Struct SparseCSR 

pub struct SparseCSR<T: TensorElement> { /* private fields */ }

Sparse tensor in CSR (Compressed Sparse Row) format

CSR format is optimized for row-wise operations and matrix-vector multiplication. It stores:

• row_ptr: Array of size (num_rows + 1) indicating where each row starts
• col_indices: Column indices for each non-zero value
• values: Non-zero values in row-major order

Example

use torsh_tensor::sparse::SparseCSR;

// Create a 3x3 sparse matrix in CSR format
// [[1.0, 0.0, 2.0],
//  [0.0, 3.0, 0.0],
//  [4.0, 0.0, 5.0]]
let row_ptr = vec![0, 2, 3, 5];  // Row pointers
let col_indices = vec![0, 2, 1, 0, 2];  // Column indices
let values = vec![1.0, 2.0, 3.0, 4.0, 5.0];  // Values
let shape = vec![3, 3];

let sparse = SparseCSR::new(row_ptr, col_indices, values, shape).expect("CSR creation should succeed");

Implementations

impl<T: TensorElement> SparseCSR<T>

pub fn new( row_ptr: Vec<usize>, col_indices: Vec<usize>, values: Vec<T>, shape: Vec<usize>, ) -> Result<Self>

Create a new CSR sparse tensor

Arguments

• row_ptr - Row pointers (length: num_rows + 1)
• col_indices - Column indices for non-zero values
• values - Non-zero values
• shape - Shape [num_rows, num_cols]

pub fn from_coo(coo: &SparseTensor<T>) -> Result<Self>

where T: Copy,

Convert from COO format to CSR format

pub fn to_coo(&self) -> Result<SparseTensor<T>>

where T: Copy,

Convert to COO format

pub fn to_dense(&self) -> Result<Tensor<T>>

where T: Copy + Zero,

Convert to dense tensor

pub fn matvec(&self, vec: &[T]) -> Result<Vec<T>>

where T: Copy + Add<Output = T> + Mul<Output = T> + Zero,

Matrix-vector multiplication (optimized for CSR)

pub fn get_row(&self, row: usize) -> Result<(Vec<usize>, Vec<T>)>

where T: Copy,

Get a specific row as a sparse vector

pub fn nnz(&self) -> usize

Getters

pub fn shape(&self) -> &[usize]

pub fn device(&self) -> DeviceType

pub fn row_ptr(&self) -> &[usize]

pub fn col_indices(&self) -> &[usize]

pub fn values(&self) -> &[T]

Trait Implementations

impl<T: Clone + TensorElement> Clone for SparseCSR<T>

fn clone(&self) -> SparseCSR<T>

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl<T: Debug + TensorElement> Debug for SparseCSR<T>

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

Formats the value using the given formatter.