Struct CsrBatch 

pub struct CsrBatch {
    pub data_raw: Option<Vec<f32>>,
    pub data_norm: Option<Vec<f32>>,
    pub indices: Vec<u32>,
    pub indptr: Vec<u32>,
    pub shape: (usize, usize),
}

A batch of cells as one CSR matrix.

The layout a training loop wants: contiguous data, indices and indptr that hand straight to scipy.sparse.csr_matrix or torch.sparse_csr_tensor without another pass. Whichever layer was not requested is None; both share one sparsity pattern, so indices and indptr are never duplicated.

Fields

data_raw: Option<Vec<f32>>

Raw counts as f32, in CSR order, when requested.

data_norm: Option<Vec<f32>>

Normalised counts as f32, in CSR order, when requested.

indices: Vec<u32>

Gene indices, one per non-zero.

indptr: Vec<u32>

Row pointers, of length shape.0 + 1.

shape: (usize, usize)

(cells in the batch, total genes).

Trait Implementations

impl Clone for CsrBatch

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for CsrBatch

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

Formats the value using the given formatter.