_empty(&self) -> bool

Whether the sparse vector is empty.

pub fn dot(&self, other: &SparseWeights) -> f32

Dot product between two sparse vectors.

Efficiently computed by intersecting the sorted-by-index entries. Since entries are sorted by weight (not index), we use a HashMap-based merge for correctness.

Trait Implementations

impl Clone for SparseWeights

fn clone(&self) -> SparseWeights

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for SparseWeights

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

Formats the value using the given formatter.

impl PartialEq for SparseWeights

fn eq(&self, other: &SparseWeights) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl StructuralPartialEq for SparseWeights