Struct SparseGradient 

pub struct SparseGradient {
    pub indices: Vec<u32>,
    pub values: Vec<f32>,
    pub original_len: usize,
}

A compact sparse representation of a gradient vector.

Indices use u32 to halve storage compared with usize on 64-bit targets, which is safe for gradient lengths that fit within 4 billion elements.

Fields

indices: Vec<u32>

Positions of kept (non-zero) elements in the original flat gradient.

values: Vec<f32>

Values corresponding to each kept index.

original_len: usize

Length of the original flat gradient from which this was derived.

Implementations

impl SparseGradient

pub fn sparsity_ratio(&self) -> f64

Fraction of elements that were not kept.

A value of 1.0 means all elements were dropped; 0.0 means all were kept.

pub fn to_dense(&self) -> Vec<f32>

Reconstruct the full dense gradient vector (zeros at dropped positions).

Trait Implementations

impl Clone for SparseGradient

fn clone(&self) -> SparseGradient

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for SparseGradient

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for SparseGradient

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for SparseGradient

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

where __S: Serializer,

Serialize this value into the given Serde serializer.